A lot of new legal businesses now call themselves AI-native, but they don’t offer the same service or operate in much the same way. The label covers Garfield, which uses AI to run small debt claims in England, and Crosby, which reviews commercial contracts for technology companies. It covers Eudia and Norm, two technology businesses that have built regulated law firms around their platforms, while some traditional firms use the same language for internal tools that help their lawyers work faster. They may all point toward the future of legal services, but they aren’t versions of the same model, which is why it remains surprisingly hard to say what an AI-native law firm actually is.

The most careful definition I have seen comes from a piece Shreya Vajpei and Saranya Mishra published on the IBA’s site late last year. An AI-native firm, they argue, uses AI rather than a pyramid of lawyers as its main method of delivering legal services, and was built that way from the beginning rather than adapted later. That’s a good start because it focuses on what produces the work and how the firm was designed. I would only pull the idea apart. A firm calling itself AI-native has made at least five separate choices, and most of the businesses using the label have made some of them but not others.

The first is about production; what happens by default when a matter arrives. In a traditional firm, a lawyer opens a blank page or, more likely, the last version of the same or similar document, and the system helps. In a machine-first firm, the system produces the first version and the lawyer steps in where needed. What makes that possible is usually less exciting than a better model. Someone has taken a type of matter, broken it into a sequence of steps, defined the input and output at each stage and decided where a lawyer must intervene. A debt claim becomes a chaser, then a letter before action, then a claim form, particulars and a directions questionnaire. Building that sequence is design work as much as technology work, and it explains why the clearest examples sit in narrow areas where the rules and process are stable.

A firm that buys a general assistant and gives every lawyer a licence has made a different choice. The tool may save a great deal of time, but the lawyer still decides what happens next and the work still moves through the lawyer. Capacity has improved; the production model hasn’t really changed. The actual test is whether the firm can describe a matter as a workflow that the system can run. Without a defined workflow, AI remains a tool used by lawyers rather than the system that runs the work.

The second choice is staffing, and it is largely independent of the first. Many firms described as AI-native use a flat group of experienced lawyers rather than a large associate pyramid. Senior associates from large firms, former general counsel and well-known partners who lend credibility. This often gets presented as a result of AI, although firms have used the model without much technology for years. Pierson Ferdinand now has almost 300 partners and virtually no associates in the United States, following a model FisherBroyles developed long before anyone had used a large language model. AI makes a flat structure easier to sustain because experienced lawyers need less junior support, but it didn’t invent the structure, and a flat firm isn’t necessarily machine-first.

The third choice is price and, underneath it, risk. Most of these firms sell fixed fees, often as part of a broader argument against the billable hour, but the more useful way to understand the price is as a transfer of scope risk. Under an hourly model, the client pays when the matter takes three times longer than expected. Under a fixed fee, the firm absorbs that cost. The model works well when the service is narrow and the matters look alike, which is why contract-focused firms begin with NDAs and MSAs for technology companies whose paper tends to repeat. As the client base widens, the difficult cases become harder to predict. That doesn’t undermine the model; it simply means the limit is set by how well the firm can price those cases, not by how quickly the software can draft.

The fourth choice is ownership. A conventional partnership distributes most of its profit to its lawyers each year, holds little capital and has limited enterprise value because its main assets can leave. The current wave of structures creates something investors can own. A technology company sitting alongside a lawyer-owned firm, an Arizona alternative business structure that permits nonlawyer ownership, or a company that raises capital to acquire legal service providers. These aren’t technology decisions, but they change how the business behaves. A partnership tends to focus on what it can distribute this year, while a venture-backed company cares more about growth and future value. That difference affects which clients it takes, how it prices the work and how much it is willing to spend building the system.

The fifth choice is measurement, which may be the least obvious and most useful. These firms can usually tell you how long a matter takes now, what it took before, how often a lawyer has to intervene and what each completed matter costs. Their economics depend on knowing. Most established firms can tell you who has access to an AI tool, but many can’t say who has changed how they work or what the change has done to delivery time. Without those numbers, decisions about capacity, hiring and price rest on little more than a general sense that the firm is becoming more efficient.

Across all five choices, the one thing that has barely moved is accountability. The regulatory story is much less dramatic than the technology. LawFairy, a rule-based immigration platform, is regulated as a law firm under the same framework as other firms. Garfield was authorised with safeguards, including a restriction on proposing case law, because the SRA considered the risk too high. In CILEX v Mazur, the Court of Appeal confirmed that unauthorised people may carry out litigation tasks under proper supervision, provided an authorised individual retains control and responsibility. That wasn’t an AI case, but the principle carries across; the work may be delegated, while responsibility stays with an identifiable professional whose insurance and career sit behind it. The cost of that supervision won’t fall with volume as easily as software costs do, so accountability may set the practical limit even when the technology could go further.

So here’s the definition I would suggest. It works better as a matter of degree than as a category. A firm is AI-native to the extent that it has broken its work into workflows a system can run, and built its staffing, pricing, ownership and measurement around those workflows rather than adjusting them after the fact. Being AI-native isn’t a badge, and a firm can sit in a different place on each of the five axes. A firm can price work like a native provider while producing it like a traditional one, which is probably the most common combination and, over time, the hardest to sustain. The point of the framework is not to decide who earns the label, but to make firms explain what they have actually changed.

The starting point is defining the work, not buying the software. These firms can run machine-first processes because someone drew a hard line around a category of matters. Debt claims below £10,000, immigration applications that follow a stable set of rules or short commercial contracts reviewed against a clear playbook. Once the work is defined, the firm can decide where people are needed, what price it can promise, how much capital the system requires and what it needs to measure.

Traditional firms aren’t shut out. New entrants can start without old billing habits or partner expectations, but BigLaw already has the clients, brand, insurance, knowledge base and lawyers who know where the standard rule stops working. Those assets are difficult to recreate. The Financial Times reported in May that Kirkland & Ellis had set aside $500 million to build its own AI platform, which is a useful reminder that “native” can’t sensibly mean “founded after ChatGPT.” If an established firm rebuilds a practice around its own workflows, it may end up far more native than a startup using generic tools and billing every lawyer-hour.

The harder part for BigLaw is changing the incentives. Partners are paid for originations, hours and the strength of their personal franchise, so turning their judgment into a firm-owned workflow can feel like giving away what makes them valuable. The likely result is that two models will live inside the same firm. Bespoke advice and difficult negotiations will remain partner-led, while recurring work moves through a separate system with its own team and pricing. That can work, but not if the new unit remains a side business while the partnership continues rewarding everyone for feeding the hourly model.

So when someone calls a firm AI-native, the useful questions are fairly simple. What produces the work when a matter arrives? How is the firm staffed? Who bears the risk when the scope expands? Who owns the business and the system behind it? What does the firm measure? If the answers are still lawyers using a general assistant, more associates as volume grows and more hours on the bill, the firm has adopted AI without changing its model (and that’s ok). The machine doesn’t need to replace the lawyer for the institution to change. The real shift begins when a firm can say exactly which parts of its work now scale without another lawyer and show the numbers to prove it.

Last week, the Financial Times reported that Paul Weiss, Quinn Emanuel and Proskauer have all spoken with private equity groups or bankers about what it might look like to take outside capital. White & Case has senior lawyers studying the model, while McDermott Will & Schulte is continuing work it began last year. This doesn’t mean that a deal is around the corner. The conversations are preliminary, no large US firm has announced a transaction, and Paul Weiss has been especially careful to play down its involvement. It says it listened to a couple of pitches at the request of firms it already does business with, held no follow-up meetings and is not pursuing the idea. Fair enough. But the names are hard to ignore. Outside investment is no longer something being discussed only around personal injury firms, regional practices and experimental legal platforms. However tentatively, it is now being discussed inside some of the most profitable law firms in the world.

When I wrote Private Equity Is Coming for Law Firms, the central idea was that investors did not need to own the legal practice itself. They could own the business around it; the technology, brand, data, real estate, marketing operation and nonlawyer staff, then charge the lawyer-owned firm for using that platform. The live deals were mostly happening well below BigLaw, which made it easy to treat the model as a niche workaround for firms with weak systems or founders looking for an exit. The recent conversations do not prove that BigLaw will adopt it, and certainly do not prove that it should. They show that the structure is now credible enough for elite firms to spend time on it. That is a meaningful step, especially because these firms have far less obvious need for the money.

We now have a fairly good sense of how the structure would work. The partners keep a lawyer-owned entity that employs the lawyers, accepts clients and provides legal advice. A separate management services organization, usually called an MSO, owns or manages some combination of the firm’s nonlegal assets and operations. The law firm enters into a long-term services agreement with the MSO, and the investor buys an interest in that MSO rather than in the practice. The services fee supplies the investor’s return. In most jurisdictions, the agreement has to preserve the lawyers’ control over professional judgment and avoid sharing legal fees with nonlawyers. That usually pushes the economics toward fixed fees, cost-plus arrangements or separately priced services supported by fair-market-value analysis, rather than a simple percentage of firm revenue or profits.

On paper, the structure looks fairly straightforward. The investor owns the services business, while the lawyers own the law firm. In practice, I am not so sure. Technology, staffing, data, firm knowledge, pricing support and practice management may be described as administrative, but they sit close to the way legal work gets done. An MSO does not have to tell a lawyer how to advise a client to influence the firm. It can shape which systems get funded, how much support a practice receives, where the firm opens an office and how aggressively it pushes for growth.

So, why would Paul Weiss, Quinn Emanuel or Proskauer sell expensive equity in the first place? These are hardly businesses starved of cash. Paul Weiss reported $2.63 billion in revenue and $7.51 million in average profit per equity partner for 2024. Quinn Emanuel reached roughly $2.8 billion in revenue and $9.5 million in profit per equity partner in 2025, while Proskauer reported $1.58 billion in revenue and more than $5 million per equity partner. A firm with numbers like these can pay for an AI team, upgrade its systems or open another office from cash flow, and it can borrow at a much lower cost than selling a permanent claim on its economics.

That is why the usual explanation, that firms need money for AI and talent, only goes so far. Outside equity can turn future earnings into cash today. It lets firms invest without reducing current partner distributions, provides liquidity to existing partners and creates equity that can be used to attract and retain talent. The firm gets permanent capital and a new compensation tool, while existing partners get something law firm partnerships rarely offer: the chance to put a market value on the institution they helped build and sell part of it before they retire.

There is a good case for that. Partners often contribute capital when they join and get it back when they leave, but they do not own a transferable share of the firm’s goodwill. They may spend years building the brand, training lawyers and funding new systems, only to retire without sharing in the value they helped create. That encourages firms to maximize current profits and leave long-term investment to the next generation. Investors would say they bring more than money. They bring operating experience, deadlines and a willingness to fund projects that may take years to pay off. Partnerships are often slow to make those investments, and outside capital could change that while giving partners a reason to build value beyond the current year.

Talent may be the more compelling argument. The market has become very good at pricing individual partners, often through guarantees far above what the rest of the partnership earns. Firms can respond with more cash, but rivals can match it. Equity in an MSO could vest over time and grow with the firm, giving rainmakers a reason to stay and help build the platform. As recent departures show, firms need to offer valuable assets that partners can only keep by staying together.

Whether MSO equity can become that asset is another matter. A law firm’s revenue may look recurring, but much of it still follows individual partners. Clients do not belong to the MSO, lawyers generally cannot be bound by noncompetes, conflicts can make valuable work disappear and entire practice groups can leave with the revenue expected to support the services fee. The investor may own the brand, systems and office lease, but not the people who make them valuable. The platform may be legally separate from the practice. Its value is much harder to separate.

It also creates the possibility of a fairly awkward cycle. An investor-backed firm uses MSO equity to retain a group of rainmakers, but a rival offers enough cash to lure them away. The firm loses their revenue but remains bound by the long-term services agreement, while the MSO still expects a return. The remaining partners must then replace the lost business while continuing to pay the MSO. Outside capital may help firms compete for talent, but it will not make partners less mobile and could make their departure more costly.

Client confidentiality adds another complication. Investors will want enough financial and operating information to monitor what they own, while clients may be far less relaxed about an outside organization having access to matters, billing data or the systems that hold their documents.

The hardest issue, at least as I see it, is who gets the value and who pays for it over time. Current equity partners control the vote and would likely receive the sale proceeds, while some may reinvest in the MSO. Younger partners, nonequity partners and associates may receive none of that initial cash, yet eventually find themselves in a firm committed to paying the management fee that supports the investor’s return. One generation could monetize the firm while the next inherits the contract. That does not make every deal unfair. A firm can reserve meaningful MSO equity for future partners, require current partners to reinvest, use the money for real investment and preserve buyback or termination rights. It can also make sure the management fee does not slowly shift value away from the partnership. But none of this happens automatically. The same partners receiving the payout will negotiate those protections.

That is why the use of proceeds will matter so much in the first BigLaw deal. Calling the money “growth capital” means little if partner distributions increase by a similar amount. If most of the capital stays in the business and future partners share in the upside, the firm can make a credible case that it is investing for the long term. If the money mostly goes to current partners while the MSO keeps the assets and fee stream, it is really a partner liquidity event.

This is still mostly speculation, and much will depend on how any deal is structured. Which assets move to the MSO, and how are they valued? Does the firm’s brand move with them? Who owns technology developed after closing, and what access does the investor have to the data behind it? The agreement would also need to address how the management fee changes, what happens if a major practice group leaves and whether the firm can walk away without paying a prohibitive amount. The firm would also need a say over any sale of the MSO. A long-term agreement with a familiar minority investor may feel very different once that investor sells to a more aggressive owner.

Then there is the exit. A conventional fund eventually needs to sell the MSO, recapitalize it or arrange a buyback. Any of those options would force the firm into another transaction involving a services provider tied to its daily operations. Permanent capital, a minority investment or a debt-like instrument may fit better, and the FT reports that several structures are being discussed. But the less control, revenue participation and exit certainty the investor gets, the less likely it is to pay the valuation partners will expect. The structure may work legally long before it makes economic sense for both sides.

These conversations may go nowhere, BigLaw firms listen to bankers, test valuations and study competitors all the time, and Paul Weiss has been unusually clear that it is not currently pursuing a deal. We should take that pushback seriously. We should also notice that the subject has moved from conference panels into meetings with the leaders of elite firms. Once a partnership starts asking what its nonlegal business is worth, how that value could be divided and what equity might do in the fight for talent, it has already begun to think about itself differently.

I can see why firms are looking. The market is changing faster than their partnership model, and outside capital could help them invest and compete. But the fact that a deal is possible does not mean it makes sense. The first BigLaw transaction should be judged less by the size of the cheque than by what it leaves future partners. If they inherit a stronger firm and share in its value, outside capital may earn its place. If they only inherit the obligation to generate returns for an investor and liquidity for earlier partners, BigLaw will have found a very expensive way to fund its succession.

On July 28, Gibson Dunn announced that William Savitt was joining the firm as co-chair of its litigation practice, and that five Wachtell Lipton litigation partners were coming with him. It takes a moment for the weight of that sentence to land, because Savitt was not a Wachtell partner in the ordinary sense of the word. He was co-chair of the firm’s executive committee, one of the two people handed the institution when Ed Herlihy and Daniel Neff stepped back in late 2023, the lawyer who forced Elon Musk to close the Twitter acquisition in 2022 and who, just this past May, won the jury verdict defending OpenAI’s restructuring against Musk in Oakland. He had spent about three decades at the firm, and in the usual grammar of law firm departures, people in that seat are not the ones who leave. They are the reason other people stay.

The move follows a string of departures that would have sounded strange at Wachtell even a few years ago. Since early 2025, the firm has lost at least nine other partners to competitors: John Sobolewski and Zach Podolsky to Latham, then Emily Johnson, Mark Stagliano and Erica Aho to Latham as well; Joshua Feltman, who chaired the restructuring practice, to Kirkland; Alison Preiss to Simpson Thacher; Noah Yavitz to Ropes & Gray; Viktor Sapezhnikov to DLA Piper. American Lawyer data shows the partnership shrank by more than 8% in 2025, the steepest contraction among the elite New York firms. Bloomberg Law has even started referring to Wachtell's "free agency era," a phrase that would have been almost unintelligible not long ago.

Anyone reaching for the easy version of the story should first sit with an inconvenient fact. Wachtell is having the best year in its history. Profits per equity partner reached $12.15 million in 2025, up 34.5%, the first time any law firm has cleared twelve million, and the firm remains first in the American Lawyer rankings, where it has sat for most of three decades. It advised on roughly $621 billion of deals last year from a single office with fewer than 300 lawyers, and when reporters called about Savitt, the firm’s statement said it was performing at its highest level across every metric, which happens to be true.

So the departures and the record results have to be held in the same hand, and that is what makes this worth writing about. This is not an essay about a firm in decline. The evidence simply doesn't support that conclusion. It is an essay about a compensation model, the classic lockstep partnership, and about whether the conditions that made it work still exist. Wachtell matters here because it is not merely an example of lockstep. It is the strongest case that was ever made for it, sustained longer and executed better than anywhere else, which means that whatever is happening at Wachtell tells us more about the model than events at any other firm possibly could.

It helps to start with what lockstep actually means, because the term gets used surprisingly loosely. In a pure lockstep system partners are paid by seniority and by nothing else. Not by billings, origination credit, hours or who happened to land the client. The formula described in the Harvard Business School case on Wachtell allotted new partners 33 points, rising with tenure to 100, with the founders at 125, and that was the entire system. A partner's income depended on the firm's profits and the year they became partner, not on what they accomplished that particular year.

It’s an unusual arrangement in a business that increasingly measures everything. Lockstep deliberately refuses to put a price on the partner sitting in front of it. Marty Lipton’s thirty-five principles captured the philosophy plainly: compensation followed an absolute lockstep, clients belonged to the firm rather than individual partners, and there was no partnership agreement, only a handshake among friends. Under that arrangement a partner has no reason to hoard a client relationship, no reason to fight over credit, and no reason to steer a matter toward their own file rather than toward the colleague best equipped for it. Daniel Neff once acknowledged that Lipton was significantly underpaid relative to his market value, and described the effect this had on everyone else quite deliberately. If the senior partner who built the firm accepts being underpaid in order to maximize the institution’s chances of lasting, the people below him understand how they are expected to behave.

Economically, lockstep did two things at once. It deferred compensation, asking partners to accept less than the market might pay during some years in exchange for more over the course of a career. And it pooled risk. Your income depended on the firm’s success rather than your own practice, which freed partners to do the work that best served the institution rather than the work that best served themselves.

What Wachtell got from this was very real. The firm's task-force model, partners from different specialties in the same room solving the same problem, only works because nobody's compensation depends on whose name ends up on the matter. As Daniel Neff put it, the victories belong to Wachtell Lipton rather than to any individual lawyer. The firm also stayed intentionally small. About two associates per partner when competitors had four to six. One office. Few retainer relationships. A willingness to decline work. Fees based on value rather than time. That is how Kraft ended up paying $20 million for two weeks of work in 1988 and considered it money well spent.

Beneath all of this was something so embedded in the firm that it was easy to take for granted. Wachtell simply did not hire lateral partners. In its first fifty years, it admitted exactly one partner from outside, and almost nobody left either. Lipton used to say that, to his knowledge, no partner had ever left the firm to join another law firm for more money. I used to think that was a story about loyalty. Now I think it was really a story about markets. Nobody could put a credible price on a Wachtell partner because nobody could separate the partner from the institution. The clients belonged to the firm, the work was collective, and the value sat inside the whole rather than in any one person.

The usual explanation is that lawyers have become less loyal, younger partners think differently, or firm culture just isn't what it used to be. I'm not sure that's where the story starts. It seems to me the market changed first. Over the past decade, the lateral market has become remarkably good at putting a price on individual partners. And once you can price something, you can negotiate over it.

You can almost see when the market changed. In 2018, Kirkland hired Sandra Goldstein from Cravath on a package reported at $11 million a year for five years. The figure was striking, but the more important point was that it became public. For the first time, there was a visible market price for one of New York’s elite partners.

A few years later, Paul Weiss raided Kirkland’s London private equity team with reported packages of up to $20 million a year, and Kirkland responded in kind. By 2025, the Am Law 200 recorded more than 3,000 lateral partner moves, while Kirkland alone hired 116 partners.

At some point this stopped looking like opportunistic hiring and became part of the business model. The largest firms now recruit partners the way companies pursue acquisitions. They identify targets, underwrite the economics and make offers they believe will generate a return.

Then the market reached Wachtell itself. When Joshua Feltman moved to Kirkland earlier this year, the reported package was around $80 million guaranteed over three years. The exact figure deserves caution, since these numbers pass through recruiters, competitors and plenty of speculation before they become public. But whether it was $70 million or $90 million is beside the point. It was comfortably above what the average Wachtell partner earned in the most profitable year any law firm has ever had.

Think about what a guarantee like that does to a lockstep system. Lockstep pays partners according to the firm’s internal bargain, not what each of them could earn in the market that year. Some partners contribute more than they receive. Others receive more than they contribute. Over the course of a career, those differences are supposed to balance out. A large lateral offer changes the calculation. It puts a visible market price on one partner. Suddenly the gap between what the partnership pays and what the market would pay is no longer theoretical. It’s sitting in front of you in a term sheet. The offer doesn’t even have to be accepted. Once you know what someone else is willing to pay, staying becomes an active choice rather than the default.

There is another detail in the Savitt move that deserves attention. Gibson Dunn’s profits per partner last year were about $8.9 million, well below Wachtell’s $12.15 million. On paper, the co-chair of Wachtell’s executive committee left the most profitable law firm in the world for a less profitable one. Of course, averages no longer tell the story. At the top of the market, partners are increasingly valued individually rather than collectively. Once that happens, Wachtell’s profitability becomes less important than what a particular partner can earn somewhere else.

That said, I don't think the story is quite that simple.

First, today's guarantee market exists because we've had an extraordinary decade for private capital and high-end transactional work. We've seen hiring booms before, in 2007 and again in 2021, and they looked much less compelling once the cycle turned. A three-year guarantee feels very different when deal flow is cut in half.

There is also a reason firms have always been cautious about paying huge premiums for laterals. Practices don’t always travel as well as recruiters hope. Clients stay behind, split their work or simply prove more attached to the institution than to the individual partner. That was always part of the thinking around Wachtell. The firm’s value was supposed to come from the institution, not from a collection of portable books of business. Whether a group of litigators who built their careers inside that model will generate the same economics elsewhere is something Gibson Dunn is now betting on.

And then there is the other side of the market. It is easy to assume the boutiques are the ones under pressure because they are losing partners. But the firms doing the hiring are running an experiment of their own. Nobody knows how resilient that model will be through a prolonged downturn. The more fragile model may turn out not to be the boutique, but the platform trying to hold together hundreds of individually priced stars.

Still, compensation isn't the whole story. The work has changed too.

Wachtell was built for a particular kind of mandate: high-stakes, board-level matters that arrive without warning, are resolved quickly and are billed on value rather than volume. For that kind of work, the boutique model remains incredibly effective. Wachtell’s deal flow is proof enough of that. But a growing share of the legal market looks different today. Private capital has created permanent, high-volume deal pipelines. Restructurings can stretch over years and cut across finance, tax, litigation and regulatory teams. National security and antitrust reviews can turn a transaction into a months-long campaign spanning multiple jurisdictions. Those mandates reward depth too, but they also reward breadth, geography and the ability to field large multidisciplinary teams. In other words, they reward platforms. It may not be a coincidence that Kirkland and Latham, the two firms that have been most aggressive in hiring Wachtell partners, are also the two firms that have built exactly those kinds of platforms.

The case for Wachtell is still a strong one. Even after the recent departures, we’re talking about roughly fifteen partners over eighteen months from a partnership of around eighty that continues to promote new partners every year. At firms like Kirkland or Latham, that level of turnover would barely make the news. Wachtell still retains the overwhelming majority of the lawyers who produced the most profitable year any law firm has ever had.

People have been predicting the end of the Wachtell model for decades. They said it couldn’t survive Marty Lipton’s generation. They said it couldn’t survive recessions. They said it couldn’t survive succession. They said it couldn’t survive a legal market built around giant global firms. So far, every one of those predictions has been wrong. It is entirely possible that we look back on the departures of 2025 and 2026 the same way. A handful of individual decisions made during an unusually aggressive lateral market, not evidence that the model itself had stopped working.

But even under the most charitable reading, one thing has changed. Lipton’s old claim that no partner had ever left for more money was never really about loyalty. It was about the firm’s moat. For decades, there wasn’t a meaningful market for Wachtell partners because their value was so closely tied to the institution itself. Today there is. The moat was never that partners couldn’t leave. It was that nobody could tell them, with any confidence, what leaving was worth. Now they can.

Wachtell isn’t unique in this respect. In many ways, this is the latest chapter in a much longer story. For most of the twentieth century, the strongest professional partnerships held people together because the institution offered something the market couldn’t easily price. Careers were built over decades, rewards were deferred and there simply wasn’t much of an outside market for individual partners.

Investment banks changed. Consulting firms changed. Accounting firms changed too, each in its own way. As the market became better at valuing individual talent, those institutions gradually adapted. Some paid their stars more. Others watched those stars leave.

Law held onto the old model longer than almost anyone else. Lockstep held onto it longer than the rest of the legal market. And Wachtell held onto it longer than anyone.

That’s why these departures matter. Not because they prove the model has stopped working, but because they show it operating in a market it was never designed for. The things that made Wachtell exceptional, shared clients, genuine collaboration, very little internal politics, and partners who really do think of themselves as one firm, are still real. Clients clearly value them. The question is whether those advantages remain large enough to outweigh a market that has become much better at putting a price on individual partners.

If you’re trying to understand whether a lockstep firm is healthy, don’t start by counting departures. Firms have always lost partners, and they always will. The more revealing question is what the firm does to persuade people to stay. Lockstep works only as long as the partnership creates more value together than its partners could create separately, even after taking into account what the market is willing to pay for individual stars. In many firms, that is still true. It just can no longer be taken for granted.

Willkie became one of the first firms to sign a direct partnership with OpenAI, Clifford Chance turned 400,000 documents into a proprietary knowledge asset, Arnold & Porter created a C-suite role for AI, and CMS paid up for one of the UK’s best-known AI lawyers. Meanwhile, the first real post-merger numbers landed: HSF Kramer and DLA Piper both reported results that will be studied closely by every firm still deciding whether scale is a strategy.

Lead Stories

Willkie signs with OpenAI, and the middle layer of legaltech should be nervous

Willkie Farr & Gallagher announced a direct partnership with OpenAI on July 22, becoming one of the first law firms to work with the lab rather than through an intermediary. The arrangement runs through Willkie Works, the firm’s internal AI organization: firm programmers get access to OpenAI’s Codex development tools, ChatGPT Enterprise rolls out firmwide, and OpenAI’s frontier models will power the firm’s proprietary platforms, including its Wendell Intelligence research and drafting system.

Willkie is betting that the durable advantage lies in owning the platform layer, not renting it. Its chief digital officer was explicit: the winners will combine frontier models with proprietary workflows and governance, not merely buy access.

OpenAI hired Ironclad co-founder Jason Boehmig in June to lead a new legal vertical, and it already works with Eudia on US government legal teams. The labs are coming into legal directly. Firms that build on top of them now are choosing a side in a market where their vendor may become their competitor’s platform, or their own competitor. Freshfields made the same wager with Anthropic in April. The unanswered question is what exclusivity, if any, these deals carry. If the answer is none, the advantage is execution speed.

The merger bills come due: HSF Kramer and DLA Piper show their numbers

Results season delivered the first hard evidence on two of the market’s biggest bets. HSF Kramer, reporting for the first time since the Herbert Smith Freehills and Kramer Levin merger, posted revenue of $2.4bn and profit per equity partner of $2.1m, up 7% on HSF’s pre-merger figure. DLA Piper closed its final year as a Swiss verein with revenue up 8.4% to $4.6bn and PEP up nearly 20% to $4.1m, before its new global LLP structure took effect. CMS pushed global revenue close to €2.2bn.

Why this matters: PEP is the price of talent. A transatlantic merger that dilutes partner profits bleeds laterals to competitors within two cycles; one that grows them buys credibility for further consolidation. HSF Kramer’s 7% uplift is a defensible first print, though a single year proves little about integration, and headline PEP says nothing about dispersion between London, New York, and the network. DLA’s conversion from verein to unified LLP is the more consequential move. Vereins made global scale cheap by avoiding profit-sharing and shared liability. Collapsing that structure is a statement that clients, and lateral candidates, now discount the loose federation model.

Two AI races, often confused: advising on it and running on it

Three announcements this week, easy to lump together, are actually two different competitions. On the client-facing side, CMS hired John Buyers, Osborne Clarke’s international AI chief, into its London technology practice on July 19, its third AI or digital-assets hire in three months. That is a bet on AI as a practice area: regulation, procurement, and transactions in a market where genuinely experienced AI lawyers are scarce and price discovery on their compensation is just beginning.

On the operational side, Arnold & Porter appointed Roger Maeda as its first chief artificial intelligence officer on July 21, and Clifford Chance disclosed it has built a 400,000-document AI knowledge library with Epiq and Microsoft. The Clifford Chance move is the one to study. A firm’s precedent bank was always its real intellectual capital; structuring it for machine retrieval converts an archive into a compounding asset that new tools can be pointed at.

Other Notable Moves

• Paul Hastings added an international arbitration partner duo in Paris from White & Case. The firm keeps buying credentialed teams at pace; earlier this month it took Cahill’s private credit co-leader in New York.

• Microsoft is rolling out Harvey across its global legal team, giving the legal AI company one of its most significant in-house deployments to date. Microsoft’s Corporate, External, and Legal Affairs team will use the platform to streamline legal workflows and improve efficiency across the organization.

• Orrick has added a three-lawyer infrastructure finance team from Willkie Farr & Gallagher in Paris, strengthening its project finance practice in the French capital. Antoine Bouzanquet joins as head of French project finance alongside two associates, bringing Orrick’s dedicated Paris project finance team to eight lawyers.

• Alius Law hired veteran aviation disputes partner Richard Mumford from Clyde & Co in London. The latest in a steady stream of senior disputes departures from Clyde & Co this year, which is starting to look less like churn and more like a pattern.

• Rebaza, Alcázar & De Las Casas brought in tax partner Enrique Pintado in Lima, folding in the boutique he founded in 2024. Pintado ran the national legal office of Peru’s tax authority Sunat for seven years; hiring the former referee is a classic play for a contentious tax bench.

• DANIEL Law named two IP partners in Brazil, including patent litigator Juliana Castelo Branco from rival Licks Attorneys, a direct raid between Brazil’s two dominant patent-litigation shops.

What to Watch

• AI spending is shifting from procurement to ownership. Willkie building on OpenAI directly, Clifford Chance structuring its precedent bank as a data asset, Arnold & Porter putting AI in the C-suite. The first phase of legal AI was buying licenses. The second is building balance-sheet capability, and it favors firms with capital, engineering hires, and governance discipline. The risk runs both directions, firms that build may be building on sand if the labs commoditize their work, and firms that wait may find the compounding advantages of proprietary data are already gone.

• Results season continues. The Magic Circle and remaining US-UK combinations report over the coming weeks. Watch PEP dispersion and any commentary on integration costs; HSF Kramer's 7% has set the bar for merger apologetics.

• Specialist benches are the current LatAm playbook. This week’s moves in Peru, Chile and Brazil were all depth plays. Contentious tax, patent litigation, insurance disputes. After two years of cross-border merger activity led by the Spanish firms, the domestic elite is responding by getting harder to replicate in high-margin specialist work rather than by matching footprint.

In case you missed it, here is this week’s essay:

Hope you have a great weekend.

RWI deals have moved more and more in favor of sellers. Sellers barely stand behind their representations anymore. The indemnity is often gone or capped at a small amount, the escrow has largely disappeared, and if something goes wrong after closing, everyone knows the buyer will usually look to the insurer. But the seller still gives the reps. It signs the purchase agreement, prepares the disclosure schedules and brings the reps down at closing.

Synthetic reps take that one step further. The reps move out of the purchase agreement and into the RWI policy itself. The buyer negotiates them with the insurer, and the seller never actually makes the representations about the business it is selling. A few years ago, that was unusual. Now it is becoming a real part of the RWI market.

The numbers help explain how we got here. In the middle market, RWI premiums typically run around 2% to 3% of the limit purchased, retentions have fallen to around 0.5% to 1% of deal value, and buyers now pay the premium in 83% of deals, up from 45% in 2018. Underwriters also tend to price deals about the same whether the seller keeps some indemnity exposure or none at all. So a true walkaway deal, which was once something sellers had to push for, has become fairly standard in insured sponsor transactions. Once the insurer was taking most of the risk behind the reps anyway, the next question was pretty obvious. Why spend weeks negotiating representations from a seller that will have little or no liability if any of them turn out to be wrong?

A synthetic deal can take a few different forms. Sometimes the entire rep package is synthetic. In other deals, the seller still gives fundamental reps, such as title and authority, while the policy provides the business reps. Synthetic tax indemnities have also become more common for sellers that do not want to retain tax exposure. Some insurers offer a pre-agreed package of reps, which makes the process faster and more predictable but can produce narrower coverage. Others let the buyer draft the reps and negotiate them with the underwriting team, much as counsel would negotiate a traditional rep package with the seller. That takes more time, but it can produce broader coverage if the buyer knows where to push.

The bigger difference is what happens to disclosure. There is usually no traditional disclosure letter or set of schedules prepared against the synthetic reps. Instead, the data room, diligence reports and underwriting Q&A have to do much more of that work. The rep package itself may be built around what the buyer actually diligenced, with knowledge qualifiers tied to defined documents rather than the actual knowledge of someone at the seller. Insurers generally expect a full buy-side diligence process precisely because they do not have a seller disclosure exercise to rely on. The rep package can be narrower, underwriting can require more work and pricing can run above standard RWI. The additional cost and underwriting burden have been coming down as more carriers become comfortable with the product, but synthetic RWI is still a structured solution, not a commodity.

There is still an important limit to the trend, this is not yet how ordinary U.S. domestic buyouts get done. In a typical U.S. private equity transaction, the seller still gives the reps, prepares disclosure schedules and goes through the usual disclosure process, even if the buyer’s real source of recovery is an RWI policy. Synthetic coverage remains concentrated in transactions where the traditional structure does not work particularly well, which is also where the product started.

In an insolvency sale, for example, an administrator may have been involved with the business for only a short period, owes duties to creditors and may have no practical ability to stand behind a claim after closing. For years, buyers in those situations bought businesses largely on an “as is” basis and priced the uncertainty into the deal. Around 2018, insurers began offering another option: instead of simply insuring reps given by the seller, they would provide the rep package themselves.

From there, the idea started to make sense in other situations. Passive shareholders in a take-private may know very little about the company’s operations. A JV partner being bought out may no longer have access to the information needed to give detailed business reps. A sponsor winding up a fund may want to distribute the proceeds and close the fund rather than leave money sitting in escrow for another 18 months. In each case, asking the seller to give a full set of business reps creates a practical problem. Synthetic coverage gives the buyer protection without forcing the seller to make statements it may not be in a position to verify.

Europe has moved further in this direction than the United States. Clifford Chance noted in its 2025 market review that synthetic W&I is becoming more commonplace, although it remains a more specialized product than conventional W&I. There is also a middle ground that may tell us more about where the market is going than fully synthetic deals themselves. In competitive European auctions, nil-recourse structures have become common. Sellers may give warranties but cap their liability at £1 or €1, leaving the W&I policy as the buyer’s only meaningful source of recovery. Some sponsor-led sales go further and refuse to give business warranties altogether.

Formally, these structures are different from synthetic W&I, but economically the distinction is getting harder to see. If a seller gives a warranty but has €1 of liability for getting it wrong, while an insurer provides the actual recovery, the seller is still the warrantor on paper but no longer carries the risk behind the warranty. Synthetic W&I simply asks whether that remaining formality is still necessary.

Continuation funds have become one of the best tests of whether this model can work at scale. GP-led secondary volume reached roughly $115 billion in 2025, according to Jefferies, up 53% in a year and crossing $100 billion for the first time. Continuation vehicles made up most of that activity, and nearly 80% of the 100 largest sponsors by assets under management have now completed at least one continuation vehicle transaction.

These deals also happen to be unusually well suited to synthetic insurance. When a sponsor moves a portfolio company from an older fund into a continuation vehicle that it also manages, the traditional idea of a seller standing behind a full set of business reps becomes awkward. The old fund wants liquidity, the GP has relationships on both sides of the transaction, and the incoming investors want meaningful protection without leaving the selling fund exposed to years of post-closing claims. Insurance gives the new investors another source of recourse and, just as importantly, has forced underwriters to learn how to evaluate deals where seller liability and seller warranties are limited or absent. They rely instead on diligence, management access, the data room, underwriting Q&A and, where relevant, the work done by lead investors and other advisers.

That may be the more important legacy of continuation funds for ordinary M&A. They have given insurers a large and growing market in which to learn how to underwrite the business rather than simply insure the seller’s promises about it. At the same time, the insurance product continues to expand around the edges. Synthetic tax coverage can replace indemnities a seller refuses to give. Some insurers are offering broader coverage for obligations that historically sat outside ordinary RWI. New-breach coverage is also beginning to address a particularly awkward problem in split signing and closing deals: a representation that was true at signing but becomes false before closing. None of these developments eliminates the purchase agreement, but taken together, they show how much risk that once had to be negotiated directly between buyer and seller can now move into an insurance contract.

For buyers, there is a trade-off here that is easy to miss. On paper, a synthetic rep package can look as good as, or sometimes better than, what the seller would have agreed to give. But the reps themselves are only part of what the traditional process gave the buyer. When a seller has to sign a representation, someone has to check whether it is true. Management gets questions, counsel works through the reps with the business, disclosure schedules start filling up with exceptions, and a negotiation over a knowledge qualifier often forces someone to find out whether the underlying statement is actually correct.

That process produces information. It is easy to think about representations only as a way to allocate risk after closing, but they also create a process before closing. They force the people who know the business to verify facts, identify exceptions and put information in front of the buyer. RWI already separated that process from much of the financial liability behind it. Synthetic reps separate it further.

In a synthetic deal, nobody on the sell side is standing behind the business reps in the same way. The insurer is underwriting based on information available to it, including a data room it did not prepare and diligence it did not perform itself. The buyer’s diligence therefore has to do more work. It is no longer just helping the buyer understand the business and negotiate the deal. It is also part of the basis on which the insurer decides what it is willing to cover.

Insurers deal with that additional risk through the policy. The policy may treat the entire data room as disclosed against the reps, so if a problem appears somewhere in the materials, the buyer may have trouble recovering even if nobody focused on it before signing. Insurers may also apply deemed knowledge or other qualifiers to reps that look broader when you first read them. Depending on the structure and the carrier, subrogation rights against the seller can also look different from the broad fraud-only waiver buyers are used to seeing in traditional RWI policies.

Claims experience makes these details more important. Around 18% of RWI policies produce a claim, roughly half of claims are made more than a year after closing, and claims seeking damages based on a multiple of the underlying loss increased from 5% of claims in 2020 to 32% in 2024. If there is a claim two years after closing, the insurer will look not only at the words of the rep but also at what the buyer knew, what its advisers reviewed, what appeared in the data room and what came up during underwriting. A broad rep on the first page of the policy may therefore provide less protection than it seems if the rest of the policy and the diligence record narrow what is actually covered. The old disclosure process was doing more work than we sometimes gave it credit for.

This is where synthetic reps become more interesting than the product itself. Historically, the structure was straightforward. The seller knew the business, gave the reps, disclosed the exceptions and bore the financial risk if those reps were wrong. RWI has already broken one part of that chain. The seller still knows the business, gives the reps and prepares the disclosures, but the insurer increasingly bears the financial risk. Synthetic RWI asks whether another part can be separated too: can the seller provide the information, the buyer diligence it, and the insurer provide the protection without the seller formally giving the business reps at all?

For now, there are good reasons why the answer in an ordinary U.S. buyout is usually no. A conventional seller can actually give reps. Management knows the business. Seller’s counsel already has a disclosure process. RWI pricing is competitive, and the seller usually has very little economic exposure anyway. Replacing a process that works with one that can require more buyer diligence, produce a narrower rep package and cost more to insure does not necessarily make the deal better. Synthetic reps currently solve a problem that most U.S. buyouts do not really have.

But a similar argument could once have been made about zero-recourse RWI. Sellers had always stood behind their reps because that was how private M&A worked. Then insurance became cheaper, underwriting became faster, carriers became more competitive and sponsors realized they could give sellers a cleaner exit without giving up meaningful buyer protection. Seller recourse disappeared surprisingly quickly from insured sponsor deals.

The same change does not have to happen with reps themselves, and it probably will not happen in the same way. The more likely path is gradual. European auctions already show one version of it. Seller liability moves toward zero. W&I becomes the buyer’s sole meaningful recourse. Some sellers stop giving business warranties. Synthetic coverage fills specific gaps. Insurers take on risks that previously had to sit somewhere in the purchase agreement. The documents may remain formally traditional long after the economics have changed.

Fully synthetic RWI therefore does not need to become the standard U.S. product for synthetic structures to influence ordinary buyouts. Some of the underwriting techniques can migrate first. So can synthetic tax coverage, new-breach protection and other targeted solutions that remove individual pieces of seller exposure. If those products become cheaper and easier to underwrite, the question may gradually change from why a deal should use synthetic coverage to why it still needs the traditional structure.

Latin America is probably further away. RWI is less deeply embedded in the region’s middle market, many transactions still rely more heavily on negotiated seller recourse, and the underwriting ecosystem is not as developed as it is in the United States or Europe. Synthetic reps are unlikely to become broadly common before ordinary RWI itself becomes more established.

Larger cross-border sponsor deals may move differently, though. A U.S. or European sponsor buying or selling a Latin American business through an international auction already brings global insurers, international counsel and deal terms developed in more mature RWI markets into the transaction. For the right seller, particularly a passive shareholder, a fund nearing the end of its life or a transaction where meaningful seller recourse is difficult, there is no particular reason synthetic structures need to wait for the broader local market to catch up. That will probably produce selective adoption before anything resembling normalization.

The more interesting question, then, is not whether every seller will eventually stop giving reps. It is how much of the traditional M&A risk-allocation system insurance can absorb. Seller indemnity has already largely disappeared from insured sponsor deals. European auctions have pushed seller recourse toward zero. Continuation funds have shown that insurers can underwrite transactions where a traditional warrantor barely exists. Synthetic products are now testing whether the reps themselves can move too.

There are still good reasons to keep the seller involved. Insurance can replace the balance sheet behind a representation much more easily than it can replace the process that made the representation useful in the first place. Someone still needs to check whether the company owns its IP, complies with its material contracts, pays its taxes and has disclosed its liabilities. That work does not disappear because an insurer writes the policy.

For deal lawyers, that changes where the attention needs to go. You are still negotiating the buyer’s protection, but part of that negotiation may now happen with an underwriter rather than opposing counsel. You need to understand which reps an insurer will cover, which deemed qualifiers matter, how the diligence record affects coverage and which exclusions are likely to matter if there is eventually a claim. Good diligence becomes more important, not less, and so does reading the policy with the same care you would normally reserve for the SPA. If the insurer is the party your client will actually turn to after closing, the details of that contract are part of the deal.

Synthetic reps remove the seller from one part of the process, but they do not remove the questions the reps were designed to answer. The seller may eventually stop standing behind those answers financially, and in some deals it may stop formally giving them at all. Someone still has to find the answers.

Synthetic RWI is forcing the market to figure out how much of the old structure we actually need, and how much of it we keep doing simply because that is how M&A has always been done.

The week’s dominant theme was transatlantic realignment under pressure. Europe’s elite firms are remaking themselves to compete on US terms. Freshfields is cutting equity partners, Linklaters was rebuffed by Cleary and responded by raiding Paul Weiss, and A&O Shearman published the first clean numbers on whether a transatlantic merger actually pays.

Lead Stories

Linklaters' transatlantic week: rebuffed by Cleary, then raiding Paul Weiss

Linklaters approached Cleary Gottlieb about a transatlantic merger earlier this year and was turned down. Linklaters denied making an approach; Cleary said it is focused on its own growth. Then on July 14, Linklaters announced it had hired a senior New York litigation duo from Paul Weiss; Christopher Boehning, lead counsel to FIFA and a 30-year Paul Weiss veteran, who joins as chair of global sports and US strategic disputes, and Daniel Levi, a 25-year veteran of the firm.

Why it matters: after A&O Shearman, HSF Kramer, Hogan Lovells Cadwalader and Ashurst Perkins Coie, every remaining Magic Circle firm is understood to be hunting for a US partner, and the Cleary episode confirms where the leverage sits. Elite New York firms do not need the merger; London firms increasingly believe they do. What makes the week interesting is the counterpunch. Paul Weiss has spent two years stripping talent out of London rivals. A European firm successfully recruiting two senior New York litigators out of Paul Weiss is the strongest evidence yet that the organic route can land elite US talent, not just fill seats. Linklaters says its US revenue has more than doubled since FY21 and its US partnership is up 77% since FY22.

Unanswered: whether the Cleary approach was a one-off or the first move in a longer campaign.

Freshfields cuts equity partners to fund its US push

The Financial Times reported, and trade press confirmed on July 13, that Freshfields has removed equity partners in recent weeks and downgraded the equity shares of others, with London, Paris and the German offices all affected. Dozens of the firm’s roughly 500 partners reportedly received fewer equity points. The cuts follow last year’s move to a flexible, performance-based compensation system and the creation of a non-equity tier earlier this year.

Why it matters: this is the clearest signal yet that a Magic Circle firm is remodelling its European partnership to fund US-market compensation. The lockstep social contract, under which senior European partners were insulated from year-to-year performance swings, is being dismantled in real time to keep US rainmakers paid at New York rates. Every Freshfields departure this year now reads differently, including this week’s: Emma Rachmaninov, a Freshfields lifer and one of London’s leading financial services regulatory lawyers, left to co-head Simpson Thacher’s European financial services and funds regulatory practice, announced July 15.

Equity cuts save money but tell every productive mid-career partner that tenure protects nothing, which is precisely the pitch US firms make when they call. Freshfields is betting the retained profits outweigh the retention risk. The German and Paris offices, where the firm’s brand premium is largest and US competition thinnest, are where that bet is safest. London is where it is not.

Other Notable Moves

• Simpson Thacher opens in Dallas with an Akin raid (US). The firm hired a three-partner hybrid capital team from Akin Gump (Ryan Cox, Payson Lyman, Joshua Morry) and opened a Dallas office on July 14, with Cox and ex-Kirkland debt finance partner David Nemecek as co-managing partners. Another marker in the Texas land grab and in the private credit talent war.

• Norm Law lands a genuine funds rainmaker (US). John Budetti, former global investment funds chair at Paul Hastings and a senior Kirkland funds partner, joined AI-native firm Norm Law on July 14 as co-head of global private capital funds. The highest-profile Big Law transactional lawyer yet to move to an AI-native platform.

• Harvey buys its way into asset management (US). Harvey acquired investment platform Benchmark on July 16, its third acquisition of 2026, and says it now serves 50+ asset managers including KKR and Blue Owl, with $100 million-plus in net-new ARR added in Q2. Legal AI vendors are moving past law firms into clients’ own deal workflows.

• Cooley keeps assembling a privacy bench (US). Michael La Marca, nearly a decade at Hunton Andrews Kurth, joined Cooley’s cyber, data and privacy group in New York, reported July 14, following May’s hire of Perkins Coie’s privacy co-chair.

• Cravath’s London founder generation steps back (UK/US). Philip Boeckman, EMEA capital markets co-head and a builder of Cravath’s London office since 2000, is retiring, reported July 15. A real succession question for one of Wall Street’s smallest, most senior London operation.

• Paul Hastings formalises a global sports practice (UK/US). Announced July 13, chaired from London by ex-Slaughter and May partner Mark Zerdin, with funds partner Chidi Oteh and former Baltimore Ravens GC Brandon Etheridge. The third elite firm in a year to institutionalise sports, following the private capital flooding into the sector.

• A&O Shearman posts the first clean merger scorecard (UK/US). Results announced July 16: roughly flat GBP revenue, double-digit profit growth, and PEP back to pre-merger levels after this year’s partnership and business-services streamlining. This is the benchmark every other transatlantic tie-up will be measured against.

• Hernández & Cía. rearms in Lima (Peru). Nydia Guevara rejoined the firm as banking, capital markets and fintech partner on July 16, arriving from rival Rodrigo, Elías & Medrano. A rare partner move between elite Lima independents, weeks after Pérez-Llorca completed its absorption of Miranda & Amado.

• DLA Piper takes a Kirkland project finance partner for LatAm (US/LatAm). André Teixeira joined DLA’s project finance practice in Washington on July 14, hired from Kirkland to build cross-border energy and infrastructure finance work into the region.

• Chile’s privacy law starts redrawing the mid-market (Chile). IP boutique Villaseca Abogados and tech firm Fernández & Yuraszeck announced a strategic alliance on July 14 built around Chile’s incoming GDPR-style data protection law. The first visible consolidation play driven explicitly by the new regime.

• Brazilian independents hire for AI and energy (Brazil). BBL Advogados added Ana Leticia Allevato (from Mayer Brown-affiliated Tauil & Chequer) for IP, privacy and AI, and KLA brought in M&A and infrastructure partner Mauricio Veiga, both reported during the week. AI regulation and energy transition work are now driving lateral traffic in São Paulo and Rio.

What to Watch

• The Magic Circle’s US problem has become a partnership problem. For a decade the US question was strategic: merge, build, or wait. This week showed it has become internal. Freshfields is cutting European equity to fund US pay. Linklaters is simultaneously shopping for a merger and proving it can hire against Paul Weiss. A&O Shearman’s results suggest a merger can work, but only after cutting its way back to pre-merger profitability. The common thread is that the cost of competing in New York is now being charged to European partners, and each firm is choosing a different way to present the bill.

• AI stopped being a vendor story. The week’s three AI items were a rainmaker move (Budetti to Norm Law), an acquisition aimed at law firm clients rather than law firms (Harvey and Benchmark), and a Big Tech product launch (Amazon’s Quick for Legal, July 15). Each targets a different layer of the legal value chain, and none depends on law firms buying anything. The competitive threat to firms is no longer that they adopt AI too slowly; it is that capital and talent are starting to route around them.

• LatAm's independents are consolidating defensively. Pérez-Llorca's absorption of Miranda & Amado has visibly accelerated movement in Lima, and Iberian and Brazilian mid-market firms are pairing up rather than waiting to be bought. The week's activity clustered in Peru, Brazil, Chile and Mexico; Argentina and Colombia were quiet.

Hope you have a great weekend.

The Weekly Brief is back. Here's a curated look at the week's most important law firm moves across the U.S., Latin America, and Europe. Hope you enjoy it.

The week the market has been circling since the spring finally arrived. On July 1, two transatlantic mergers went live on the same day, and one of them is the largest law firm combination ever recorded. Add a marquee restructuring hire in New York and a run of infrastructure and energy moves in London, and the quarter closed on a note of open consolidation.

Lead Stories

Hogan Lovells Cadwalader opens for business, and resets the top of the table

The merger of Hogan Lovells and Cadwalader, Wickersham & Taft went live on July 1, creating a firm of more than 3,200 lawyers and roughly $3.6 billion in combined revenue on 2024 numbers. That makes it the world’s fifth-largest firm by turnover on day one, second-largest in Washington, top 10 in London and top 25 in New York. Miguel Zaldivar, who ran Hogan Lovells, leads the combined firm. Cadwalader’s Patrick Quinn takes global integration and Wesley Misson runs the finance practice.

Hogan brings breadth across corporate, regulatory, IP and disputes. Cadwalader brings a concentrated, high-margin finance and structured products franchise that is hard to build and harder to poach. The interesting question is whether a boutique’s economics survive absorption into a global platform. Cadwalader’s partners were used to a specific profit model and a specific culture. Integration risk in law firm mergers almost always shows up in the numbers 18 to 24 months out, when the laterals who were going to leave have left. Watch partner retention in the structured finance group through 2027, because that is where the value either holds or walks.

Ashurst Perkins Coie completes on the same day, and builds a genuine two-way platform

The other July 1 closing gave the market Ashurst Perkins Coie, a roughly $2.8 billion firm with more than 950 partners and 3,500 fee earners across 52 offices in 20 countries. Paul Jenkins of Ashurst and Bill Malley of Perkins Coie run it as global co-CEOs, with Karen Davies and Brian Eiting as co-chairs. Flagship hubs sit in London, New York, Seattle and Sydney.

This is the deal that solves a real gap on both sides. Ashurst had wanted durable US depth for years and never quite bought it. Perkins Coie had a strong domestic and West Coast technology practice but a thin international platform for clients going global. The combination is more complementary than competitive, which is the profile that tends to integrate cleanly. The open question is cultural rather than strategic. Perkins Coie’s identity is tied to the Pacific Northwest and to tech clients who chose it for reasons that were never only about capability. Whether a London-led global structure preserves that, or slowly dilutes it, will decide if this becomes a top-20 firm in substance or only in the rankings.

Paul Weiss lands Sprayregen, and the restructuring talent war escalates

On June 25, Paul Weiss confirmed that James “Jamie” Sprayregen is joining to co-lead its restructuring and debt capital solutions practice, alongside Brian Hermann and Andrew Rosenberg. Sprayregen built Kirkland’s restructuring group starting in 1990 and turned it into the dominant debtor-side practice in the country, with a résumé that runs through Caesars, Toys “R” Us and United Airlines. He left Kirkland in 2024 for Hilco Global, and now returns to private practice at a direct competitor.

This is the clearest signal yet that restructuring is where the lateral market is heating up going into a softer credit cycle. Firms are positioning for distressed volume before it arrives, and the scarce input is not associates but a handful of names that creditors and boards already trust. The hire also fits a broader reshaping at Paul Weiss under new chair Scott Barshay, who took over after Brad Karp’s retirement this year. Co-head Paul Basta steps back from his role by year end as part of succession. Bringing in an outside star to anchor that transition is a bet that reputation transfers with the person. Sometimes it does. The Kirkland restructuring machine, though, was always bigger than any one partner, which is the risk buried in a hire this expensive.

Other Notable Moves

• Paul Hastings - hires private equity partners Lyndsey Laverack and Jade Williams-Adedeji from Covington (London). Deepens real estate and infrastructure PE coverage in a market where sponsor-side infra work is in demand.

• Cooley - twin partner hire from Kirkland into its infrastructure, energy and real estate group (London/US). Signals continued build-out of energy transition and infra capacity, and more pressure on Kirkland’s bench.

• Ritch Mueller - adds a competition and antitrust partner from local outfit Conesa & Moreno Abogados (Mexico). One of Mexico’s elite firms deepening a regulated-markets antitrust bench as merger control and monopolistic-practice work stays active.

• CMS - adds arbitration partner George Burn from BCLP (London). Energy, infrastructure and emerging-markets disputes focus; a straightforward strengthening of an already deep arbitration bench.

• Martinelli Advogados - recruits a tax partner from Bastos-Tigre, Coelho da Rocha, Lopes e Freitas Advogados (Brazil). Continues a sustained partner hiring run, now building out tax as Brazilian firms compete on full-service depth.

• Skadden → OceanSound Partners - Ken Wolff, co-head of Skadden’s private equity group, leaves to become partner and chief legal officer at the PE firm. A reminder that the best sponsor-side lawyers increasingly have a buy-side exit.

• Marval O’Farrell Mairal - integrates Neuquén-based Pujante Hernaez Abogados, adding two partners, to strengthen oil and gas, energy and infrastructure coverage (Argentina). A regional tuck-in aimed squarely at Vaca Muerta, where hydrocarbon development is driving demand for local energy counsel.

• Polsinelli - adds Christian Fabian as a PE M&A shareholder in Chicago, its sixth M&A shareholder hire this year. Mid-market build-out continuing at pace.

• Pérez-Llorca - London leadership handover as Álvaro Ramírez de Haro returns to Madrid and José Luis Romeu takes over the office; finance partner Ander Valverde also returns to coordinate Madrid finance. Iberia’s most acquisitive firm is managing its London bet, not retreating from it.

What to Watch

• Partner retention at both merged firms. The real test of Hogan Lovells Cadwalader and Ashurst Perkins Coie is not the launch press but who is still there in 18 months. Watch the high-margin groups, structured finance at the former and West Coast tech at the latter.

• Kirkland’s response. Two of this week’s moves came at Kirkland’s expense, including its founding restructuring name. A firm that hires as aggressively as Kirkland rarely absorbs that quietly.

• The next merger approach. With two mega-deals now live and working, boards at other mid-to-upper firms will feel the pressure to have their own answer. The interesting names are the independents that have long insisted they would never merge.

• Restructuring build-outs. If Paul Weiss is paying up for a restructuring anchor, peers reading the same credit signals will move too. Watch for further senior distressed hires through Q3.

• Iberia and cross-border US–Spain flow. Pérez-Llorca’s London reshuffle and continued US interest in Spanish and Latin American work are worth tracking as a durable corridor, even in a week where the headline news sat in London and New York.

In case you missed it, here is our last essay:

Hope you have a great weekend.

The SEC’s enforcement division has opened a probe into continuation funds, Reuters reported this week, and the things it is reportedly looking at are conflicts of interest, how managers value the assets they move, and whether investors are being told a consistent story. You didn’t need inside sources to guess that list, because all three problems come from the same place, which is that the manager runs both sides of the trade. The more interesting question is why that single fact produces exactly those worries, and why the structure has spread so fast in spite of them.

Private equity has had a problem for a couple of years now that it prefers not to talk about. The companies it bought before interest rates climbed are ready to sell, the funds that hold them are running out of time, and the buyers who used to take them off its hands, rival firms, big corporates, the public markets, have mostly stopped showing up at the prices the marks require. With more than 30,000 bought-and-unsold companies now sitting in portfolios, a fund that can’t sell can’t return cash to its investors, and a firm that can’t return cash struggles to raise the next fund that keeps it alive. So it started doing something that sounds odd said out loud but has come to feel ordinary in practice, which is selling companies to itself.

A firm runs a fund that owns a company. Instead of selling it to an outside buyer, it sets up a new fund, which it also runs, and moves the company across. The investors in the old fund choose when the deal closes, to take their cash at the agreed price or roll their stake into the new fund and stay in, while the fresh money comes from firms that buy and sell stakes in private funds for a living. The manager usually puts its own proceeds back in and the same people keep running the company as before, so what you get is less a sale than a reshuffle of who owns the business and on what terms. Nothing about the company changes; what changes is that the manager now sits on both sides of the table, selling for the old fund and buying for the new one, and setting the price for both.

For most of the last decade this was a quiet, slightly shabby corner of the market, where a struggling manager parked the assets it couldn’t sell to keep a dying fund alive, but that reputation has gone. Manager-led secondary deals, the category these fall into, were worth $106 billion last year, up from $70 billion the year before, and continuation funds make up most of it. The structure that used to signal trouble is now where managers keep their best companies, often building a whole new fund around a single one, and you don’t go to that trouble for a business you’re trying to offload. You do it for the one you’d hate to lose, which raises the obvious question of why a firm would work this hard to buy a company it already owns.

The answer begins with what an exit is for. Everything in this business is built around the moment a company is finally sold to someone else, the moment that turns a number on a report into money in an investor’s account and tests whether the number was ever honest, since an outside buyer paying a real price is the one thing you can’t talk into existence. A continuation fund keeps almost everything about that moment and removes the moment itself. The manager records a sale and books the gain, the investors who want out get their cash, and the manager takes its cut of the profit now, even though the company was never sold to anyone outside the firm, never priced by a market, never made to face a buyer who could look at the marks and walk away. It counts as a sale while nothing was really sold.

A manager earns its cut of the profit, the carry, only after the whole fund has returned investors their money plus a minimum return, so in a fund where some deals went badly the winners cover the losers first, and the gain on one great company is netted against the duds before the manager sees a cent. That netting is deliberate, a way of tying the manager’s pay to the whole portfolio it built, not its luckiest pick, and lifting the one winner out into a new fund quietly breaks it. The manager banks its cut on that company alone, at a price it had a heavy hand in setting, without waiting for the laggards to recover, and then resets the clock and starts earning a fresh cut on the same company all over again.

The industry’s justification to this is alignment, and it isn’t a bad one. In most of these deals the manager rolls all of its own proceeds back into the new fund, keeping its money at risk on the company instead of walking away, which matters, because a manager backing its own conviction is a better sign than one heading for the door. But rolling the proceeds back in is a separate thing from the carry it banked along the way, since it has still turned a maybe-someday cut into a sure-thing-today one, and done so by trading with itself. Some readers will recognize the shape from the other direction, because when I wrote about rollover, the founder took a second bite that sat behind the sponsor, junior and contingent and controlled by the other side, while here the sponsor takes a second bite of its own, on the cleanest possible terms, against a counterparty that happens to be itself.

The valuations are the part the SEC keeps coming back to, and the suspicion is that the manager marks the company wherever its carry needs it to be. The numbers give that suspicion something to chew on, because one study of more than 1,000 companies later sold out of these funds found they had been valued, on average, 28% higher than the same manager had valued them 6 months earlier. An honest person holds two readings of a gap like that, the unkind one being that the company gets marked low going in, so the gain and the carry look bigger coming out, and the kind one that these really are good businesses, and 6 months of growth plus a genuine outside bid produced a higher and truer number. The data doesn’t say which is doing more of the work, but a 28% gap showing up that often is large enough to make you ask what a private valuation measures when the same hand writes both sides.

To be fair, the price is not whatever the manager decides over coffee. There is usually a real outside check on the number. In a normal sale, a buyer and a seller want opposite things, and the price comes out of that tension. A continuation fund does not quite work that way. It looks like an M&A deal where buyer and seller are tied to the same firm, and also like a fundraise where the money is being used to buy something the manager already owns.

What stands in for the missing counterparty is the lead buyer, usually one of the large secondary firms that does these deals at scale. It funds a big piece of the new vehicle, negotiates the price, and gives everyone else a number to work from. That outsider, more than the paperwork around the deal, is what keeps the price connected to reality. But the comfort only goes so far. The lead buyer is trying to buy well. It is not there to protect the investors selling out of the old fund. So yes, it pushes on price, but it is nobody’s fiduciary, and the manager still controls how competitive the process is in the first place.

A thin auction, or one where the manager’s own next fund quietly buys in alongside, can produce a price softer than the word “negotiated” suggests. The fairness opinion that usually comes with the deal does not fix that problem. It is a letter from a bank hired by the manager saying the price sits in a defensible range. That has value, but less than the name implies. It catches the indefensible. It does not catch the merely very good, which is where most prices live.

Now sit in the seat of an investor in the old fund, and the SEC’s disclosure question starts to make a lot of sense. When the deal lands, you get a letter, a deadline, and a choice: take your cash or roll into the new fund. In practical terms, you are being asked to sell, or buy again, a company you already own, at a price your manager helped set, on a clock your manager controls, using information your manager controls.

Investors have won real protections here. They usually get time to decide, access to the same information incoming buyers receive, a default rule that treats silence as a request for cash instead of pushing inattentive investors into the new vehicle, and the right to keep their old economics if they roll. Still, the bind survives, because both doors lead back to the same manager.

Roll, and you hold the same company for years more, usually deferring the tax instead of escaping it. Cash out, and you sell back, at the manager’s price, the one company the manager’s own behavior suggests may be the best asset in the fund. There is no version of the choice where the other side of the table is not your manager.

The body meant to watch all this is the fund’s advisory committee, a small group of the larger investors who vote to approve the conflict, and the process has tightened, so the review is private, paid for by the fund, with everyone on the same information. It is still a soft check, because the same big investors sit on one fund’s committee after another and end up on both sides of these trades, with one quoted sitting on around 60 at once, and when the regulator looked at these bodies it doubted they had the independence or the time to police conflicts this size. The gatekeeper is real, and it is staffed, more often than not, by the manager’s own repeat investors.

You might assume something this conflicted, at this scale, would already be tightly regulated. The SEC tried. In 2023 it adopted a rule that would have required an independent fairness or valuation opinion for these deals and disclosure of the adviser’s ties to the manager. In 2024, a federal court struck the rule down, so the requirement disappeared almost as soon as it arrived. With no clear rulebook in place, the enforcement division has stepped in after the fact, asking the questions that were always going to matter: conflicts, valuations, and whether buyers and sellers were told the same thing.

That is not the same as a finding. Being investigated does not mean anyone did anything wrong. It also does not solve the core problem in real time. Enforcement can punish a bad mark years later, but it cannot sit at the table when the price is set. At the moment that matters, when the manager decides the price and investors choose whether to roll or cash out, the real limits are still the fund documents, the process, and the industry’s own conventions. Those conventions matter because reputation matters, and the big secondary buyers are repeat players with long memories. But convention is still convention. The SEC showing up increases the risk of getting caught later. It does not change who is doing the work on the day.

The important metric here is Distributions to Paid-In Capital (DPI), which basically shows how much cash a fund has actually returned to investors. DPI has always carried weight because cash is supposed to be hard to dress up. Either the money reached investors or it did not. A continuation fund complicates that logic. It produces real cash, but not from a true exit into the market. It comes from new investors buying into a vehicle managed by the same firm, often to keep holding the same company. The cash is real. What it proves is less clear.

That is the part worth watching. Private equity is several years and several hundred billion dollars into a test of whether it can keep marking, paying out, and earning fees on companies it has not really sold. The old test at the end of the deal was much simpler, an outside buyer paid a price. Continuation funds defer that test. Maybe the structure holds because the companies are good and the secondary buyers are tough on price. Or maybe it holds until a wave of these funds has to sell into a market that disagrees with the marks.

We are not in a place where we can say continuation funds are fundamentally bad. Often they are useful, and now and then they really are the best company a manager has, kept on for honest reasons. The useful point is narrower than the scandal version wants it to be. A price paid by an outside buyer tells you something because a stranger put real money behind it. The price inside a continuation fund tells you less because the firmest conviction behind it belongs to the party on both sides. The cash that reaches you is just as real either way. What thins out is what that cash is supposed to prove about the company. That is worth holding onto the next time a pitch leads with a clean line of returned capital that turns out, on a closer look, to be a company sold quietly from one of the manager’s hands into the other.

In early 2025 the Delaware Court of Chancery decided a dispute, the Dura Medic case, that most U.S. deal lawyers will likely recognize from their own files. A private equity firm bought a medical equipment supplier. The sellers represented that the company had received no written notice of healthcare-law violations in the prior three years, except for one listed in the disclosure schedules. After closing, the buyer found others, and one of them turned into a costly government review. The buyer sued. The sellers’ defense was the one sellers always reach for, we told you about this. There was a diligence call before signing, they said, where we walked you through these very notices. You knew. You closed anyway. You cannot come back now and demand to be paid for a risk you accepted with your eyes open.

The court rejected that defense, and the way it did so is what makes the case worth writing about. It did not simply say the buyer wins. It explained why a representation works the way it does, and in doing so it turned a different clause, one that sits at the back of every purchase agreement, into one of the more consequential allocations of risk in the document.

So first, let’s back up and ask what a representation actually does. A representation in a purchase agreement is not a statement the buyer is asked to believe. It is a contractual assignment of risk. When the seller represents that there are no undisclosed regulatory notices, the seller is agreeing to carry the cost if that turns out to be false, whether or not anyone knew it was false at signing. That is the commercial point of the exercise. The buyer is buying a promise of recourse along with the asset. A breach of contract claim, unlike a fraud claim, does not require the buyer to prove it justifiably relied on anything. The seller, having sold the buyer the right to rely on its words, is in no position to argue afterward that the reliance was unreasonable. This is the core of Delaware contracts law, and it is why the state has settled into what people call a pro-sandbagging default. If the contract is silent on whether a buyer’s pre-closing knowledge defeats a claim, the buyer generally keeps the claim.

Delaware was generally understood as pro-sandbagging even before Dura Medic, but the doctrine still carried enough noise to keep the issue alive and the market divided. Most opinions leaned on the buyer’s way, but none of them settled the question outright. Dura Medic is the cleanest statement yet, and it goes one step further in a way that changes how we draft.

The sellers argued that even if their diligence-call disclosure never made it into the agreement, it should still modify the written representation, because the buyer actually received the information. The court said no, and it pointed at the integration clause to explain why. An integration clause is the boilerplate stating that the written agreement is the entire agreement and supersedes everything said or written before it. Every deal has one, almost nobody negotiates it. And it does precisely what it says; information conveyed outside the four corners of the document, in diligence sessions, site visits, management presentations, employee interviews, emails, cannot operate to change what the representations say. If the disclosure is not in the schedules, it does not qualify the rep. The real significance of Dura Medic is not merely that Delaware remains pro-sandbagging. It is that the court used a standard integration clause to explain why extra-contractual disclosures could not modify contractual representations.

For sellers this is the part to sit with. Disclosure that lives in a data room or a conference call and never lands in a disclosure schedule does not narrow your exposure. It may help you defend a fraud claim, where the buyer’s knowledge actually matters, but it does nothing for the breach-of-contract indemnification claim that is the real fight in most deals. If you want your diligence disclosures to count, write them into the schedules. And if you want protection against a buyer who knew and stayed quiet, you need an express anti-sandbagging clause, the kind that requires the buyer to show it lacked knowledge of the inaccuracy before it can recover. Silence now clearly favors the buyer.

The harder question, and the more interesting one for anyone working outside the United States, is whether any of this travels. Delaware reached its result through a particular worldview. The parties are sophisticated, the written deal is close to sacred, and the court enforces the allocation the contract describes rather than rescuing a party from a bargain it made. That worldview is not universal, so it is worth seeing how other jurisdictions handle the same question.

New York, the other dominant choice of law in US deals, is close but not identical. New York cares about the source of the buyer’s knowledge. If the buyer learned from the seller that a representation was false and closed anyway without preserving the point, a court may treat the warranty as waived. If the buyer learned of the problem from a third party, or otherwise preserved its rights, the claim is stronger. The result is a regime that asks how the buyer found out, which is exactly the distinction Delaware’s integration-clause logic tends to erase. Under Dura Medic, disclosure outside the agreement is close to irrelevant unless it made its way into the schedules. Under New York law, the path the information took can decide the case, although express contractual language can change the result.

English law leans the other way, but it gets there through a different way. English practice is less focused on abstract sandbagging defaults and more focused on the disclosure process itself. Sellers disclose against the warranties through a formal disclosure letter, often supported by the data room, and what is fairly disclosed there qualifies the warranties whether or not it sits in a schedule in the American sense. US-style sandbagging provisions are far less common in English deals, partly because the disclosure letter does much of the work that sandbagging clauses do in US agreements. The point is not that English law has a simple anti-sandbagging rule. It is that the market’s disclosure machinery usually pushes the analysis in a more seller-protective direction.

Continental civil law moves further still, although not in one uniform way. In jurisdictions like Germany, France, or much of Latam, courts are more likely to analyze the issue through good faith, waiver, abuse of rights, consent, or similar doctrines. A buyer’s positive knowledge of a defect at signing may weaken or bar a warranty claim, especially where the buyer closed without clearly reserving its rights, unless the parties have clearly agreed otherwise. The written agreement matters, but it sits inside a framework of good faith the parties cannot entirely draft away. This is the deepest divide. Delaware says the contract is the risk allocation. Civil law says the contract is a risk allocation that may still be read through the parties’ conduct.

I personally think that last qualifier, “unless the parties have clearly agreed otherwise,” is the whole game in Latin America. The protection a buyer wants may be enforceable in much of the region. It just lives in the drafting rather than the default.

Mexico is a useful example. There is no statute on sandbagging and the precedents are thin, so a buyer who stays quiet about a known breach would, in principle, face a harder claim than it would under Delaware law. But that does not mean Mexican law is blind to how sophisticated parties draft cross-border M&A contracts. At least in Mexico, there are precedents that leave room to interpret contracts by reference to comparative law and commercial usage, especially where the parties deliberately use concepts drawn from Anglo-American deal practice. If a Mexican SPA uses terms like representations, warranties, disclosure schedules, materiality scrapes, indemnity caps, or sandbagging, there is a serious argument that sophisticated parties understood those terms in the market sense they carry in that tradition.

That matters. An express pro-sandbagging clause gives the buyer a much better argument than silence ever would. It tells the court or tribunal not only that the parties allocated the risk expressly, but that they did so using a known transactional vocabulary. That contractual freedom holds in much of the region, subject to good faith, abuse-of-rights limits, and the specific rules of the governing law. What you cannot do is lean on the local default, which tends to run the other way through good faith, and you cannot draft around dolo, willful deceit, which generally cannot be waived or capped no matter how the clause is written.

Leaving legal systems behind, the question remains the same in any jurisdiction. When does information change who bears a risk, and what form does that information have to take to count. Sandbagging is that question asked about the buyer’s knowledge at signing. Knowledge qualifiers are the same question asked about the seller’s. When a representation is hedged with “to the seller’s knowledge,” the parties are agreeing that the seller carries the risk only of what it knew or should have known, and the fights over whose knowledge counts and whether constructive knowledge is included are fights over where that line falls. The disclosure schedule is the same question in physical form, the agreed list of facts moved from the seller’s side of the ledger to the buyer’s. The decision is, at bottom, a ruling about what counts as a valid entry in that ledger. The schedule counts. Disclosures made outside the purchase agreement do not.

Which is what makes the disclosure schedule the real battleground, and that fight does not end at signing. Any deal lawyer reading this has probably lived through a grinding negotiation over what the seller can add to the schedules between signing and closing. Because in that gap, something usually surfaces. A customer gives notice. A regulator opens a file. The seller wants to supplement the schedules to capture the new matter, and to have that supplement do two things, satisfy the condition that the representations be true again at closing, and cut off the buyer’s indemnification claim for the disclosed item. Buyers resist both, and the market-standard mirrors the logic of Dura Medic. The seller can supplement for transparency, but the supplement is informational only. It neither cures the closing condition nor eliminates indemnity for anything that existed at signing. The principle is the same one running through every paragraph above. New information does not move the risk simply because the seller hands it over. It moves the risk only when the contract says it does.

There is a middle ground, though, and it appears when the seller has leverage. The usual compromise pulls those two functions apart. Supplements for matters that genuinely arise after signing, and not from the seller breaching a covenant, are allowed to update the representations for purposes of the closing condition, so the buyer has to close unless the new matter is large enough to trip the material adverse change threshold. But the buyer keeps its indemnity claim for those items. The trade often runs the other way too, if a supplement hands the buyer a right to walk and it closes anyway, it is treated as having accepted the matter and gives up the indemnity claim for it.

Here’s the practical advice. Treat disclosure as an act of drafting, not an act of telling. What you or your client say in a meeting, send in an email, or load into a data room is not where the risk gets allocated. The risk gets allocated in the words the parties sign, and the clear trend is that courts will hold them to those words and little else. That is the lesson from Dura Medic. Information changes the allocation of risk only when it enters the contract in a form the contract recognizes.

For a seller, that means the schedules are the disclosure that counts. Get the diligence record into them, ask for an express anti-sandbagging clause if you want protection against a buyer who knew and said nothing, and negotiate a real mechanism for the gap between signing and closing. Do not assume a court’s sense of fair play will fill what you left out, because in the systems that handle most large deals it will not. For a buyer, the lesson is understanding that the protections you lean on may be doing more work than you realize, and that work can shrink or disappear the moment the governing law moves from one tradition to another. Knowing what each provision actually buys you matters far more than assuming it travels.

If you work across borders, that is the deeper point. The same sentence carries different weight depending on who reads it. A buyer’s pre-closing knowledge can be irrelevant, decisive, or somewhere in between, and which one it is gets settled by the law the parties choose, the forum they pick and the contract language, often long before anyone imagines a dispute. Deal lawyers need to know what they are actually drafting, and stay open about how those words will land in the jurisdiction that ends up reading them. The same clause can answer the only question that really matters here, what your client knew and when that knowledge counted, in opposite ways depending on where the deal is read.

I think Delaware has the better instinct here. Where sophisticated parties are free to order their own affairs, the market is better off treating the purchase agreement as the thing that matters, and as the only instrument that allocates risk. The contract is where the value of the entire negotiation sits. Every hour spent arguing over a representation, a knowledge qualifier, a number in a schedule, is an hour spent deciding who carries which risk, and that work means something only if the document is later read to mean what it says.

This cuts both ways, which is why Delaware’s instinct strikes me as the more honest one. A buyer relies on the express representations in the agreement, and it should. That reliance is what lets it pay a price and move forward without re-verifying everything the seller put on the table. But the same logic has to give the buyer real recourse when those representations turn out to be false. That is the point Dura Medic makes through the integration clause. You cannot ask the market to take the contract seriously when the deal is signed and then let something said in diligence quietly rewrite it once a problem shows up. If the words allocate the risk, they have to allocate it all the way through.

In most of my M&A essays I’ve spent a lot of time on who is buying companies: private equity funds, search funds, family offices, strategics, and the trends shaping the buy side. I’ve paid much less attention to the sell side, and particularly to how the seller actually gets paid. That’s what I want to get into here, because it has changed more than many sellers realize.

A founder selling the company they built thirty years ago does not, in most mid-market private equity deals today, simply receive a price. They receive cash for most of their stake and shares in the buyer’s new vehicle for the rest, often somewhere between 15% and 30% of the total deal economics, and they walk out of closing as a minority shareholder in a company that now carries acquisition debt and is controlled by the people who just bought them out. This is rollover equity, and it shows up in well over half of mid-market sponsor deals now. In founder and family businesses, where the buyer needs the seller’s knowledge and relationships, it is closer to a default assumption.

The reason buyers want it is genuine, and it is worth understanding before anything else, because the whole structure gets sold on it. It is called alignment, and the idea is simple. When a fund buys a business it has never operated, it is buying blind. The founder knows which customer relationships are personal rather than institutional, which manager is one bad month from quitting, which numbers in the projections are solid and which are hopeful. If the founder takes all their money and leaves, every one of those risks lands on the buyer. But if the founder keeps a meaningful piece of their wealth invested in the business, his incentives change. Founder hands over a cleaner company, stays engaged through the transition, and tells the truth, because now he or she gets hurt too if things go wrong. In exchange, Seller gets the famous second bite of the apple. The business gets sold again in five or so years, hopefully bigger and at a better multiple, and their retained stake pays out a second time, sometimes for more than the first sale did. Everyone’s interests point the same direction. That is the pitch, and as far as it goes, it is true.

What the pitch leaves out is that alignment describes the holding period, and the money is made or lost on terms that operate at the exit, where the interests of the fund and the founder stop pointing the same direction. I want to spend this essay on the handful of things about rollover that practitioners understand and the standard explanations mostly do not say out loud. None of this is an argument against the structure. It is an argument for negotiating it like it matters, because in a succession sale the rolled equity is very often where the real money ends up being, and it is consistently the worst-negotiated part of the deal.

Start with the most basic problem, which is what the seller’s shares actually are, because this is where the alignment story gets complicated in practice. In a lot of mid-market deals the answer is simple and perfectly fair. There is one kind of equity, everyone holds the same shares, and the seller sits next to the sponsor on exactly the same footing. If that is your deal, the headline percentage means what it says. The trouble starts in the deals, more common as you move up in size, where the buyer builds two layers into the company. One layer gets paid back first when the business is eventually sold, with a fixed return that accumulates whether or not the business performs. The other layer, the common, is whatever is left after the first layer has been paid. The mistake almost everyone makes is to assume the sponsor takes the senior layer and the seller takes the junior one, like a clean split down the middle. That is not how it works. The sponsor holds a slice of both, in the same proportion, so the senior layer sits on top of everyone, including the sponsor’s own common. No sponsor puts all its money in the senior piece, because that would cap its return at the fixed rate and hand the real upside to whoever owns the common. Owning the common upside is the whole reason it is doing the deal.

Here is where sellers can end up worse off than buyers they think they are matching. The fair way to roll into a two-layer company is to take the same mix the sponsor takes, a slice of both layers in the same proportion, so the two of them rise and fall together. What often gets papered instead is that the seller’s roll goes entirely into the common, sitting behind the whole senior layer, while the seller pays the same price per share the sponsor paid for its blend of both. Same price, more junior position. And this is not a matter of opinion. When valuation firms are later hired to put a real number on a roll like that, for the fund’s books or for tax, they routinely mark it below the figure written into the deal documents, precisely because a junior, common-only position is worth less per share than the senior-and-common blend it was priced against. The buyer’s own side of the market, asked to be honest for accounting purposes, concedes that this seller overpaid for the roll on day one. The founder who rolled 20% of his proceeds into the common alone is holding something worth less than 20%, and the only question is the size of the gap.

In a great outcome, none of this matters much. If the business doubles, there is enough value that the preferred gets paid, the common gets rich, and the founder’s second bite arrives as promised. The disadvantage shows in the middle outcomes, which are, by definition, the most common ones. Picture a deal where the business grows modestly, the debt gets serviced, and the company sells five years later for somewhat more than the fund paid. The sale proceeds go first to the lenders, then to the sponsor’s preferred plus its accumulated return, and the common shareholders split what remains. In that scenario the fund does fine. It got its money back with a contractual return before the founder saw anything. The founder, holding common, can come out with a fraction of what their ownership percentage implied, or close to nothing, in a company that did not fail. They took the risk of an owner and got paid like an afterthought. The fund and the founder lived through the same five years and the same business performance and had completely different financial experiences of it, and the difference was decided on a page of the shareholders’ agreement that nobody on the sell side read carefully.

That page is negotiable, which is the part that should bother you. The single most valuable thing a rolling founder can ask for is to roll into the same mix the sponsor holds, a slice of both layers in the same proportion, so they move with the buyer instead of sitting alone in the common. Sponsors grant exactly this to the institutional investors who co-invest beside them all the time, because sophisticated money demands it as a matter of course. Some sponsors have started offering it to sellers too, having figured out that alignment is a strange word for an arrangement where the founder only wins in the scenarios where the fund was going to win anyway. But it is never the opening offer. The opening offer almost always puts the seller in the common alone, and a seller who does not know to ask will simply never be told that better was available. If the structure will not move, the fallback is blunt but simple: roll less. A smaller roll on bad terms loses less than a big roll on the same terms, and sponsors who will not improve the security will often accept a smaller check rather than reopen the architecture.

The single most common rookie mistake sellers make on price is to throw everything into the total number and treat the rollover percentage as a logistical detail to be settled later. Buyers run the same math in the opposite direction. A buyer who pays part of the price in its own shares, shares it creates and prices, can afford generosity on the number everyone toasts at closing. Bump the price by a few million, recover it by enlarging the roll, and the seller has been paid the bump in an instrument the buyer manufactured. This is also why rollover gets recommended as the natural way to bridge a gap when buyer and seller disagree on value. It works, but notice what the bridge is built from; the exact portion of the price the parties could not agree on gets paid in the most uncertain instrument in the deal. The seller wins the valuation argument and collects the winnings in lottery tickets. The practical lesson is that price and rollover terms are one negotiation, run by the buyer as one negotiation, and a seller who negotiates them separately will lose on the second whatever they won on the first.

And rollover does invite that kind of split attention, because it sounds so reasonable when a founder describes it. They talk about keeping a piece of the business, and keeping feels safe, conservative, almost humble. But nothing is being kept. The company they owned, debt-free and under their control, stops existing at closing. What actually happens is that the founder takes a quarter of their life’s liquidity and invests it, fresh, into a different asset. The same operations, now loaded with acquisition debt, inside a structure they do not control, with no ability to sell, no ability to force a sale, and no say in when the exit comes. If a financial advisor proposed that exact investment to a sixty-year-old client the week after a liquidity event, the meeting would be short. It can still be the right call. Nobody knows the asset better than the person who built it, and concentrated bets on things you understand deeply are how founders got wealthy in the first place. But it should be decided as what it is, a large new investment competing against everything else that money could do, rather than as the sentimental default it usually becomes.

And then there’s the timing assumption behind the rollover pitch that, in my view, deserves real scrutiny right now, more than at any point I can remember. The second bite story rests on a timeline everyone takes for granted. The fund holds for five years or so, sells, and the founder gets the payoff. That rhythm is currently broken across the industry. Private equity exits have fallen every year since 2021. Funds are returning cash to their own investors at the lowest rates on record, roughly half the historical pace. Sponsors are holding companies longer, and increasingly they are selling portfolio companies to vehicles they themselves control, which resets the holding clock for another five or more years. Now look at the rolling founder’s position with clear eyes. They have handed capital to a fund manager, they wait for that manager to produce liquidity, and they absorb the delay when liquidity does not come. That is exactly the position of a fund investor, with one difference; the actual fund investors negotiated protections. They have committees, reporting rights, rules about conflicts when a sponsor sells an asset to itself, and the ultimate leverage of refusing to invest in the next fund. The founder has whatever made it into the shareholders’ agreement. The rollover seller is, functionally, the most junior and least protected investor in the entire structure, recruited at precisely the moment the protected investors are complaining loudest about not getting paid. So when I look at a rollover today, my first question is what happens to this stake if there is no sale on the promised timeline, and whether the founder can be swept into one of these sponsor-to-sponsor internal sales without real consent. Most documents say nothing about it. You can guess who benefits from that silence.

One last thing deserves its own paragraph, because it's the one that catches founders even when they got everything else right. When the seller stays on to run the business after closing, which in succession deals is the norm, the buyer will usually connect the rolled shares to the job. The documents sort departing employees into good leavers and bad leavers, and they let the company buy back a bad leaver’s shares, often at the price originally paid rather than what the shares are now worth. Bad leaver gets defined broadly, sometimes broadly enough to include simply resigning too early. Read plainly, this means a founder can forfeit the entire second bite by leaving a job working for people they met during the negotiation, while the business itself thrives. Equity on these terms is not an investment. It is deferred salary with conditions, dressed as ownership. The clause arrives late in the process, inside documents nobody budgeted legal hours for, and I have watched people who spent three weeks fighting over a price adjustment worth a tenth as much wave it through unread.

I want to close where the pitch begins, because none of this is a case against rolling. The alignment is real, the structure exists because it solves a genuine problem on both sides of a succession, and the second bite happens. The point is narrower and, I think, more useful. Every protection described in this essay is standard, known, and obtainable, and the reason sellers do not get protected is almost never that the buyer refused. It is sequencing. The equity documents show up after the price is agreed, after the celebration, on the buyer’s paper, reviewed by lawyers hired to close a sale rather than to vet a ten-year minority investment, against a fee budget already spent on the purchase agreement. The fix costs almost nothing compared to what it protects. Treat the roll as a second deal. Price it like one, question it like one, and staff it like one, starting the day the letter of intent puts a single dollar into shares instead of cash.

On May 12, Anthropic launched what it is calling Claude for Legal. Twelve practice-area plugins spanning commercial, corporate, employment, privacy, regulatory, and AI governance work, plus more than twenty MCP connectors integrating Claude with the document management and research platforms that law firms and in-house teams already run on. DocuSign, Ironclad, iManage, NetDocuments, LexisNexis, Thomson Reuters, Box, Everlaw. The announcement also included deep integration with Microsoft 365, meaning Claude now operates as a context-carrying agent across Word, Outlook, Excel, and PowerPoint.

No surprises on the market response. Legal tech stocks dropped when Anthropic’s initial legal plugin launched back in February. Over 20,000 legal professionals registered for the company’s April webinar, the largest legal session Anthropic has ever held. Freshfields committed to a multi-year agreement deploying Claude across 33 offices and 5,700 employees, reporting a 500% increase in usage within six weeks. Quinn Emanuel, Holland & Knight, and Crosby Legal confirmed they are running Claude on live matters. Bloomberg Law reported that legal is now the top power-user job function inside Anthropic’s Cowork platform. And just days after the announcement, sources told Artificial Lawyer that OpenAI is developing its own legal vertical, reportedly branded Codex for Legal, with hires from the legal tech world and a plan for forward-deployed engineering teams.

I have been using Claude for transactional work since late last year, and I have tested enough of the competing tools to have opinions about most of them. This is my honest assessment on what the launch actually means, where Claude delivers real value, where the hype is running ahead of the product, and what practicing lawyers should actually do about it.

To understand what Claude for Legal represents, you need to understand the market it is entering. In the first quarter of 2026 alone, Harvey raised $200 million at an $11 billion valuation and Legora closed $600 million in total Series D funding at $5.6 billion. Harvey is now valued at roughly twice the size of the entire legal AI software market as estimated by Research and Markets. These numbers represent a bet on future revenue growth that is either prescient or deeply speculative, and the data does not yet resolve which. What is clear is that the legal AI space has moved past the experimentation phase. Harvey serves over 100,000 lawyers across 1,300 organizations. Legora reports more than 1,000 firms and in-house teams across 50 markets.

Harvey has built its competitive position around specialization and service. The platform offers AI agents that execute legal workflows end-to-end, from contract analysis through regulatory research to litigation preparation. Its base offering runs at roughly $1,200 per lawyer per month, and the company dedicates about 10% of its headcount to former lawyers in customer success roles who work directly with client firms, driving adoption, managing change, and building the custom agents that turn a general capability into something a real estate partner or a regulatory associate actually uses day to day. Harvey’s strategic alliance with LexisNexis gives it access to citation-backed research through Westlaw’s database. Its integration with Intapp provides ethical wall enforcement. The pitch is that legal work is too specialized, too high-stakes, and too institutionally complex for a general-purpose model to handle without a purpose-built layer between the model and the lawyer.

Legora, the Swedish platform formerly known as Leya, has taken a different path. Where Harvey emphasizes deep US law firm penetration, Legora has built broader international coverage across 50 markets with a collaboration-oriented platform it positions as the operating system for legal AI. Its recent acquisition of Qura, a company that built structured AI-native legal databases designed for precise legal reasoning rather than surface-level retrieval, signals a push into legal research that goes beyond what retrieval-augmented generation alone can deliver. Among law firms surveyed, Legora reports 4.3 non-billable hours saved per lawyer per week, with 42% reporting that they have won new work as a direct result of using the platform. If they hold at scale, these metrics represent a genuine shift in how firms compete for mandates.

Then there is Thomson Reuters’ CoCounsel, which scored highest in independent benchmarks early on and carries the enormous advantage of being connected to Westlaw’s proprietary legal database. CoCounsel is now also integrated with Claude, creating a strange dynamic of Thomson Reuters simultaneously competing with and distributing through Anthropic’s ecosystem. That decision alone tells you something about where the center of gravity is shifting. Thomson Reuters apparently concluded that being inside Claude’s environment is more valuable than trying to keep Claude out.

What Claude for Legal does differently from all of these is not that it is smarter on any given legal task. On Harvey’s BigLaw Bench, Claude Opus 4.7 scores 90.9 %, the highest of any model, while GPT-5.4 sits at 84.2 %. That is a meaningful gap. But Harvey and Legora both use Claude as their underlying model, so the raw intelligence advantage is shared across the ecosystem. The difference is that Anthropic is not building a walled legal product. It is building what Legal IT Insider described as an orchestration layer. A single interface capable of accessing legal research tools, document management systems, transaction platforms, and specialist legal AI products simultaneously. The pitch is not that Anthropic’s legal AI is better than Harvey’s or Legora’s, but rather that you should not need a separate legal AI at all.

That is what triggered the stock selloffs in February, and it is what makes this launch genuinely consequential rather than just another product release. If Claude becomes the environment where lawyers live, the place where they draft, research, communicate, and manage documents, then every other legal technology company becomes a connector within Anthropic’s platform rather than a standalone destination. Harvey, Legora, Thomson Reuters, and the rest become features, not products. The specialized legal AI companies understand this risk. Their counter-argument is that they offer model flexibility, the ability to switch between Claude, GPT, and Gemini depending on which performs best for a given task, combined with the kind of implementation depth, compliance infrastructure, and institutional understanding that a research lab cannot replicate overnight. That argument is defensible today. Whether it remains defensible as the foundation models continue absorbing domain-specific capability is the question that will define this market for the next two to three years.

On the capabilities themselves, three things matter for the kind of work most lawyers actually do, and on these three Claude has a genuine edge that goes beyond benchmark scores.

Context window is the first. Claude Opus 4.7 and Sonnet 4.6 offer a one-million-token context window through the API and enterprise tiers, roughly 1,500 to 2,000 pages in a single session. For transactional lawyers, this means you can load an entire deal binder, a full set of disclosure schedules, or a substantial discovery production into one session and work across documents without the model losing track of defined terms, cross-references, or interdependent provisions. Anyone who has tried to do cross-document analysis with models capped at 128,000 tokens knows that the difference is quite significant. It changes the kinds of questions you can ask and the kinds of answers you can trust.

Writing quality is the second, and while this sounds subjective, its practical impact is measurable in hours. Claude produces prose that reads closer to how a senior associate actually drafts than any other model I have worked with. Client memos, executive summaries, board-ready translations of dense regulatory analysis, contract drafting. The first draft consistently requires lighter editing. In transactional practice, a surprising share of lawyer time goes into converting technical legal analysis into something a deal team can act on.

The connector architecture is the third, and it is the one that addresses the problem that has dogged every legal AI tool since the first hallucinated citation made the news. Anthropic’s approach is what it calls grounding. The MCP connectors allow Claude to pull from live, verified sources rather than generating answers from training data. In practice, this means the model can reach into Westlaw for primary law, CourtListener for court opinions, iManage for a firm’s own document repository, and NetDocuments for institutional knowledge. The difference between an AI that reads a real document and one that synthesizes from memory is the difference between a tool you can trust on a live matter and one you use for brainstorming and hope for the best.

All of that said, there is a meaningful gap between what Claude for Legal can do in a demo and what it takes to use it well on real work. This is where I part ways with most of the commentary I have read over the past two weeks. I will continue to insist on the role the lawyer plays and why learning how to use these tools is not only important to get an edge, but also to know where their limits are.

Claude for Word can now redline a contract against market positions, flag non-standard terms, and generate markup with tracked changes. For a first pass on a standard NDA, a services agreement, or an employment contract, the output is genuinely usable. But in a negotiated transaction, the markup is only the surface. The real negotiation is happening in what you choose not to mark up, which provisions you let go to signal flexibility, which ones you push on to test the other side’s priorities, and how you sequence your requests to preserve leverage for the issues that actually matter to your client. A model that redlines everything that deviates from market is technically correct and strategically useless. Knowing when to leave something alone is a skill that no benchmark measures.

This is not a defense of the status quo or an argument that lawyers should feel comfortable ignoring these tools. The productivity gains on pattern-recognition work are very real. Due diligence checklists, initial contract redlines against standard playbooks, research memos on settled legal questions, disclosure schedule extraction, regulatory filing summaries. These are hours that can be redirected toward work that actually requires judgment, and firms that are not redirecting them are falling behind. The point is that the same quality that makes Claude so effective on routine work, its confident, authoritative prose, is exactly what makes it risky on non-routine work. The output does not flag its own uncertainty the way a careful associate would hedge a memo on a novel question. It sounds the same whether it is right or subtly wrong, and in high-stakes practice, subtly wrong is the most dangerous kind of wrong there is.

There is also a governance dimension that the launch excitement has largely glossed over. In February, Judge Rakoff ruled in United States v. Heppner that exchanges between a criminal defendant and Claude’s consumer platform were protected by neither attorney-client privilege nor the work product doctrine. What happened was that Heppner used a consumer version without counsel’s direction, and Anthropic’s privacy policy at the consumer tier allows data collection and potential disclosure to third parties, including governmental authorities. The ruling is not a categorical prohibition on AI-assisted legal work. But it established that how a lawyer or client engages with these tools, the tier of service, the data handling terms, whether counsel directed the use, affects whether fundamental legal protections survive. The same week, a Michigan court in Warner v. Gilbarco reached the opposite result on work product for a pro se litigant’s AI-generated documents, applying a more functional analysis. The two decisions are not compatible in their doctrinal frameworks, which means the law on AI and privilege is not settled. It is actively being litigated, and firms that have not built governance policies around these distinctions are taking on risk that no product capability can offset.

For the mid-market firms and in-house teams that make up most of this readership, the practical question is how to adopt these tools at a cost structure and governance standard that makes sense for the work you actually do.

Claude for Legal’s plugins and connectors ride on top of standard paid Claude plans rather than a separate legal product with its own pricing tier. That means the entry cost is meaningfully lower than a dedicated Harvey or Legora deployment. For a mid-market firm without a large innovation team, the access Claude provides means a huge deal. The Microsoft 365 integration means lawyers can work within Word and Outlook without changing their daily workflow. The plugins are open-source and extensible, which is a strength for firms with technical capability and a real limitation for firms without it. Artificial Lawyer noted after the launch webinar that using the plugins, even in the no-code environment, will require more technical support than many lawyers can provide for themselves. Freshfields has a dedicated innovation lab. Quinn Emanuel has the resources to co-develop agentic workflows with Anthropic directly. Most firms are not Freshfields or Quinn Emanuel.

This is where Harvey and Legora retain an advantage that the Claude for Legal announcement does not erase. Both companies have built organizations around implementation, not just capability. Harvey embeds former lawyers within client firms to build custom agents tailored to specific practice groups. Legora has grown to 400 employees across six offices with forward-deployed teams that manage the full lifecycle from exploration through rollout. When a firm licenses Harvey or Legora, it is buying more than software. It is buying the institutional translation layer that turns a powerful model into something an M&A closing team or a regulatory compliance group actually uses consistently. Anthropic is a research lab that has built an extraordinarily powerful model. That is not the same thing as understanding that the real estate partner and the M&A associate have fundamentally different workflows even when they are using the same tool.

There is also the pricing question that connects to what I have written about before. As I argued in previous essays, the firms that translate AI-driven efficiency into competitive pricing will build a strong advantage over those that simply pocket the savings. Claude for Legal makes this question more urgent because it lowers the cost of the efficiency itself. If the tools to accelerate diligence, contract review, and regulatory analysis are available on a standard subscription rather than a five-figure monthly platform fee, the firms that compete on price while maintaining quality will have more room to do so.

If I strip away the market noise, what happened in mid-May is that the foundation model companies decided legal work is valuable enough to compete for directly rather than leaving it to the vertical startups and incumbents. Anthropic moved first with Claude for Legal. OpenAI and Microsoft are following with Codex for Legal and Copilot. The specialized legal AI companies are well-funded, well-positioned, and building real institutional relationships with their clients. But they are now operating in a market where their infrastructure providers are simultaneously their competitors, their partners, and their distribution channels.

For practicing lawyers, the practical takeaway is more immediate and less dramatic than the headlines suggest. Claude for Legal is the most capable general-purpose AI tool available for legal work today. Its context window, writing quality, and connector architecture give it real advantages on the tasks that consume the majority of a transactional lawyer’s repetitive hours. It is not a replacement for Harvey or Legora if what you need is a managed, deeply integrated deployment with dedicated implementation support. It is not a substitute for the judgment that experienced attorneys bring to the work that actually determines outcomes. And it is not exempt from the privilege, governance, and ethical questions that every AI tool in legal practice must answer.

What’s clear is that Anthropic has put itself in a position where it wins regardless of which legal AI company its customers choose. Harvey is built on Claude. Legora is built on Claude. Eve, Solve Intelligence, and a growing list of smaller players are built on Claude. Thomson Reuters integrated CoCounsel with Claude. And now Anthropic is also selling directly to the same firms those companies serve. It is the supplier, the partner, and the competitor simultaneously, and every connector it adds to Claude for Legal makes the direct offering a little more attractive relative to the intermediaries that depend on its model. The legal AI companies will tell you this is a healthy ecosystem. Some of them may even believe it. But the economics of platform dependency have played out in other industries, and the pattern is not kind to the companies building on top. Whether that dynamic ultimately reshapes the legal AI market depends on a question that has nothing to do with model quality or benchmark scores. Would law firms rather assemble their own AI stack from a powerful general-purpose platform, or pay someone else to assemble it for them? My instinct, having watched how this profession adopts technology, is that most firms will pay. Not because they lack the capability to build, but because the cost of getting it wrong on a live matter is high enough that the implementation layer matters. If that is right, Claude for Legal’s most lasting impact may not be as a product that lawyers use directly. It may be that it raised the floor for what every legal AI tool is expected to do, and that benefits every lawyer in the market whether or not they ever open a Claude subscription.

Adoption vs Integration

AI adoption and AI integration are not the same thing, and most of the confusion in the legal market comes from treating them as if they were. AI adoption means that lawyers at the firm use AI tools. They open Claude, ChatGPT, Harvey, Legora or CoCounsel, they prompt it with a question or a document, and they take whatever the model produces and work it into their output. This is useful. For individual tasks, particularly legal research, contract summarization, and first-pass drafting, it can be genuinely productive. But it does not change how the firm operates. The work still enters the same way, moves through the same steps, gets reviewed by the same people in the same sequence, and gets priced on the same basis. The AI is an accessory to an unchanged process.

AI integration is a whole other story. Integration means the firm has redesigned how work enters, how it gets scoped, how documents are produced and reviewed, how precedent is retrieved and reused, how matters are priced, and how institutional knowledge accumulates. AI is embedded in the operating layer of the firm rather than sitting on top of it as an optional enhancement. Clearly, the returns are categorically different. Adoption produces incremental efficiency gains for individual lawyers, maybe 10% to 20% on specific tasks. Integration produces compound gains across the firm. Faster turnaround, more consistent quality, better pricing accuracy, and the ability to handle more matters without proportionally increasing headcount.

The legal AI vendor market has inadvertently reinforced the adoption-over-integration dynamic. Harvey, Legora, CoCounsel, and the growing ecosystem of legal-specific AI platforms are built to be adopted. They are designed so that an individual lawyer can log in and start prompting. That is a reasonable product strategy, but it optimizes for the wrong unit of change. The relevant unit is not the individual lawyer’s productivity on a single task. It is the firm’s capacity to deliver legal services at a given quality level, at a given speed, for a given cost. Changing that requires changing the system, not just adding a tool.

The Case for General-Purpose Models

The legal AI market has consolidated around the assumption that law firms need legal-specific AI platforms. The pitch makes sense. Legal work is specialized, the stakes are high, and you need a model trained on or fine-tuned for legal content to get reliable results. There is some truth here. Legal-specific platforms offer curated databases, jurisdiction-aware search, and pre-built workflows for common legal tasks. For large firms with substantial budgets and dedicated legal technology teams, these platforms may be worth the investment.

For most firms, the more important question is what general-purpose models can already do, because the answer has changed dramatically in the last eighteen months. A model like Claude, which is the one I use most and will refer to throughout this essay, operating with a 200,000-token context window, can ingest a full purchase agreement, a disclosure schedule, and a set of prior transaction documents simultaneously and produce analysis that would have taken a junior associate a full working week. It can hold an entire firm’s precedent library in context, compare clause language across dozens of agreements, and flag deviations from house style with high reliability. The quality of output on legal reasoning, contract analysis, and document drafting tasks from general-purpose frontier models now matches or exceeds what most legal-specific wrappers produce, because most of those wrappers are themselves built on the same underlying models.

The three things that legal-specific platforms genuinely sell, beyond the model itself, are workflow structure, governance frameworks, and confidentiality assurance. These are real and important. But they are also things a firm can build for itself, often more effectively than a vendor can build them generically. A workflow designed around your firm’s actual matter types, your review standards, and your clients’ specific requirements will outperform a generic platform workflow. A governance framework written by the partners, reflecting the ethical obligations and risk tolerance of the firm, will be more rigorous than one generated by a vendor’s onboarding wizard. And confidentiality, the most critical concern, is increasingly addressed at the platform level. Claude’s professional plans, for example, offer zero data retention, enterprise-grade encryption, and explicit contractual commitments about data handling that satisfy the requirements of ABA Formal Opinion 512.

Then there’s economics. A Claude Pro subscription costs $20 per month per user. A Claude Max subscription, which provides extended context, higher usage limits, and access to Cowork, costs $100 to $200 per month. Harvey and comparable platforms typically run $150 to $300 per user per month or more, often with annual commitments and minimum seat counts. For a ten-lawyer firm, the difference between $2,000 per month and $20,000 per month is the difference between an accessible infrastructure investment and a significant fixed cost that requires justification against uncertain returns. The general-purpose model path lets firms invest the savings in the thing that actually drives integration: building the internal architecture that makes AI useful at a systems level.

What an AI-Integrated Firm Actually Looks Like

An AI-integrated law firm is not a firm where every lawyer uses AI. It is a firm where AI is woven into the operating model so deeply that it changes the client experience and the economics of service delivery. This translates into faster turnaround times, because work products move through AI-assisted pipelines rather than sitting in a queue. More consistent quality, because review protocols are standardized and AI-assisted rather than dependent on whichever associate happens to be available. Transparent, outcome-oriented pricing, because the firm understands its own cost structure well enough to price on value rather than time. And direct senior-lawyer involvement on every matter, because AI has absorbed enough of the production work to free senior practitioners from the volume constraint that historically forced delegation.

That last point deserves emphasis because it inverts a dynamic that has defined law firm economics for decades. In the traditional model, senior partners are capacity-constrained. They can only touch so many matters, so they delegate extensively to junior lawyers, whose work they review at the end of the process. The client pays for both the junior lawyer’s production time and the senior lawyer’s review time, and the senior lawyer’s actual involvement on any given matter is often limited to a few critical junctures. In an AI-integrated model, the senior lawyer can be more present throughout the matter because AI handles the production work that would otherwise consume their time or require delegation. The client gets more senior attention, not less, and the firm can support this model at price points that are competitive with, or lower than, traditional billing.

Humans are still necessary. This is worth stating directly because the AI discourse tends to oscillate between two poles: AI replaces lawyers entirely, or AI is just a fancy spell-checker. Neither captures what is actually happening. In an integrated model, the lawyer’s role shifts from production to supervision, judgment, and client counsel. The lawyer reviews AI-generated first drafts against their professional judgment. The lawyer makes the strategic call about deal structure, risk allocation, and negotiation positioning. The lawyer explains to the client why a particular clause matters and what the practical consequences are. AI cannot do any of this. What AI can do is ensure the lawyer spends their time on these high-value activities rather than on the mechanical work that surrounds them.

The Operating Layer: Seven Workflow Domains

Integration happens across specific workflow domains, and the practical question for any firm is which domains to redesign first and how to sequence the work. Seven areas represent the core of a law firm’s operating layer, and each one is amenable to AI integration with current general-purpose models.

Intake and matter scoping. This is where most firms should start, because it produces visible results quickly and does not require attorney judgment at every step. An AI-assisted intake workflow screens incoming inquiries against the firm’s practice area criteria and conflict parameters, routes qualified matters to the appropriate attorney, and generates a preliminary scope memo that includes relevant jurisdiction, estimated complexity, comparable prior matters from the firm’s own records, and a draft engagement framework. The attorney reviews the scope memo, adjusts it based on their professional assessment, and proceeds from a structured starting point rather than a blank page. According to data, intake-to-engagement automation is the highest-ROI starting workflow for small firms, producing trackable results without requiring deep technical implementation.

Document production and drafting. Although this is the domain where AI’s impact is most immediately visible, it’s also where firms most frequently stop short of real integration. The adoption approach is to paste a prompt into Claude and ask for a draft. The integration approach is to build a structured drafting workflow. The firm maintains a precedent library organized by matter type, jurisdiction, and clause category. When a new matter requires document production, the AI pulls relevant precedents, generates a first draft conforming to the firm’s house style and the client’s specific requirements, and flags areas where the drafter needs to make a judgment call. The adoption approach saves time on an individual document. The integration approach changes the firm’s production capacity.

Due diligence and document review. The Harvard Law School Center on the Legal Profession study that found AI complaint-response systems reducing associate time from 16 hours to minutes is the extreme case, but more modest gains of 40% to 70% time reduction on document review are well documented. For transactional firms, an integrated diligence workflow means the AI conducts the first-pass review of a data room, surfaces issues organized by risk category, cross-references findings against the firm’s diligence checklist, and produces a structured memo that the reviewing attorney can validate rather than produce from scratch. The attorney’s role becomes quality assurance and judgment, not data processing.

Precedent management and institutional learning. Every firm generates institutional knowledge through its work. How it structures particular provisions, what negotiation positions it takes in specific deal types, what issues arise repeatedly in certain industries. Almost no small or mid-sized firm captures this knowledge in any structured way. It lives in partners’ heads and in document management systems organized by client name rather than by substance. An integrated precedent system uses AI to index the firm’s work product by clause type, deal structure, jurisdiction, and outcome. When a lawyer starts a new matter, the system surfaces relevant prior work, identifies the firm’s typical approach, and flags where the current matter deviates from established patterns. Over time, this creates a genuine competitive advantage, the firm’s AI-assisted workflows get better as the firm’s knowledge base deepens.

Review and quality control. ABA Formal Opinion 512 established that lawyers using AI must understand the technology’s capabilities and limitations, maintain competent supervision, and ensure accuracy. An integrated review protocol means the firm has defined standards for AI-assisted work product: what gets checked, by whom, against what benchmarks. Rather than each lawyer making ad hoc decisions about when and how to verify AI output, the firm operates a consistent quality framework that satisfies ethical obligations and produces auditable records of human review.

Pricing and matter economics. This is where integration enables a firm to shift from hourly billing to outcome-based and fixed-fee arrangements. The reason most small and mid-sized firms cannot offer credible fixed fees is that they do not have granular data on how long different matter types actually take, what the cost drivers are, and where the variance comes from. An AI-integrated pricing workflow tracks matter-level economics across the firm’s portfolio, identifies patterns in cost and time allocation, and generates data-driven scoping estimates for new matters. This is the infrastructure that makes alternative fee arrangements viable. Without it, fixed fees are guesswork. With it, they become a competitive advantage.

Client communication and reporting. An integrated communication workflow means the firm can provide clients with real-time matter status, structured updates at defined intervals, and transparent reporting on work performed and fees incurred. This does not require building a client portal from scratch. It requires systematizing the information that the firm already generates through its work and making it accessible to clients in a structured format.

Building the Architecture: Skills, Projects, and Proprietary Workflows

The practical mechanism for moving from adoption to integration is to convert the firm’s repeatable processes into structured, reusable workflows that AI can execute under attorney supervision. In the Claude ecosystem, the primary tools for this are Projects and Skills. A Project is a persistent workspace with defined context. You load the firm’s precedents, style guidelines, jurisdiction-specific requirements, and matter instructions into the project, and Claude operates within that context for every task in that workspace. A Skill is a more structured unit. A set of instructions, templates, and reference materials bundled together so that Claude executes a defined workflow on demand. Projects provide context while Skills provide process.

For a transactional practice, this might work as follows. The firm builds a Project for its M&A practice that contains its standard purchase agreement templates, its clause library organized by deal type, its diligence checklist, and its house style guide. Within that Project, the firm develops Skills for specific tasks. A diligence review Skill that processes data room documents against the firm’s checklist and produces a structured findings memo. A drafting Skill that generates first-pass agreements from a term sheet, pulling language from the firm’s precedent library and flagging non-standard terms. A closing checklist Skill that tracks conditions precedent and generates status reports. Each of these Skills encodes the firm’s own methodology, not a generic platform’s idea of how legal work should be done.

Cowork extends this further by allowing the firm to delegate multi-step tasks. Instead of prompting Claude with a single question, a lawyer can describe a complete workflow: review this data room, produce a diligence memo organized by our standard categories, flag high-risk items for partner review, and generate a preliminary issues list for the client call. Claude plans the sequence, executes each step, and delivers the finished work product for attorney review. This is closer to delegating work to a supervised associate than it is to using a search engine. The critical difference is that the firm controls the process definition. The workflow reflects the firm’s standards, not a vendor’s defaults.

The proprietary dimension of this architecture is underappreciated. When a firm builds its own Skills and populates its own Projects with institutional knowledge, it creates something that no vendor can replicate. That is an AI system that works the way the firm works. The firm’s precedents, its negotiation patterns, its risk assessment frameworks, its client-specific preferences, all of this lives in the firm’s AI environment and makes the system more useful over time. This is a genuine moat. A competing firm can subscribe to the same AI platform, but they cannot replicate another firm’s institutional knowledge layer. For small and mid-sized firms in particular, this is where the competitive advantage sits. Not in having the most expensive AI tool, but in having the most thoughtfully constructed AI operating system.

An Implementation Roadmap

The sequence matters more than the speed. Firms that try to integrate AI across every workflow simultaneously end up with fragmented, inconsistent implementation that lawyers do not trust. Firms that start with a single workflow, prove the value, and expand methodically build durable integration that compounds over time.

The first month should focus on foundations. The firm should write its governance framework, because informal AI use does not scale. It should test the leading general-purpose models, select the one or two it will actually use, and configure the chosen platform in a way that satisfies the firm’s professional responsibility obligations on confidentiality, data handling, supervision, and client information. At the same time, the firm should identify the first workflows to redesign. For most firms, intake and matter scoping is the right place to start. It is frequent, relatively low-risk, and produces visible benefits across the firm.

Then it’s time for the firm to build and test those first few integrated workflows. This means creating the Project workspace with the firm’s relevant precedents and standards, developing the Skill or structured prompt sequence that executes the workflow, running it on live matters under close attorney supervision, and iterating based on what works and what does not. The firm should expect the first version to be imperfect. The goal is not perfection on day one. The goal is a working system that lawyers trust enough to use consistently and that the firm can improve over time.

From months three to six, the firm can expand into a broader workflow domain. The natural next step after intake is document production, because it relies on the same precedent library, style rules, and matter context. After that comes review and quality control, because formal review protocols create the trust infrastructure that lets the system scale. By the end of this phase, the firm should have three to five integrated workflows in production, a governance framework tested through actual use, and a growing base of institutional knowledge that makes the system more useful with every matter.

Only then should the firm move into pricing and client-facing communication. These are the domains that can change the firm’s market position, but they require internal data first. After a few months of AI-assisted workflows, the firm starts to understand how long matters actually take, where the cost drivers sit, and how much variance to expect. That gives it a stronger foundation for fixed-fee or outcome-based arrangements. It can price with more confidence, report progress with more structure, and show clients that AI is not just reducing time. It is making the delivery model more disciplined, transparent, and predictable.

The Competitive Landscape and the Window

The data on the competitive dynamics is unambiguous. Firms with extensive AI integration report 65% capacity gains and 44% improvements in client satisfaction, according to Clio. The 30% of mid-sized firms that report difficulty integrating new technology into existing workflows are not facing a technology problem. They are facing a process design problem. They bought tools without redesigning the work those tools are meant to support.

For small and mid-sized firms, the current moment represents a genuine opportunity, because the gap between what general-purpose models can do and what most firms are actually doing with them is enormous. The large firms are investing heavily in proprietary platforms and legal engineering teams. They are building at scale. But they are also building within the constraints of large institutional structures. Partnership governance, legacy technology stacks, and practice cultures that resist process standardization. A 15-lawyer firm can move faster, experiment more freely, and redesign workflows without navigating the politics of a 500-partner institution. That agility is a real advantage, but only if the firm uses it.

As AI capabilities continue to advance, the cost of catching up increases. A firm that starts building its integration architecture now is accumulating institutional knowledge, refining its workflows, and developing proprietary Skills that become more valuable over time. A firm that waits another two years will face a larger gap against competitors who started earlier, a more crowded vendor market that makes evaluation harder, and clients who have come to expect AI-integrated service delivery from their legal providers.

The real obstacle for most firms is not technology, budget, or even skepticism about AI. It is the gap between knowing that AI matters and knowing what to build. That gap is closeable, and it does not require a massive investment or a team of engineers. It requires a clear-eyed assessment of how work actually moves through the firm, a willingness to redesign the processes that no longer serve the firm’s interests, and the discipline to build systematically rather than chase the latest product announcement.

This essay covers the framework, not every implementation detail. Those details depend on a firm’s practice areas, client base, matter types, and the way work actually moves through the organization. The principles, however, are the same. Map the work, identify the workflow domains where AI creates genuine leverage, build the governance layer, and sequence the implementation so that each phase creates the foundation for the next. But translating those principles into a working system for a specific firm requires going deeper than any single essay can.

I have spent the last year studying and building these integration architectures. The firms that get this right tend to share a few habits. They start with workflows, not product demos. They treat governance as the thing that makes AI usable rather than a compliance exercise. And they understand that the proprietary advantage comes from encoding their own institutional knowledge into AI-assisted systems, not from buying the most expensive platform in the market. None of this is instant. Firms still need to do the work. But the work compounds and, over time, the firm starts to build something more valuable than AI access: a system that gets better with every case and every transaction.

If this resonates, and particularly if you find yourself wanting the more granular implementation detail behind each of the workflow domains discussed here, there is more to come. I will continue covering the dynamics of AI integration in legal services. And for firms that want to move from framework to execution, I am always happy to talk through what that looks like in practice.

Cross-border middle-market M&A is growing faster than the overall market. A Vlerick Business School study analyzing over 36,000 completed deals in 2024 found that cross-border mid-market transactions outperformed the broader M&A market in both volume and deal value growth. Nearshoring is pushing record foreign direct investment into Latin America. PE platform strategies are generating cross-border add-on activity that barely existed a few years ago. The succession wave is producing targets that attract buyers from outside the seller’s home jurisdiction. As a result, U.S. mid-market deal lawyers are seeing these transactions more often, usually in legal systems built on assumptions very different from the ones they trained in.

The challenge is that the entire system, the way courts interpret contractual obligations, the way regulators engage with transaction approvals, the way risk allocation concepts that US lawyers take for granted translate (or fail to translate) into a civil law framework, operates differently. A US purchase agreement is a product of decades of precedent-based refinement. Every provision has been tested in court. The concepts have settled meanings. Terms carry legal content shaped by case law. Move that same agreement into a civil law jurisdiction, put it before a court with no tradition of interpreting M&A-specific contractual language, and those carefully drafted provisions may produce outcomes that nobody at the negotiating table anticipated.

This essay looks at some of the points where that friction shows up in practice when a U.S.-trained deal lawyer works with local counsel on an acquisition in a civil law market. It is not a survey of regulatory differences or a checklist of foreign legal requirements. It is a lawyer’s account of where the two systems collide and what the attorneys who handle these transactions well do differently.

The US private M&A market is, by any measure, the most developed transactional market in the world. The documentation is standardized to a degree that no other jurisdiction matches. The industry has created a shared understanding of what “market” means across a wide range of negotiating points. Indemnification caps, basket structures, survival periods, materiality qualifiers, knowledge standards, closing conditions. When a US buyer’s counsel and a US seller’s counsel negotiate a purchase agreement, they are usually working from a common playbook. They may disagree on where to land, but they share a vocabulary, a set of benchmarks, and an institutional understanding of how the document is supposed to work.

That infrastructure is often missing in the markets where many cross-border mid-market deals now take place. The local M&A team may be highly competent, often staffed by lawyers trained at top U.S. or European schools and fully capable of handling complex transactions. But the broader ecosystem, the courts, the regulators, the accounting profession, the deal insurance market, and sometimes even the counterparty, does not have the same depth of transactional experience. In the US, a judge interpreting an indemnification provision in a purchase agreement has decades of Delaware precedent to draw on. In many civil law jurisdictions, the judge hearing a post-closing dispute may have little or no exposure to a purchase agreement at all. Terms that US lawyers rely on, concepts refined through years of litigation and market practice, land in a legal system that does not have the interpretive framework to apply them the way the lawyers intended.

This is not a criticism of the local legal system. Civil law jurisdictions have their own sophisticated frameworks for commercial transactions, rooted in statutory codes rather than case law. The problem is that a U.S.-style purchase agreement carries assumptions about interpretation and enforcement that do not always travel well. A U.S. lawyer who misses that point will draft provisions that look familiar and prove thin when tested. A U.S. lawyer who understands it will work with local counsel to translate the client’s protections into mechanisms the local system will actually recognize, apply, and enforce.

Some of this is abstract until you see it in a specific provision. A few concepts come up on almost every cross-border deal where the U.S. draft meets a civil law system, and they illustrate the problem better than any general description. The example I always use is efforts covenants. It may be the obvious one, but it captures the problem almost perfectly. U.S. purchase agreements rely on a hierarchy of efforts standards: “best efforts,” “reasonable best efforts,” “commercially reasonable efforts,” and “reasonable efforts.” In the U.S., those phrases are not the same. Case law has given them meaning, and the choice between them signals how much the obligor is expected to do, and how much of its own commercial interest it may protect. In many civil law systems, that hierarchy does not exist. Local law may impose a general duty of good faith or diligent performance, but it usually does not recognize the same calibrated scale that U.S. lawyers use. A court in a civil law jurisdiction may still enforce the clause, but not with the assumptions a U.S. lawyer had in mind. That is the problem. The issue is not that the provision is unenforceable. It is unpredictable, which for transactional purposes is nearly as bad.

Even when the purchase agreement is governed by New York law, the deal does not become a New York transaction. The asset, the target, the corporate records, the employees, the permits, the regulatory interfaces, and often the enforcement risk still sit in a local civil law system. That creates a false sense of comfort for U.S. counsel. New York law may govern the contract, but it does not automatically govern the steps required to perfect title, implement a reorganization, obtain third-party consents, enforce restrictive covenants, collect against local assets, or make a local regulator behave as the signing model assumed it would. A buyer may have a clean New York-law indemnity claim and still face a slower, messier reality when the practical remedy depends on local courts, local corporate formalities, or local administrative authorities. The real challenge is that the governing law clause can settle the interpretation of the paper while leaving the operational life of the deal, and much of the actual execution risk, inside a different legal system altogether.

Regulatory burden and timing is another challenge. One of the most common execution failures on cross-border deals is simply getting the sign-to-close timeline wrong. US deal lawyers are accustomed to a regulatory environment that, at certain deal sizes, is largely permissive and approvals are generally predictable in timing. The US lawyer’s instinct is to model a sign-to-close interim process and timeline based on diligence completion and financing, with regulatory approvals as a background item.

In many of the jurisdictions where cross-border mid-market deals are concentrated, the regulatory burden is heavier, less predictable, and more consequential for deal timing. Competition authority filings may be required at lower thresholds than in the US. Foreign investment screening regimes may apply to transactions that would not trigger CFIUS. Industry-specific regulatory approvals, in sectors like healthcare, financial services, insurance, telecommunications, energy, and education, may require affirmative clearance before closing. And the response times from local regulators are often longer, less predictable, and less responsive to the deal team’s preferred timeline than US lawyers expect.

All of this plays out through people, not just legal systems. And the dynamic that actually determines how these issues get resolved on most cross-border lower middle-market deals is the relationship between the advisory teams. In many of these transactions, the seller keeps only local counsel, and that counsel may have little to no transactional experience. The buyer, by contrast, often comes in with U.S. lead counsel and a local sophisticated co-counsel. That alone creates a complicated dynamic between advisors. The buyer’s local co-counsel becomes the bridge between two legal systems and, at times, the intermediary between two advisory teams that are not really speaking the same language.

The translation role is literal in some cases and conceptual in all of them. When US lead counsel sends a 60-page purchase agreement drafted under New York law to the seller’s local lawyer, the seller’s counsel is reading a document built on concepts that do not exist in the same form under local law, and in some cases do not exist at all in their day-to-day practice. He may be an experienced commercial lawyer with knowledge of local corporate law, but they may have limited exposure to M&A documentation. At that point, the buyer’s local co-counsel often ends up doing far more than jurisdiction-specific support. They explain U.S. concepts to the seller’s lawyer, translate the seller’s concerns back to U.S. lead counsel, and flag the places where the draft assumes a legal or transactional framework the local system does not actually share.

That is unusually valuable work, and it requires a specific kind of practitioner. Not just someone who knows the local law, but someone who is genuinely fluent in both systems and can spot substantive mismatches. Just as important, it requires someone with the judgment and credibility to be trusted by both sides as a serious communicator, even though they formally sit on only one side of the table.

For US lead counsel, the choice of local co-counsel is one of the most consequential decisions on the deal. The selection criteria should not be limited to jurisdictional expertise. It should include cross-border deal experience, familiarity with US-style documentation, and the ability to engage substantively with both the US lead counsel’s approach and the seller’s local counsel’s concerns.

On a large-cap cross-border deal, there is enough structure to absorb this kind of friction. Specialist counsel on both sides of the table, deeper diligence budgets, developed regulatory workstreams, deal insurance. In the lower middle market, most of that is not there. The deal depends on a handful of people getting the translation right. And this is exactly the part of the market that is growing.

Nearly 25,000 cross-border mid-market deals were completed over the past six years, and that number is likely to keep rising. So this is not really about having some vague international exposure. U.S. deal lawyers need a practical understanding of how at least one civil law system handles transactional concepts, enough to know when to push, when to pause, and when local input needs to come in early. They need strong local co-counsel relationships built through repeat collaboration. And they need to be willing to adjust a U.S. purchase agreement to fit a different legal system, instead of assuming the local law will somehow adapt to the draft. Every cross-border deal needs someone who can see both systems clearly enough to connect them. That is a skill worth building, and increasingly, it is just part of the job.

My last essay on search funds and business succession got a strong response, so I want to go a bit more technical on a related idea I have been thinking about for some time. I would be interested to know whether it resonates with other advisors.

The locked-box mechanism, long standard in much of European private M&A, deserves much more serious attention in the U.S. and Latin American lower middle market, where it remains surprisingly rare. I think that might be a mistake. Here is why.

Two Ways to Set a Price

Every M&A deal has to answer a simple question: what is the business worth at the moment ownership changes hands? The headline number in the LOI is almost never the final number. Between signing and closing, and sometimes for months after closing, the actual amount the seller receives is determined by a pricing mechanism that allocates the economic risk of operating the business during the transition period. How that mechanism works has an outsized effect on deal outcomes, post-closing relationships, and the probability of disputes. Yet in most lower middle-market transactions, the choice of mechanism receives less negotiating attention than almost any other relevant provision of the purchase agreement.

Under completion accounts, the standard in the US, the parties agree on an enterprise value and then determine the equity value by reference to the target’s balance sheet at the date of closing. The buyer pays a provisional price on closing day, typically based on estimated working capital, net debt, and cash. Within 60 to 90 days, the buyer’s accountants prepare closing accounts reflecting the target’s actual financial position at the moment of transfer. If working capital is below the agreed target, the price adjusts down, dollar for dollar. If above, it adjusts up. The seller carries the economic risk until the keys change hands, and the buyer pays only for what it demonstrably receives.

Under a locked box, the price is fixed at signing by reference to an agreed historical balance sheet, typically the most recent audited accounts. From the locked-box date forward, the buyer is treated as the economic owner. The seller commits to a no-leakage covenant, meaning no dividends, no management fees, no value extraction beyond specifically permitted payments like ordinary-course salaries, pre-agreed distributions, or expense reimbursements. If the seller breaches, the leakage is repayable dollar for dollar and uncapped. There is no post-closing true-up. There are no completion accounts. The price was set at signing, and it does not change.

Both mechanisms translate an agreed enterprise value into an equity price that reflects the actual financial condition of the business. The underlying economics are the same. The difference is in the timing, the information dynamics, the allocation of risk during the transition period, and, critically, the probability and cost of post-closing disputes. Those differences matter far more in the lower middle market than most deal commentary acknowledges.

Where the Markets Stand

The geographic split is hard to miss. In Europe, the locked box is the dominant mechanism in PE exits and competitive auction processes, used in roughly 60% of deals without a traditional purchase price adjustment. The CMS European M&A Study, covering 582 transactions across 27 jurisdictions in 2024, reported increasing locked-box adoption in non-PPA scenarios. In the UK, the locked box has become the default for auction processes and is increasingly adopted by trade sellers. Across the continent, the mechanism is standard in the Nordics, Germany, and France.

In the United States, the locked box is virtually nonexistent. A recent Law360 analysis noted that despite being well understood by US deal lawyers for over two decades, the mechanism remains rare in domestic transactions. The ABA Private Target Deal Points Study found that 86% of US private target transactions incorporate post-closing purchase price adjustments, the vast majority of which are working capital true-ups. In Latin America, completion accounts are the default by even wider margins, imported from US deal practice without much debate about whether the mechanism fits the local context.

The European picture is not a one-way trend, and intellectual honesty requires acknowledging this. The CMS 2024 study noted that completion accounts have regained prominence in the European mid-market, now appearing in 48% of total deals, and that locked-box usage has actually declined in smaller transactions. Why? Because at the lower end of the market, the financial reporting quality of the targets often does not support a locked box. When the seller’s accounts are informal, unaudited, or inconsistent, the buyer’s risk of relying on a historical balance sheet increases substantially. This is a real limitation, and it maps directly onto the lower middle-market environments in the US and Latin America where many founder-owned businesses operate.

But the decline in locked-box usage at the bottom of the European market does not weaken the case I want to make. If anything, it sharpens it. The locked box rewards preparation. A well-prepared seller that invests in clean accounts, quality-of-earnings analysis, and structured vendor due diligence earns the right to demand price certainty. The mechanism fails not because of any inherent flaw, but when the financial reporting foundation is too weak to support it. The implication for practitioners in the US and Latin America is not that the locked box is unsuitable for their markets. The implication is that the advisory ecosystem needs to invest in the preparation that makes the locked box viable, because the alternative, the completion accounts process, imposes costs and risks on lower middle-market sellers that are disproportionate to the value at stake.

The Post-Closing Fight Nobody Planned For

Working capital adjustments account for more than half of all post-closing disputes in US private M&A, according to multiple deal studies. That has nothing to do with bad faith. The mechanism depends on definitions, accounting judgments, and measurement timing that are inherently contestable, and in the lower middle market the contest is structurally lopsided.

Here is what it looks like on a $15 million deal. The buyer and seller negotiate a working capital “peg”, usually a trailing 12-month average. That average may be distorted by seasonality, one-time items, growth trends, or pre-closing operational changes the seller made in anticipation of the sale. The peg itself is the first battleground, and it is typically finalized late in the process when deal fatigue is highest and both sides are inclined to accept vague language rather than delay closing.

After closing, the buyer prepares the completion accounts. The buyer now controls the books. The buyer’s accountants make the judgment calls on inventory valuation, the allowance for doubtful accounts, accrued liabilities, deferred revenue. These calculations involve estimates and assumptions that competent professionals can apply differently and arrive at materially different numbers. The seller reviews the buyer’s calculations, discovers a working capital deficit of $200,000 to $400,000 that was not anticipated, and faces an unpleasant choice. Accept the buyer’s number, spend $50,000 to $100,000 disputing it, or escalate to an independent accountant whose resolution process adds months and cost. On a $15 million deal where the total advisory fees already run 3 to 5% of transaction value, adding another point for a post-closing dispute that could have been avoided is a failure of deal design.

The information asymmetry is huge and favors the buyer. The buyer prepares the accounts, controls the timing, decides how to classify borderline items. The seller, who no longer has the books and has already transferred control, is in the position of challenging the buyer’s work product with less information and less access. In large-cap transactions both parties have teams that can engage on equal footing. In a $15 million deal where the seller is a founder who has never been through a sale process, the asymmetry is severe. And every practitioner who has advised on these deals knows that 70% of the time, the seller accepts the buyer’s calculation without dispute, often because contesting it costs more than the shortfall. The completion accounts mechanism, designed to protect buyers in complex transactions with volatile balance sheets, becomes a tool of asymmetric extraction when applied to deals where the seller is structurally outgunned.

I want to be precise about what I am claiming and what I am not. I am not arguing that buyers systematically use the completion accounts process in bad faith. Most do not. I am arguing that the mechanism creates incentives and information advantages that, in the lower middle market, disproportionately harm sellers who lack the resources, the sophistication, or the appetite to contest post-closing adjustments. And I am arguing that the cost of the process itself, the accounting fees, the legal fees, the management distraction, and the relationship damage of a post-closing dispute, is often irrational relative to the amounts at stake. When a dispute over $150,000 in working capital requires $75,000 in professional fees to resolve, the mechanism has failed its purpose.

Why Founders Need This

As I wrote in my last essay, the profile of the seller in a search fund acquisition or founder-led exit is remarkably consistent across the US, Spain, and Latin America. A founder between their 50s to 70s who built a profitable business over decades, has no viable internal successor, and wants to exit in a way that preserves the company’s culture and provides a clean break. These sellers are not financial sponsors optimizing for IRR. They are individuals making one of the most consequential decisions of their professional lives, and their priorities, roughly in order, are certainty that the deal will close, certainty about what they will receive, confidence that the buyer will take care of the employees and customers, and a clean break that allows them to move on without months of post-closing entanglement.

The locked box delivers on the first two in ways that completion accounts cannot. The founder knows, at signing, exactly what they will receive. No anxious 90-day wait wondering whether the buyer’s accountants will find a shortfall. No dispute over whether the trailing 12-month average was the right peg. No argument about whether the seller’s accounting treatment of a borderline item was “consistent with past practice.” The price was agreed. The economics are settled.

The clean-break benefit matters just as much. Under completion accounts, the seller remains economically entangled for months after closing. The escrow sits there as a reminder that the deal is not done. The true-up process requires the seller to engage with the buyer’s financial team, review calculations, hire accountants, navigate a dispute mechanism at exactly the moment they are trying to emotionally step away from a business they ran for 30 years. Every advisor on founder exits has seen deals where the post-closing adjustment process poisoned a relationship that was collaborative through signing and closing. The locked box eliminates this entirely. The economic relationship ends at closing, which is what the seller wanted all along.

Buyers benefit too, in ways that are underappreciated. Price certainty means the buyer can model returns and plan integration without a pending adjustment hanging over the financials. Valuable management time goes to customers and team-building rather than to an accounting process. In search fund acquisitions, where the searcher is stepping in as a first-time CEO and the first 90 days are critical, the distraction of a working capital dispute can be genuinely damaging. Several practitioners have noted that in exchange for accepting the locked box, buyers typically negotiate more robust warranty coverage on financial statements and current trading, which provides meaningful protection through a different, less adversarial channel.

The LatAm Case

The case for the locked box is arguably even stronger in Latin America. In my last essay I noted that Latin American search fund acquisitions already rely on seller financing and deferred payments because traditional bank debt is uneconomic at local rates. Earn-outs account for 8.2% of median deal value. These are markets where creative structuring is the norm. The completion accounts mechanism, imported wholesale from US deal practice, depends on financial reporting quality that many founder-owned businesses in the region do not have, dispute resolution infrastructure that is less developed than in the US, and professional advisory fees that represent a larger share of the deal.

A $200,000 working capital dispute on a $10 million deal in Mexico, resolved through an independent accountant process, can easily cost $50,000 to $75,000 in fees. If you add this to the total advisory fees, it is simply not a reasonable cost of doing business. Yes, the locked box mechanism requires the seller to invest upfront in financial preparation, clean reference accounts and, ideally, a sell-side quality-of-earnings report. But that investment replaces the post-closing dispute risk while simultaneously producing a better-prepared sale process that attracts better offers and closes faster.

For the Latin American advisory ecosystem, the opportunity is to frame the locked box not as a European import but as a structuring tool that aligns with commercial realities already present in the market. Sellers who want certainty, buyers who want clean integration, and a deal environment where the cost of post-closing friction is disproportionately high relative to the amounts at stake.

Getting Creative: The Hybrid

For practitioners looking to introduce locked-box economics without asking a buyer to accept the full European model on a first encounter, the hybrid structure is the obvious entry point. And it is already gaining traction in cross-border deal practice.

The deal uses a locked-box baseline, price fixed at signing, no-leakage covenant, no broad post-closing NWC true-up. But the parties carve out a limited set of pre-agreed balance sheet items, typically one or two lines where the buyer has a legitimate measurement concern, like an inventory count or a specific receivable, and subject only those items to a post-closing verification. Everything else stays locked. The effect is that you eliminate the broad NWC fight, which is where 90% of the cost and conflict lives, while preserving the buyer’s ability to verify the specific items that actually carry risk.

Think of it as completion accounts with a scope limit, or a locked box with a safety valve. The buyer gets targeted protection where the diligence left genuine uncertainty. The seller gets price certainty on everything else. The post-closing engagement, if it happens at all, is narrowly defined, faster, and cheaper than a full NWC dispute. Importantly, both parties agree upfront on exactly which items are subject to verification, which eliminates the definitional ambiguity that drives most completion accounts disputes in the first place.

In practice, these hybrids tend to settle into a few recognizable structures. The cleanest version is a locked-box with a narrow true-up for cash and debt. The seller still gets most of what makes the locked-box attractive. A price largely fixed by reference to historical accounts, a no-leakage regime, and less room for sprawling post-closing accounting arguments. The buyer, meanwhile, avoids taking blind risk on the two balance-sheet items most capable of moving real value between signing and closing. Other versions are even more tailored. Some deals leave the price locked except for one or two specifically negotiated variables, such as a tax reserve, a project liability, or a regulatory receivable. Others use a collar approach, where working capital is left alone unless it falls outside an agreed range. What matters is that the completion mechanism is not allowed to metastasize into a full reopening of the balance sheet.

The advisors who are starting to use these hybrid structures in cross-border deals report that they reduce negotiation time on the purchase price mechanism, lower post-closing advisory costs, and produce fewer disputes. The structures are bespoke, which means they require more thought upfront than a standard completion accounts template, but that upfront investment pays for itself many times over when the alternative is a $75,000 dispute over NWC that both parties could have avoided.

Where to Draw the Line

The locked box, whether full or hybrid, is not the right mechanism for every deal. Transactions with long sign-to-close gaps driven by regulatory approvals or financing contingencies shift too much interim risk to the buyer. Businesses with high seasonality or volatile working capital present a harder case, because the reference accounts may not capture a representative period of the financial cycle. And the financial reporting quality constraint is genuine. The locked box is only as reliable as the reference accounts it is built on. If the seller’s financials are unaudited, inconsistently prepared, or lacking the granularity to define and police leakage, the mechanism creates more risk than it eliminates.

The response to that last constraint is not to give up on the mechanism. A sell-side quality-of-earnings analysis, which is becoming standard practice in well-run sale processes regardless of the pricing mechanism, provides exactly the financial foundation that a locked box requires. The investment in preparation is modest relative to the post-closing costs it replaces.

The broader point is that the choice of pricing mechanism is a deal design decision, not a drafting exercise. It should be discussed at the LOI stage, not delegated to counsel after the commercial terms are set. It should reflect the specific deal, based on the quality of the financials, the expected timeline, the volatility of working capital, and what both parties actually care about most. The US market’s blanket default to completion accounts forecloses that conversation before it begins.

Europe has been using the locked box for two decades, across thousands of transactions, in the world’s second-largest M&A market. The mechanism is proven. The hybrid structures emerging in cross-border practice are making it accessible to markets that have never tried it. The deal types where it fits best, founder exits, succession transactions, search fund acquisitions, are the fastest-growing segment of lower middle-market M&A. For sell-side advisors, the locked box can be positioned as a competitive differentiator. The buyer who offers price certainty to a founder navigating the most emotionally charged transaction of their career has a real edge over a buyer presenting a provisional price subject to a true-up the founder does not fully understand. For buy-side advisors and search fund operators, the locked box signals seriousness. The price is real. There is no hidden adjustment waiting in the wings.

The tools are there. The question is how long certain markets will continue defaulting to a mechanism that, in the specific context of founder-led exits and succession transactions, serves neither buyer nor seller as well as the alternative that is sitting right there, already proven, waiting to be adopted.

Search funds sit at the intersection of entrepreneurship, private equity, and lower middle-market M&A, and they are reshaping how a generation of business owners exits the companies they built. The model is better understood in the US than anywhere else, but Europe (particularly Spain) and Latin America are catching up fast, and the structural conditions in those markets may actually be more favorable. This is an attempt to explain the mechanics, map the differences across geographies, and make the case that practitioners in the M&A ecosystem should be paying closer attention.

What a Search Fund Actually Is

A search fund is an investment vehicle through which an entrepreneur, typically a recent MBA graduate or an experienced operator, raises a small pool of capital from a group of investors to fund a search for a single company to acquire, operate, and grow. The model was conceived at Stanford GSB in 1984 by Irv Grousbeck and has since produced more than 700 tracked funds in the US and Canada alone, with another 320 internationally. In 2023, a record 94 core search funds launched in the US and 59 launched internationally, a pace that would have been unimaginable a decade ago.

The mechanics work in two phases. In the first, the searcher raises what is effectively seed capital, typically $400,000 to $700,000, sold as units to a group of 10 to 20 investors. This capital covers the searcher’s salary, travel, legal costs, and deal expenses during an 18- to 24-month search window. Investors at this stage receive the right, but not the obligation, to participate in financing the eventual acquisition. In the second phase, once a target is identified, the searcher raises acquisition capital from the original investor group (who hold pro-rata rights) and potentially new co-investors, layering equity with senior debt and, frequently, seller financing. The searcher then steps in as CEO and operates the business for a hold period typically ranging from five to ten years.

The targets are not distressed turnarounds or venture-scale bets. Search funds acquire profitable, cash-flowing businesses in the $5 million to $30 million enterprise value range, typically with recurring revenue, defensible market positions, and 20 to 50 employees. Healthcare services, business services, technology, and manufacturing are the most common sectors. The median acquisition in the US has a purchase price around $14.4 million, EBITDA margins of 22.5%, and a 7x EBITDA multiple. Internationally, the median sits at $11.7 million with slightly higher margins at 24%.

The returns have been remarkably consistent. Across all 681 tracked US and Canadian search funds since 1984, the aggregate pre-tax IRR is 35.1% with an ROI of 4.5x. Exited companies have achieved a 42.9% IRR. Nearly seven in ten acquired companies generated positive returns. For an asset class that operates in the lower middle market and relies on first-time CEOs running companies they did not build, those numbers deserve serious attention.

The Succession Gap

The search fund model works because of a structural imbalance that exists in every developed and developing economy: aging founders who built valuable businesses have no clear path to exit that preserves what they built. Traditional private equity is too large, too transactional, or too focused on financial engineering for businesses at this scale. Strategic acquirers often absorb the company into a larger platform, eliminating the identity and culture the founder spent decades creating. And family succession, statistically, fails more often than it succeeds.

The numbers on family business succession are strikingly consistent across the three markets that matter most for search funds. In Spain, over 85% of companies are family-owned, and fewer than 30% have a formal succession plan. A study from the Spanish Family Business Institute found that only 12% of non-member family businesses have any succession plan at all, and 35% of family offices expect a generational shift within the next decade. Eighty percent of Spanish family businesses do not survive to the third generation.

In Mexico, the picture is nearly identical in structure but arguably worse in execution. Family businesses represent over 85% of all companies, and 100% of the largest Mexican companies are family-controlled. According to INEGI, only three out of ten family businesses survive to the second generation. A 2023 study by the CIFEM-BBVA and IPADE found that 52% of family businesses face meaningful survival risk from internal governance failures, and only 3% have an explicit CEO succession plan. The IFEM report at Tec de Monterrey surveyed over a thousand business leaders across Mexico and Latin America and found that 86% of entrepreneurial families lack a transitioning plan for the next generation, 68% have no shared vision or only an informal one, and 74% have made no provision for the senior generation’s role after transition.

What makes these statistics actionable, rather than merely alarming, is that they describe a market failure with a known solution. The founder who built a $10 million revenue manufacturing business in Guadalajara or a $15 million logistics company in Barcelona faces a narrow set of options: pass it to a child who may not want it or be prepared for it, sell to a competitor who will strip the business for parts, hold on until declining health forces a fire sale, or find an operator who will buy the business, step in as CEO, and run it with the same care the founder did. Search funds are the only institutional model designed specifically for that last option.

Three Markets, Three Capital Stacks

The search fund community often discusses the model as though it operates the same way everywhere. It does not. The deal mechanics differ fundamentally across the US, Spain, and Latin America, and those differences shape the risk profile, the return dynamics, and the type of legal and financial advisory work the transactions require.

The United States: SBA Infrastructure and Standardization

The US search fund market benefits from a financing ecosystem that does not exist anywhere else: the SBA 7(a) loan program. In a typical US search fund acquisition at the lower end of the market, the capital stack follows an 80/10/10 structure. The SBA-backed lender provides roughly 80% of the purchase price as senior debt, the seller finances 10% through a subordinated seller note (typically on standby for 12 to 24 months post-closing), and the buyer contributes 10% as equity, often raised from a group of five to twelve investors. For larger traditional search fund deals, the structure shifts toward 30 to 40% senior debt, 50 to 60% investor equity, and 10 to 20% seller financing, but the SBA remains the anchor of lower middle-market acquisition finance.

This infrastructure matters more than most international practitioners appreciate. The SBA guarantee de-risks the transaction for the lender, enabling leverage levels and terms that would be impossible in a purely commercial lending environment for a first-time CEO acquiring a business. The all-in borrowing cost as of early 2026 sits around Prime plus 2.75%, with terms of seven to ten years. The personal guarantee requirement for anyone holding 20% or more of the equity creates a natural discipline in the capital structure and keeps investor stakes fragmented, but it also means the searcher bears meaningful downside risk.

The SBA framework has also created a degree of standardization in deal documentation that reduces transaction costs. Purchase agreements, investor rights, and the interplay between the SBA loan covenants and the equity structure follow recognizable patterns. Counsel on these deals, while not handling the complexity of a large-cap M&A transaction, needs to understand a specific and specialized regulatory overlay that most generalist corporate lawyers do not encounter. The SBA recently updated its Standard Operating Procedure regarding search fund eligibility, tightening rules around investor control provisions and the treatment of preferred equity with redemption features. These changes are creating real uncertainty in the market about which deal structures remain compliant, and the advisory work around structuring investor terms to satisfy SBA requirements while preserving investor protections is becoming increasingly nuanced.

Spain: Bank-Oriented and IESE-Anchored

Spain’s search fund ecosystem is the most developed outside North America, with 67 first-time funds launched to date, making it the single most active international market by fund count. The IESE ecosystem is the institutional backbone. IESE was the first business school outside North America to offer an ETA-focused elective, and over 60 IESE graduates have raised search funds across 20 countries. The biennial IESE International Search Fund Conference has become the primary networking event for the non-US search community, and the proximity of IESE’s International Search Fund Center to the Spanish market has created a density of institutional knowledge, investor relationships, and deal flow that no other European market can match.

The deal structures in Spain reflect the broader European financing environment: more conservative leverage ratios than the US, stronger reliance on traditional bank financing, and family offices playing a more prominent role as equity co-investors. There is no SBA equivalent in Spain, so the debt portion of the capital stack comes from commercial banks that apply conventional underwriting standards. This means lower leverage, often 40 to 50% of the purchase price, with the remainder split between investor equity and, increasingly, seller financing or deferred consideration. The due diligence processes tend to be longer and more thorough than in the US, partly reflecting the European M&A culture and partly because the targets are often family businesses with decades of informal governance, incomplete records, and intermingled personal and business assets.

From a legal perspective, the Spanish search fund transaction requires counsel to navigate the intersection of corporate acquisition law, family business governance, and increasingly, tax-efficient structuring for the seller. Many founders are selling not because they want liquidity but because they have no successor, and the emotional dynamics of these transactions differ materially from a PE buyout. The seller often stays involved in some capacity, whether through a consulting agreement, a board seat, or a retained minority stake, and the documentation needs to manage those ongoing relationships carefully. Succession tax reform in Spain has eased some of the fiscal burden on these transitions, but the planning remains complex.

Latin America: Creative Structuring Without Infrastructure

Latin America’s search fund ecosystem is younger but growing rapidly. Since the first recorded Latin American search fund in 2008, 169 funds have initiated capital-raising across the region, resulting in 59 acquisitions and 14 exits through mid-2025. Mexico led the early wave with 50 first-time funds and accounts for seven of the 15 known positive international exits, more than any other country outside the US. Brazil has surpassed Mexico in recent cohorts by total fund count, and Colombia, Chile, and Peru are showing increasing activity.

The deal mechanics in Latin America are fundamentally different from both the US and Spain, and this is where the practitioner perspective becomes essential. There is no SBA program. Local interest rates make traditional acquisition leverage uneconomic in most countries. Brazil’s benchmark Selic rate reached 14.75% in 2025, and Mexican rates, while lower, still price senior debt at levels that compress returns significantly if you try to replicate a US-style leveraged structure. Local banks typically require hard collateral, and many search fund targets are asset-light service or technology businesses where value sits in cash flow rather than tangible assets.

As a result, Latin American search fund acquisitions have evolved a distinct capital structure. Debt represents a median of only 36.5% of deal value, compared to 70 to 80% in the US. Earn-outs, rare in US search fund deals, account for a median of 8.2% of the purchase price. Seller financing and deferred payments play a much larger role, effectively substituting for the institutional debt infrastructure that does not exist. In Brazil specifically, vendor loans indexed to inflation or local benchmark rates have become the standard debt substitute, with meaningful portions of the purchase price paid over several years post-closing.

For M&A lawyers advising on these transactions, the structuring challenges are considerable. The capital stack is bespoke on every deal, with no standardized templates equivalent to the US SBA framework. Cross-border investors, who provide a significant share of the equity in Latin American search funds, often require USD-denominated structures or hedging mechanisms that add complexity and cost.

The searcher profile in Latin America is also evolving in ways that affect the transactions. Solo searchers now represent 78% of the 2023-2024 cohort, up from 41% in 2017-2018, and 51% of searchers have non-MBA backgrounds. A growing number of experienced operators and former founders are entering the model, bringing deeper industry knowledge but sometimes less familiarity with institutional deal processes. This creates both opportunity and risk for the advisory ecosystem.

What the Stanford and IESE Studies Do Not Capture

The biennial studies from Stanford GSB and IESE are invaluable for tracking the macro trajectory of the asset class. But there are dimensions of these transactions that the studies, by design, do not measure, and that practitioners should understand.

The first is the quality of due diligence. Search fund acquisitions are, by definition, transactions where a first-time buyer is acquiring a company from a founder who may have run the business informally for decades. Financial statements may be unaudited or prepared on a cash basis. Customer concentration may be higher than reported. Key-person risk, usually centered on the departing founder, is the single largest operational risk in these deals, and it is notoriously difficult to diligence. The searcher is often replacing the owner-operator entirely, which means the person who built the customer relationships, managed the key accounts, and embodied the company’s culture is walking out the door. Lenders and investors know this, and the best transactions build in structured transition periods, consulting agreements, and earnout mechanisms that keep the seller economically aligned through the handover. The worst transactions underestimate how much of the company’s value is tied to the founder’s personal relationships.

The second is the post-acquisition failure mode. The studies report that roughly 30% of acquired companies generate negative returns or a total loss. What they do not detail is the mechanism of failure. In my experience with lower middle-market transactions, the most common failure modes are not strategic miscalculation or market collapse. They are integration failures: the new CEO loses the trust of the existing team, a key customer departs because their relationship was with the founder, or the working capital dynamics of the business turn out to be more demanding than the financial model predicted. Good legal structuring can help mitigate them through more careful treatment of transition services, seller support, working capital design, earnouts, restrictive covenants, and information rights. But mitigation starts with recognizing that these are the pressure points most likely to matter after closing.

The third, specific to the international markets, is the exit environment. The US search fund market has a well-developed secondary market: financial sponsors, strategic acquirers, and even other search fund investors provide multiple exit paths. Internationally, the exit options are thinner. Only 21 exits have been recorded across all international search funds, and six were at a loss. For Latin American funds, the limited IPO market, the absence of a deep pool of domestic PE buyers at this deal size, and currency risk on USD-denominated investor returns create a more constrained exit environment. The Spectra Investments data shows 14 exits across all of Latin America through mid-2025, eight of which occurred in the past four years, suggesting the market is maturing but is still far from liquid.

What Smaller Markets Can Learn from the US

The US search fund market did not become institutional overnight. It took four decades, the sustained involvement of a single academic institution (Stanford GSB), and the development of a financing infrastructure (the SBA 7(a) program) that was not designed for search funds but turned out to be perfectly suited for them. The question for Spain, Mexico, Brazil, and the rest of Latin America is what elements of the US ecosystem are replicable and what elements require local substitutes.

The first lesson is the importance of a financing anchor. The SBA’s role in the US cannot be overstated. By providing a government-backed guarantee that enables 80% leverage on small business acquisitions, the SBA created the conditions for a standardized, repeatable deal structure that lowered the barrier to entry for searchers and reduced risk for investors. No Latin American country has an equivalent program. Mexico’s development banks, Nacional Financiera and Bancomext, have mandates that could theoretically encompass search fund acquisition financing, but neither has developed a product for this purpose. Spain has had some success with public subsidy programs for business succession, but nothing approaching the scale or specificity of the SBA. Developing a financing product, whether government-backed or through a consortium of private lenders, that enables higher leverage on small business acquisitions with cash-flow-based underwriting rather than hard collateral requirements, would be the single highest-impact intervention for growing the search fund ecosystem in these markets.

The second lesson is the role of business schools as ecosystem anchors. Stanford GSB and IESE have demonstrated that a single institution can catalyze an entire market by offering ETA-focused courses, producing research, hosting conferences, and creating networks that connect searchers with investors and operators. In Mexico, IPADE has begun hosting search fund forums and building awareness, and EGADE at Tec de Monterrey has the institutional capacity to play a similar role. But the coverage remains thin. The US has dozens of MBA programs that now teach entrepreneurship through acquisition as a distinct career path. Mexico and Brazil have a handful each, and the result is that the pipeline of prepared searchers is much narrower than the opportunity set warrants.

The third lesson, and the one most relevant to this audience, is the development of a specialized advisory ecosystem. In the US, a growing number of law firms, lenders, and advisory firms have built dedicated search fund practices. They understand the standard deal structures, the SBA regulatory overlay, the investor governance expectations, and the specific diligence issues that arise in founder-owned business acquisitions. This specialization reduces transaction costs, accelerates deal timelines, and improves outcomes. In Mexico and broader Latin America, the advisory ecosystem for search fund transactions is essentially nonexistent at scale. Searchers rely on generalist corporate lawyers who may be encountering the model for the first time, and the result is often higher legal costs, longer timelines, and deal structures that do not adequately protect either the searcher or the investors.

The fourth lesson is the importance of investor education and community. The US search fund investor base has expanded from a small group of Stanford-affiliated angels to include institutional investors, family offices, and former search fund operators who reinvest their proceeds. This depth of capital provides searchers with both financing and mentorship. In Latin America, Spectra Investments has emerged as the most active institutional investor, claiming involvement in more than half of all funds raised in the region, but the overall investor base remains thin and concentrated. Many potential investors in Mexico and Latin America, particularly family offices sitting on exactly the kind of capital that search funds need, are simply unaware that the model exists or how it works. The education gap is as much a constraint on growth as the financing gap.

The Opportunity

Search funds occupy a space in the M&A market that is structurally underserved by the existing institutional infrastructure. Traditional private equity firms focus on larger transactions because the economics of their fund structures require it. Investment banks focus on deals that generate fees proportional to their overhead. And the founders of $5 million to $30 million businesses, the ones actually facing the succession crisis, often cannot afford or do not know how to access the advisory services that would help them navigate a structured sale process.

This creates a genuine market opportunity for M&A practitioners, legal advisors, and capital allocators who are willing to develop expertise in this segment. The fee levels per transaction are modest compared to mid-market or large-cap M&A, but the volume, the repeatability of the deal structures, and the relationship dynamics, where the successful searcher becomes a repeat client who will eventually exit and potentially launch a second fund or become an investor, create a practice-building opportunity that is genuinely attractive for mid-sized firms and boutique advisory practices.

In Latin America specifically, the gap between the opportunity and the infrastructure is enormous. Ninety percent of Mexican companies are family-owned SMEs. The founding generation is aging. The succession planning deficit is well-documented and, if anything, worse than in Spain, which is already the second most active search fund market globally. The deal structures require creative financing solutions that reward sophisticated transactional advice.

The search fund model is not a panacea. Not every founder-owned business is a suitable target, not every searcher is a capable operator, and the returns, while strong in aggregate, include a meaningful tail of losses. The international markets in particular are still in the early innings. The exit data is thin, the financing infrastructure is immature, and the institutional knowledge base is concentrated in a handful of business schools and investor groups. But the underlying thesis, that there exists a large, growing pool of profitable, well-run businesses whose founders need an exit and whose scale makes them invisible to traditional private equity, is as sound in CDMX, Bogotá and Madrid as it is in Minneapolis. The question is not whether the opportunity is real. The question is whether the markets outside the US will build the infrastructure, develop the talent, and deploy the capital needed to capture it before the succession clock runs out.

I’d be happy to speak with founders, searchers, investors, or advisors who are also active in this space, or simply interested in the asset class. It sits in a part of the market that is still underserved, often misunderstood, and likely to matter more over the next few years than it does today.

Mid-market M&A deals are more complex than they were a decade ago. This is not a controversial observation. Anyone working on transactions in the $30 to $200 million range has watched the documentation get longer, the negotiation points multiply, and the ancillary workstreams expand. What is less obvious, and what I want to spend this essay on, is the consequence. The fee expectations around mid-market advisory work have not adjusted to reflect what the work now requires. There is a genuine pricing gap in this market, and it is not going to close on its own.

The Technical Complexity

In my view, several features once associated primarily with larger, more sophisticated transactions have now become standard in middle-market private M&A.

Representation and warranty insurance now appears in roughly 75% of PE-backed transactions and is available even in deals as small as $25 million to $30 million, with premiums that have fallen to around 2.5% to 3% of the insured amount. Its effect goes well beyond adding another product to the deal. It changes the structure of the transaction itself. In insured deals, the entire indemnification framework is reworked. Median indemnity caps fall to just 0.3% of purchase price, walk-away structures appeared in roughly a third of deals in 2024, and aligning the policy with the purchase agreement has become a substantial drafting and negotiation exercise in its own right.

Rollover equity is also standard in PE-backed acquisitions across the lower middle market, where add-on strategies account for more than 75% of buyout activity. That matters because rollover is not just a pricing tool. It is a governance, incentive, and risk-allocation mechanism wrapped into the same structure. It raises questions about valuation, liquidity, minority protections, tax treatment, and the extent to which a selling founder or management team will remain economically aligned with the sponsor after closing. In practice, those issues are rarely standalone. They tend to interact with earnouts, management retention, non-competes, and post-closing control arrangements, which makes the drafting and negotiation materially more intricate than the headline deal value might suggest.

Earnouts, meanwhile, now appear in more than 20% of non-life-sciences deals, with 68% using multiple performance metrics and median earnout potential rising to 43% of the closing payment.

Efforts standards in earnout provisions and regulatory-approval covenants require close calibration, because they often govern the most contested question in the deal after signing. How much a party is actually obligated to do to get to the intended result. Small drafting choices can materially shift both execution risk and economic outcome.

Break fees and reverse break fees now appear regularly in competitive mid-market auctions, especially in deals with a real gap between signing and closing. They are no longer features reserved for larger transactions. Once introduced, they require careful work around trigger events, fault allocation, financing failures, regulatory risk, and remedy structure. What looks like a single economic term often becomes a broader negotiation over execution certainty and the consequences of a broken deal.

Each of these features adds real legal engineering. In many deals, they appear together. These are fully negotiated, highly structured deal terms that once belonged mainly to larger transactions that now routinely appear in the mid-market.

The Practical Complexity

Technical complexity is only part of the story, and in some ways it is the easier part. A competent M&A lawyer can work through these structures without much difficulty. What makes mid-market practice genuinely more demanding than its large-cap equivalent has less to do with the deal structures themselves and more to do with the environment in which they operate.

Start with client sophistication. In a large-cap transaction, both sides of the table are usually staffed with experienced deal teams. The CFO has been through multiple transactions. The general counsel understands the structural features of the deal. The board has an established framework for evaluating the terms. Multiple advisors are running parallel workstreams in a coordinated way, and the process is generally clear to everyone involved. The legal work in that environment may be highly complex, but the clients themselves are sophisticated consumers of that advice. They know what representation and warranty insurance does. They have an informed view on earnout risk. They understand what they are trading away when they agree to a non-compete.

Mid-market transactions, particularly those involving founder-led businesses or first-generation sellers, frequently involve principals encountering these mechanisms for the first time. The seller may not understand the practical effect of an indemnification walk-away or what the R&W retention means for their actual exposure. The management team being asked to roll over equity may not grasp the tax consequences, the governance implications, or the difference between what they own today and what they will own after the restructuring. The buyer’s deal team on a lower mid-market add-on may have limited experience structuring earnouts or negotiating efforts covenants with real teeth.

This matters because the lawyer is not merely drafting and negotiating. The lawyer is also translating. Explaining what these mechanisms actually do, where the risks sit, how the leverage works, and what the client should be pushing for. That translation function takes real time. It also requires judgment. How far to simplify without distorting the substance, and how to equip the client to make decisions without overwhelming them with technical detail. The quality of that work can materially shape the outcome of the deal.

A first-time seller who does not understand deal dynamics may accept terms that a more seasoned party would push back on immediately. A management team that lacks a clear sense of market can leave protections on the table worth many times the legal fee. Preventing those outcomes is some of the most valuable work a mid-market lawyer does. It is also the kind of work that remains largely invisible in the way legal engagements are priced.

Then there is process coordination. Large-cap transactions usually run on top of a real execution infrastructure. Financial, legal, tax, and operational diligence are handled through dedicated workstreams, with a project-management layer keeping them aligned and clear protocols governing timing, sequencing, and information flow. In the mid-market, that infrastructure is often much thinner, if it exists at all. The deal lawyer frequently becomes the de facto project manager, coordinating with the financial advisor, the accountants, the insurance broker, and the client, managing the overall timeline, spotting issues that fall between advisory silos, and holding the process together in ways that extend well beyond the traditional boundaries of legal work. This happens not because it is formally assigned, but because nobody else is positioned to do it. In many mid-market deals, it is one of the conditions for getting the transaction across the line.

These are not edge cases. They describe the operating reality of mid-market M&A practice. The combination of structural complexity, inexperienced principals, and limited process infrastructure means that the mid-market advisor is frequently doing more work, across a wider range of functions, with less support, than a large-cap counterpart handling a transaction with comparable structural features. And the fee for that work is, in most cases, a fraction of what the same mechanisms generate at larger deal sizes.

The Value Proposition Problem

Most of the pricing pressures in the mid-market are not really within the control of mid-market advisors. Fee expectations are often shaped by historical reference points from earlier transactions, when the documentation was lighter, the processes were looser, and the overall advisory burden was lower. Those legacy benchmarks continue to influence what clients believe a deal should cost, even when the work now looks very different. At the same time, pricing pressure comes from both directions. Regional firms push aggressively on rates, while larger firms move down-market when they need to keep teams busy or preserve market share. The result is a fee market that reflects competitive tension far more than the real demands of the engagement. In many cases, the price is not being set by the actual complexity of the work, but by outdated expectations and market crowding.

There is, however, one important factor that advisors do control, and it is worth stating plainly. Many firms focused on the mid-market, whether true mid-sized firms or even top-tier firms pursuing that segment, still sell themselves through some version of the same message: “we bring large-cap deal quality and top-tier execution to the mid-market, at a more reasonable price”. That framing is well intentioned, and, in many cases, it is also true. A strong mid-market partner with fifteen years of transaction experience may be doing work every bit as rigorous, commercially sound, and technically sophisticated as a partner at an Am Law 25 firm.

The problem is that “same quality, lower price” is a value proposition that accepts a discount by design. It positions the firm as a substitute for something more expensive rather than as a specialist providing distinct value. And it ensures that every future fee conversation starts from the premise that the advisor should be cheaper, because that is what was promised.

Mid-market M&A is not simply a smaller-scale version of large-cap practice. It is a distinct discipline, one in which structural complexity intersects with operational difficulty in ways that larger transactions often do not. The financial information is usually less complete and less dependable. The principals are often less experienced. The execution infrastructure is thinner. And the margin for error is, in practical terms, narrower, because the value at risk often represents a much larger share of the client’s total wealth.

That changes the stakes of the advisory work. A seller who loses a $10 million earnout on a $60 million sale is not absorbing that loss in the same way a private equity fund does when it misses a $100 million earnout on a $2 billion exit. In the first case, the amount at issue may represent a meaningful portion of a founder’s life’s work. In the second, it is more likely to be one variable within a broader portfolio outcome. The advisor who structures the deal to protect against the first outcome is providing enormous value at a level that is often deeply personal to the client.

The firms that handle this well build the infrastructure needed to make the work visible, both internally and to clients. That means matter-level profitability data, phased fee structures built around clear scope assumptions, and the ability to explain to a PE deal team or a founder, in concrete terms, what the engagement actually requires and why a structurally complex transaction should not be priced like a simpler one. This matters even more in a market where trust around legal billing is weak. If only 30% of PE firms trust their outside counsel to bill accurately, and nearly 80% say legal spend lacks transparency, then the advisor who can show the work, explain the drivers of cost, and price the engagement in a way that feels disciplined and credible is differentiating itself through a form of clarity that ought to be standard, but still is not.

This is not primarily about raising rates across the board. It is about changing the conversation. About describing mid-market M&A practice for what it actually is, not what it looks like from the outside, and pricing the work based on its real complexity rather than the size of the transaction. The work has changed with it. Advisors who understand that, and are prepared to articulate it clearly to clients, will end up in a very different competitive position from those still relying on the language of discount.

Let me know where you agree or disagree in the comments.

Leave a comment

A few days ago, Anthropic released a labor market study that quickly turned into the usual debate about which jobs are most at risk from AI and whether the labor market is about to face a wave of disruption. Depending on which article you read, AI is either about to destroy large parts of the economy or has already been absorbed with little impact. Neither interpretation is particularly useful. The study itself is far more interesting.

The core contribution is a measure the researchers call “observed exposure.” Prior studies of AI and labor markets, and there have been many, typically worked from theoretical capability. They asked which occupations contained tasks that AI could plausibly perform, then modeled disruption from there. The problem with that approach is that capability and adoption are different things. An AI system might be technically capable of performing a task today that no organization has yet deployed it to perform, for reasons ranging from regulatory constraints to procurement inertia to simple ignorance of the technology’s capabilities. Theoretical exposure measures the ceiling. Observed exposure, built from Anthropic’s own usage data across millions of real professional interactions with Claude, measures where the floor actually sits.

The gap between those two measures turns out to be enormous. And for law specifically, the gap is the story.

80 and 15

Legal occupations, in the study’s data, carry approximately 80% theoretical exposure. Meaning that AI systems could, in principle, speed up or perform something close to four-fifths of the tasks that lawyers and legal professionals spend their working days doing. That figure is consistent with the academic literature and with the vendor marketing that has surrounded legal AI for several years.

However, the observed exposure figure for legal occupations is 15%. That’s a 65-percentage-point gap between what AI could theoretically do in law and what it is actually observed doing in professional settings.

For comparison, the Computer and Math category shows 94% theoretical capability and 33% observed usage, a gap of 61 points. Office and Administrative roles show 90% theoretical, 25% observed. Business and Financial operations show 85% theoretical, 20% observed. Legal’s absolute gap is among the widest of any knowledge-work category in the study, and its observed penetration is lower than every white-collar category except healthcare, which carries its own obvious friction layers around clinical judgment, liability, and regulated workflow. In simple words, legal work is highly exposed to AI capability, but unusually resistant to AI adoption.

The natural question is why. The study is careful, appropriately so, not to over-interpret. The researchers note that the gap between theoretical and observed capability reflects multiple forces simultaneously: model limitations on certain task types, legal constraints, requirements for human verification, software integration hurdles, and the simple pace of diffusion across organizations. But that explanation, accurate as far as it goes, does not fully account for why law sits where it does. For that, you have to think about what law actually is as an economic and institutional system.

Why the Gap Is Wider in Law

Legal work, particularly at the BigLaw and sophisticated mid-market level, carries a set of characteristics that slow AI adoption in ways that are not present in other high-exposure categories.

The first is privilege and confidentiality. Lawyers do not simply have ethical obligations around client data; they have professional liability consequences, regulatory oversight from state bars, and practical vulnerability to court sanctions for disclosure failures. The threshold question before deploying any AI tool in a real legal workflow is not whether the tool can perform the task, but whether using it can be done in a manner consistent with duties of confidentiality. Consumer-facing LLMs like the Claude.ai interface that Anthropic’s Economic Index draws on are not what sophisticated legal work runs through. That usage gap matters, the study’s observed exposure data reflects how lawyers interact with publicly available tools, not how legal AI infrastructure is being deployed at firms that have invested in private deployment of models behind their own data walls. The 15% figure likely understates enterprise AI activity in law while accurately capturing retail adoption.

The second is the economics of hourly billing. This is ground covered in previous essays, but it bears restating in this specific context. The billable hour creates a structural disincentive to task automation that has no parallel in most other high-exposure knowledge professions. A financial analyst who uses AI to compress three days of modeling work into four hours has made himself more productive. A lawyer who uses AI to compress thirty hours of document review into ninety minutes has, under conventional billing arrangements, significantly reduced revenue. The individual incentive structures do not point toward adoption. They point toward using AI to increase throughput on billable matters rather than to replace billable tasks with automated ones, which is exactly what the broader data on BigLaw AI usage shows. Firms are capturing AI productivity gains as margin, not passing them to clients as savings.

The third, and perhaps most underappreciated, is the nature of legal task completion. Anthropic’s researchers weight observed exposure toward fully automated usage rather than augmentative usage, on the theory that full automation is more predictive of eventual labor displacement than collaboration. That is a defensible methodological choice. But it means that legal AI tools currently in widespread use, most of which are deeply augmentative rather than autonomous, contribute less to the observed exposure score than their actual penetration in the profession would suggest. A junior associate using Harvey to accelerate first-draft research, or a transactional team using contract analysis tools to flag non-standard representations and warranties, shows up in Anthropic’s data as augmentation rather than automation. The economic consequence is real, but the observed exposure metric, as designed, doesn’t fully capture it.

None of these three factors is a permanent shield. Privilege concerns are being addressed, imperfectly and unevenly, through enterprise deployment models and data governance frameworks. The billable hour has its own long-developing crisis, which the economics data from the prior essay laid out in some detail. And the augmentation-to-automation ratio that Anthropic has been tracking across its index reports has been shifting gradually toward automation over time, with automated usage growing from 43% to 45% of professional interactions between early 2025 and late 2025. The gap is wide. It is also closing.

What the 15% Actually Covers

Knowing that observed legal AI exposure sits at 15% is useful. Knowing which 15% is more useful. The study does not decompose task-level data for legal occupations in public-facing detail, but the Economic Index reports that preceded it, and the broader usage data Anthropic has published, give enough signal to reason about this.

Document review, legal research, and first-draft generation are the tasks that legal AI tools are most visibly deployed for. These correspond to work historically performed by junior associates: cite-checking, pulling precedents, summarizing regulatory filings, drafting initial versions of routine agreements. The Harvard Law School study referenced in last week’s essay found that an AI complaint-response system at an AmLaw 100 firm reduced associate time on a standard task from sixteen hours to three to four minutes. That is not augmentation. That is replacement of a discrete, repeatable block of professional labor, and it fits the pattern of fully automated usage that Anthropic’s methodology weights most heavily. If any slice of legal work shows up in the observed exposure data as fully automated, it is this category.

Higher in the task hierarchy, the picture changes. Contract negotiation, deal strategy, regulatory advice, litigation judgment, client relationship management, and courtroom representation are listed explicitly in the study as tasks remaining beyond AI’s current reach. The researchers cite representing clients in court as a specific example of a zero-exposure task. That is accurate in the narrow sense that no firm is routing courtroom argument through an AI system. It is potentially misleading in the broader sense that the preparation, strategy, and analysis that precede courtroom moments are precisely the tasks where AI augmentation is moving fastest, even if the moment itself remains human.

The practical implication is that the legal AI deployment currently registered in Anthropic’s data is concentrated at the entry level of the work hierarchy. The 15% being automated today is disproportionately the 15% that junior lawyers have traditionally learned on. That has consequences.

The Hiring Signal and What It Actually Means

The study’s most carefully worded finding is the one about entry-level hiring. The researchers find no systematic increase in unemployment among workers in the most AI-exposed occupations since late 2022. That is the headline most commentators ran with. But they also find suggestive, if not statistically conclusive, evidence that hiring of workers aged 22 to 25 has slowed in exposed occupations, with a roughly 14% decline in monthly job-finding rates.

For law specifically, the hiring data from independent sources is less ambiguous than the cross-sector unemployment data. The National Association for Law Placement (NALP) reported in late 2025 that BigLaw’s summer associate offer volume for 2024 programs was down 19% from 2022 levels. The callback-to-offer rate fell to 46%, the lowest since 2012. Firms reduced 2024 and 2025 summer associate headcount, and NALP analysts noted explicitly that two converging trends, falling firm appetite for large entering classes and rising law school enrollment, are likely to produce measurable declines in employment outcomes for the classes of 2027 and 2028. The median number of summer associate offers per office dropped to six in recent cycles, the lowest figure since 1993.

It would be an overreach to attribute that contraction primarily to AI. The pandemic-era over hiring, partner utilization rates, M&A deal volume fluctuations, and macro uncertainty all contribute. But the Anthropic study provides a plausible explanation for part of what is happening: if the tasks that junior associates historically performed are the same tasks now showing up in the observed exposure data as automated, firms have a weaker economic case for the size of entering classes they maintained in prior decades. They are not firing first-years. They are simply hiring fewer of them, and the labor market registers that as a slowdown in job-finding rates rather than as unemployment.

This distinction matters for how the profession interprets what is happening. Unemployment data showing no crisis is reassuring. What it does not show is whether the door to entry-level legal careers is quietly narrowing, which is a different question with a different answer. The preliminary evidence suggests it is.

The Gap Will Close, and the Pace Is Not Determined by Capability

The most important analytical point in the Anthropic study, stated directly by the researchers and largely lost in the headlines, is this: the pace at which the gap between theoretical and observed exposure closes is not determined by AI capability. It is determined by adoption. And adoption in law is determined by a set of institutional, economic, and regulatory forces that are entirely separate from the question of what the technology can do.

That framing should change how law firms think about the timeline of change. The conversation in most firms is still structured around AI capability. What can these tools do reliably, where do they hallucinate, what tasks are they accurate enough for, what is the liability exposure if the output is wrong. Those are legitimate questions. But they are not the binding constraint on the pace of change. The real constraint is whether the economic and organizational conditions for adoption exist. And those conditions are evolving faster in law’s external environment than inside law firms themselves.

In-house legal departments, which the Anthropic data captures differently from outside counsel, have been doubling their generative AI usage year-over-year. Legal operations teams are building internal AI workflows precisely because they have neither the billing hour incentive problems nor the professional liability exposure on client data that outside counsel face. When in-house teams can handle mid-complexity work with AI-enabled internal capacity, the work does not flow to outside counsel in the first place. The reduction in demand that results from in-house capability expansion shows up as a quiet contraction in outside counsel work, not as a dramatic renegotiation of billing rates, which is another reason why the existing disruption is hard to see clearly in standard economic indicators.

What Law Should Take From This

The conventional response to AI disruption narratives in law has followed a predictable sequence. First, skepticism about whether the technology is actually capable of the tasks claimed for it. Then, acknowledgment that some tasks are affected but insistence that the core of legal work is relationship-dependent, judgment-intensive, and institutionally insulated. Then, a strategy of using AI to make lawyers more productive without reducing headcount or prices. Then, eventually, a reckoning when external pressure makes the prior posture economically unsustainable.

The Anthropic study is useful because it allows a more granular version of this conversation. The 15% observed exposure figure is not a number that should produce panic. It is, however, a number that has already produced consequences, most visibly in entry-level hiring, that are showing up in the data before they have registered clearly in firm strategy. The gap between 15% and 80% represents both the work the profession still has to do and the time it has to do it.

That time is not unlimited. As capabilities improve and adoption spreads, observed exposure will inevitably move closer to theoretical exposure. The real question is whether the institutional conditions inside the legal industry will adapt in time. Billing models, data governance rules, and the economic incentives facing individual lawyers all shape how quickly AI can be deployed in practice. If in-house legal teams and alternative providers move faster than law firms, some of the work will simply stop flowing to outside counsel in the first place. When that happens, the adjustment comes from shrinking demand rather than deliberate organizational change.

The study will have updates. Anthropic’s researchers are clear that this is a first measurement, not a final verdict, and that they intend to track how observed exposure evolves as usage data changes. That commitment to ongoing measurement is useful for an industry trying to manage a transformation whose pace is still uncertain. For now, the most honest reading of the data is that legal AI is neither as widely deployed as its advocates claim nor as distant as its skeptics suggest. For now, the gap is the story.

Let me know what you think:

Leave a comment

A note on what this is: I have written about some of what follows in separate pieces over the past few months. The billable hour, alternative fee arrangements, the future of junior associates, and the impact of AI on law firm economics. I have made predictions, some of which I still stand behind, others I am less sure about. The purpose of this essay is not to argue a thesis. It is to lay out the numbers as they stand in early 2026, work through what they might mean, and see where they take us.

If you have been reading The Shift, you will recognize many of the topics. If this is your first encounter, the data should speak for itself. Some of it may confirm what you already suspect. Some of it may complicate those assumptions. That’s the point. This is a data essay. You are welcome to draw your own conclusions alongside mine.

The Revenue Picture

The AmLaw 100 generated $158.3 billion in gross revenue in 2024, a 13.3% increase over the prior year. Profits per equity partner hit $3.15 million, up 12.3%. Revenue per lawyer has climbed 54% since 2019. By any conventional financial measure, BigLaw has never been healthier. The Thomson Reuters/Georgetown Law 2026 State of the Legal Market report noted that 2025 continued this trajectory, with demand growth exceeding 5% and billing rates rising 7.3%, the fastest pace in two decades.

These figures deserve scrutiny rather than celebration, because they describe the aggregate performance of an industry whose internal economics are diverging rapidly. The CounselLink 2025 Trends Report, which draws on $67 billion in actual invoiced legal spend, found that firms with 750 or more lawyers now capture 49.3% of all corporate legal spending, up from 45.5% in 2021. Market share is concentrating at the top. The firms driving these headline numbers are increasingly a distinct economic category from the rest of the AmLaw 200, let alone the broader profession.

What strikes me about the revenue data is less the growth itself than where the growth is coming from. Demand increased, yes, but the larger driver was rate. Billing rate increases contributed more to revenue growth than volume gains in both 2024 and 2025. Firms are making more money not primarily because they are doing more work, but because they are charging more per unit of work performed. That dynamic has implications we will return to throughout this essay.

The Rate Divergence

The Wall Street Journal reported in February 2026 that senior partner hourly rates at the nation’s largest firms now reach $3,400 per hour, based on data from Persuit, the legal procurement platform. Across the 50 largest US firms, partner billing rates rose 16% in a single year. At Quinn Emanuel, top partners bill $3,000 an hour, a 34% increase since 2022. Susman Godfrey partners have set 2026 rack rates at $4,000. CounselLink data confirmed that the median partner rate at firms with 750 or more lawyers exceeded $1,000 per hour for the first time in 2024. In M&A specifically, partner rates at AmLaw 25 firms average $1,680 per hour, the highest of any practice area.

First-year billing rates are not exactly stagnating. Paul Weiss bills first-years at $895 an hour, Sullivan & Cromwell at $850. In absolute terms, those figures are still rising. But the growth trajectory of junior associate rates tells a more interesting story. The Wolters Kluwer ELM Solutions Real Rate Report found that senior associate rates surged 8.1% in 2024 while overall associate rates grew 3.1%. Twenty-nine percent of all timekeepers maintained entirely flat rates year-over-year. The growth is concentrating at the top of the seniority ladder, and the gap between what a seventh-year associate commands and what a second-year generates is widening in ways that the aggregate numbers obscure.

First-year BigLaw salaries have held steady at $225,000 since the Milbank reset in 2024, after two decades of periodic step-function increases. Meanwhile, partner profits per equity partner at AmLaw 50 firms grew 141% between 2012 and 2024. I want to be careful not to draw too mechanical a causal link here, because associate salaries and partner profits respond to different market forces. But the overall picture is one of accelerating returns to seniority and flattening returns to junior labor. Whether that is a temporary market condition or a structural change is one of the central questions this data raises.

The Non-Equity Era

The traditional BigLaw pyramid assigned a large base of junior associates to generate surplus revenue that flowed upward to equity partners. The standard rule of thumb was to bill associates at roughly three times their cost: one-third to salary, one-third to overhead, one-third to partner profits. Higher leverage, meaning more associates per equity partner, meant more surplus per partner.

That pyramid is reshaping, and the data support this transformation. The associate-to-partner ratio has declined from roughly 2:1 in 2000 to approximately 1.3:1 by 2025. Associates represented 40.2% of law firm lawyers from 2020 to 2024, down from 44.5% during the 2005 to 2009 period. The base of the pyramid is shrinking.

But the most significant structural change is happening at the partnership level itself. In 2024, nonequity partners outnumbered equity partners for the first time in the AmLaw 100, at a 51/49 ratio. Nonequity partner headcount grew 10.1% while equity partner headcount grew 3.3%. A cascade of historically single-tier firms have created nonequity partnership tiers since late 2023: Cravath, Paul Weiss, Skadden, Sullivan & Cromwell, Debevoise. These were the firms that for decades defined themselves by the proposition that every partner was a real partner. Their decision to create two-tier structures represents a fundamental rethinking of how elite firms generate and distribute profits.

The nonequity partner has become the new leverage engine. These lawyers, typically nine to twelve years out of law school, bill at senior rates, carry substantial workloads, and generate significant surplus revenue without sharing in the equity pool. They are, in economic terms, what junior associates used to be: the stratum that produces more value than it captures. The difference is that nonequity partners are far more productive per capita, bill at higher rates, and require less supervision. From a pure profit-generation standpoint, one nonequity partner billing 2,000 hours at $1,200 an hour is worth considerably more than two or three junior associates billing the same total hours at $600 to $800 each, once you account for the overhead of training, supervision, and the lower realization rates that junior timekeepers typically carry.

This is worth sitting with. The firms that created these tiers are not doing it reluctantly or defensively. They are doing it because the economics work better. And if the economics work better with fewer junior associates and more nonequity partners, then the traditional argument that BigLaw needs large entering classes to feed the pyramid becomes considerably less compelling. Firms are generating more profit with a different labor mix. Whether that mix is sustainable is a separate question. Whether it is good for the profession is another one entirely.

The realization gap

Billing rates are not the same as collected rates, and the gap between the two is growing in ways that matter. Average realization across the AmLaw 100 fell to 80.93% in 2023, a five-year low, with 75% of top-40 firms reporting declines. Some firms experienced sharp drops. Crowell & Moring’s realization fell to 66.5%, Troutman Pepper to 69%. The strategy, as Morrison & Foerster’s COO framed it, is straightforward: raise rack rates aggressively, accept higher discounts, and net out ahead on collected revenue.

This matters for the billable hour debate because it reveals a pricing model that is partly performative. The stated rate is not the real price. The real price emerges through a negotiation between the firm’s billing department and the client’s legal operations team, mediated by relationship dynamics, matter urgency, and competitive alternatives. When we talk about partner rates hitting $3,400 an hour, we are talking about a number that, on average, gets collected at something closer to $2,750 after discounts and write-offs. Still enormous. But the gap between sticker and settlement price introduces a degree of opacity that makes it genuinely difficult to compare the economics of hourly billing to alternative arrangements.

For readers who evaluate outside counsel spend, realization data should be a primary analytical tool. A firm quoting $1,200 an hour at 75% realization and a firm quoting $950 an hour at 92% realization may deliver nearly identical effective rates. The posted number is the beginning of a negotiation, not the end of one.

Declining realization also creates an interesting paradox for the AI discussion. If firms are already giving back 20% of billed revenue through discounts, the margin available to absorb AI-driven efficiency gains without reducing client bills is smaller than the headline rates suggest. Firms cannot simultaneously raise rates, increase discounts, and absorb the cost of AI tooling without eventually hitting a profitability constraint. Something in that equation has to give, and the data does not yet tell us clearly which variable will adjust first.

The Client Side: Frustrated but Not Yet Empowered

The 2024 ACC Chief Legal Officers Survey reported that 42% of legal departments received mandates to cut costs while 58% simultaneously faced significant rate increases from outside counsel. Client satisfaction with BigLaw hit a 25-year low according to BTI Consulting. Forty-five percent of in-house lawyers rate BigLaw as poor or terrible value for money.

Private equity firms, which are theoretically among the most sophisticated and demanding purchasers of legal services, are applying meaningful pressure but lack the infrastructure to enforce it. Apperio’s survey of 100 PE firms found that 73% reported significant legal fee increases over the prior three years, and only 30% of US PE firms trust their external counsel to bill accurately. Seventy-nine percent described legal spend as lacking transparency. Yet PE firms spent an average of $10.5 million annually on external legal services, with 91% still analyzing legal spend through manual spreadsheet processes. The tools to enforce cost discipline barely exist on the buy side.

This gap between frustration and purchasing behavior is one of the most important dynamics in the current market. Thomson Reuters found that client pushback on rate increases remained minimal throughout 2024 and 2025. Firms with 750 or more lawyers increased their share of total legal spend during the same period clients reported historic dissatisfaction. In bet-the-company situations, general counsel still reach for the most expensive firms and pay whatever they charge. The disconnect between expressed dissatisfaction and actual purchasing behavior is one of the billable hour’s most powerful structural defenses.

I find this pattern difficult to reconcile with my recent prediction that the hourly model is near collapse. Clients are unhappy. Clients are paying more than ever. Both of those statements are true. The explanation probably lies in the nature of legal services as a trust good: clients cannot fully evaluate quality in advance, so they use price and brand as proxies, and they are reluctant to switch providers in high-stakes situations where the cost of being wrong vastly exceeds the cost of overpaying.

The question I keep returning to is whether AI will change this dynamic not by making clients more willing to push back, but by giving in-house teams enough capability to handle work that previously required outside counsel. Sixty-four percent of in-house counsel plan to bring more work in-house, and the ACC/Everlaw 2025 survey found that generative AI usage in corporate legal departments more than doubled in a single year. If AI enables in-house teams to absorb mid-complexity work that currently flows to BigLaw, the impact on firm revenue could be significant even if no one ever renegotiates a billing rate.

Alternative Fee Arrangements: The Perennial Almost-Revolution

The data on AFAs tells a story of widespread adoption and limited impact. Ninety-six percent of firms with 150 or more lawyers offer some form of alternative fee arrangement. Roughly a third of in-house legal work is now paid through non-hourly structures, up from 29% the prior year. Fixed fees are the most common variant (73% of AFA-offering firms), followed by retainers (67%) and contingency arrangements (62%).

But only 22.3% of total law firm revenue came from AFAs in 2024, according to Citi/Hildebrandt. The hourly model still generates nearly four out of every five dollars. Employment law, insurance defense, regulatory compliance, and patent prosecution have embraced AFAs to varying degrees. M&A and high-stakes litigation, where deal complexity varies wildly and stakes are highest, have largely resisted.

Do AFAs actually save clients money? The evidence is genuinely mixed, and I think this ambiguity is underappreciated. Ogletree Deakins delivered 44% savings on legal spend and a 74% reduction in pending litigation. Seyfarth Shaw’s Lean Six Sigma-based model has documented savings of 15% to 50% depending on matter type. But many firms price AFAs to maintain revenue equivalence with what they would have earned on an hourly basis, and some clients have reverted to hourly billing because they suspect fixed fees conceal higher margins.

Jordan Furlong, who has written as much about the billable hour’s demise as anyone in the legal industry, has identified a critical paradox here. Clients resist AFAs in part because the same trust deficit that makes them dislike hourly billing prevents them from believing that fixed fees are fairly priced. As he put it in a recent piece, he is no longer in the business of predicting the billable hour’s demise. He is trying to understand why it is so resilient. That intellectual honesty is worth taking seriously. The structural incentives keeping the billable hour alive are formidable. Partner compensation, associate performance evaluation, budgeting, reporting, and practice management systems all revolve around hours. Even firms that adopt external AFAs typically maintain internal time tracking because, as Deloitte Legal’s Jeremy Black has noted, it remains the simplest way to measure output. Changing the pricing model means rebuilding the management model, and that is a far larger undertaking than what goes out in the engagement letter.

Goodwin Procter’s decision to deploy Persuit’s fixed-fee benchmarking tool globally in late 2025 is notable precisely because it represents the kind of infrastructure investment that AFAs require at scale. You cannot offer credible fixed fees on M&A work without sophisticated matter-level profitability analysis, historical cost data, and risk-adjusted scoping tools. Most firms lack this infrastructure entirely, which is one reason AFAs remain concentrated in practice areas where scope is relatively predictable.

AI: The Variable Everyone Talks About, Yet No One Understands

Here is the data point that keeps me up at night. Well, not really. But stay with me. A study by Harvard Law School’s Center on the Legal Profession found that an AI complaint-response system at one AmLaw 100 firm reduced associate time from 16 hours to three to four minutes. Not 16 hours to eight. Not 16 to four. Sixteen hours to minutes. And this is just one, admittedly dramatic, example. Firms are finding more and more ways to deploy AI across core work.

Yet the billing impact remains surprisingly muted. A 2025 survey of nearly 5,000 firms found that 58% reported AI had not affected billing practices at all. Only 19% reported reduced billable hours. Among large firms specifically, only 20% acknowledged AI-driven hour reductions. Depending on your perspective, the explanation is either rational or troubling: firms are using AI to increase throughput, take on more work per lawyer, and shift the surplus to partner-level revenue rather than reducing client bills. Thirty-four percent of firms are charging more for AI-enhanced services, while only 6% charge less, according to Axiom’s 2025 survey.

The client response to this is growing sharper. The ACC/Everlaw survey found that 59% of in-house counsel have seen no savings from their firms’ AI adoption, and only 24% are satisfied with how firms use AI for cost-effectiveness. Sixty-one percent plan to push for changes in how services are priced. The ABA’s Formal Opinion 512, issued in July 2024, established that lawyers benefiting from AI efficiencies under hourly billing must bill only for actual time spent. That is an ethical constraint that, if meaningfully enforced, directly threatens the strategy of capturing AI productivity gains as firm profit.

I have argued before that AI will not kill the billable hour through client demand alone. Clients have been demanding change for thirty years. What AI does instead is break the historical pattern in which each new technology created roughly as much work as it eliminated. Online legal research gave lawyers access to more cases to review. E-discovery tools generated more data to process. Each efficiency tool expanded the scope of billable activity. Generative AI operates differently. When a complaint draft goes from sixteen hours to four minutes, there is no compensating expansion of scope. The work is simply gone. Previous technologies compressed human work. This one replaces discrete blocks of it.

So tying this back to the future of junior lawyers, Harvey AI reached an $8 billion valuation by December 2025, with daily active usage growing 81%. Thomson Reuters’ CoCounsel scored highest in independent benchmarks. LexisNexis launched Protegé targeting 15% to 20% automation of lawyer tasks by 2028. Across the profession, 77% of legal professionals using AI deploy it for document review and e-discovery, 74% for legal research, and 59% for drafting briefs and memos. These are all tasks historically performed by junior associates. Firms like Ropes & Gray now allow first-year associates to count up to 400 hours of AI training toward billable targets, hours that cannot be billed to clients. Latham & Watkins brought all 400 first-year associates through a mandatory two-day AI Academy. These programs signal that firms understand junior associate work is being displaced but are choosing to retrain rather than reduce headcount. For now.

ALSPs and the Emergence of a Two-Tier Market

The US alternative legal services provider market is projected to exceed $23 billion by 2028, growing at approximately 21% annually. The CounselLink data shows what some analysts are calling a dumbbell effect, goliath firms on one end, specialized boutiques and ALSPs on the other, with intense pricing pressure in the middle.

What seems to be emerging is a split in the market that may already be further along than many people realize. At the top end, elite strategic work, the kind of matter where the partner’s judgment, relationships, and reputational insurance justify $2,000 to $4,000 per hour, will continue to command premium hourly rates from a shrinking pool of superstar practitioners. Below that tier, commoditized and process-driven work is migrating toward fixed fees, ALSPs, in-house AI-enabled teams, and subscription models. The firms caught in the middle, those that rely on volume-based hourly billing for work that is neither truly bespoke nor efficiently commoditized, face the most pressure.

For those of us in M&A specifically, the resistance to standardization is real but not necessarily permanent. Deal complexity varies enormously, and the argument for hourly billing in complex transactions has always rested on the genuine unpredictability of scope. But phased fixed fees, broken-deal discounts, and collar structures are gaining traction. The question is whether these are marginal accommodations that preserve the hourly model’s dominance, or early indicators of a more fundamental pricing shift that the M&A market has simply been slower to adopt.

Four Previous Death Sentences

Intellectual honesty requires acknowledging that the billable hour has survived at least four waves of credible disruption predictions. In 1993, the ABA’s alternative billing task force predicted that advanced technology, then meaning word processors and personal workstations, would eviscerate hourly earnings. After the 2008 financial crisis, the so-called more-for-less era generated intense client pressure that yielded incremental AFA adoption but no structural shift. After 2010, the rise of legal process outsourcing and legal operations functions was supposed to fundamentally alter the purchasing dynamic. More recently, the initial wave of legal technology in the mid-2010s prompted another round of predictions.

Each time, the billable hour adapted. Previous technologies created new categories of billable work roughly in proportion to the work they eliminated. The question this time is whether generative AI breaks that pattern. Bruce MacEwen, one of the most respected analysts of law firm economics, stated in December 2025 that he cannot see how the billable hour revenue model survives the arrival of generative AI. That is a strong claim from someone who has watched the industry for decades and is not prone to hyperbole.

I find myself somewhere between MacEwen’s conviction and Furlong’s pragmatic skepticism. The math is different this time, and I think the people who dismiss that are not paying close enough attention to the productivity data. But the institutional inertia is also more powerful than most disruption narratives credit. The 2026 Thomson Reuters/Georgetown report carries an explicit warning that echoes previous cycles: the legal industry surged like this before, in 2007 before the financial crisis and in 2021 before the inflation crunch. Record profits can mask structural fragility. They can also simply be record profits. The data alone cannot resolve which interpretation is correct.

Where the Numbers Leave Us

Looking at these numbers together rather than in isolation, a few patterns start to emerge. I’m not sure they can be resolved yet. In fact, I don’t think they can. But they do raise some questions worth naming.

If the new leverage model works, meaning fewer junior associates, more nonequity partners, and AI handling the routine work that associates used to perform, what happens to the talent pipeline? Law firms have always justified the grueling economics of junior practice with the implicit promise of eventual partnership. If partnership increasingly means a nonequity tier with limited upside, and if the junior years involve less substantive training because AI handles the tasks that used to provide it, the value proposition for top law school graduates changes materially. We may be looking at a profession that remains enormously lucrative at the top while becoming significantly less attractive as a career bet for incoming lawyers.

If AI productivity gains continue to accrue primarily to firms rather than clients, how long can that hold? The ABA’s ethical guidance and growing client sophistication create pressure toward transparency. But the enforcement mechanisms are weak, and the information asymmetry between firms and clients regarding AI usage is substantial. The 59% of in-house counsel who report seeing no savings from their firms’ AI investments represent a dissatisfaction that has not yet translated into purchasing behavior change. When and whether it does is the single most consequential variable for law firm economics over the next three to five years.

If the 90/10 split between hourly billing and AFAs shifts toward 75/25 within five years, as some analysts project, what does that mean for firm profitability? The firms best positioned for that transition are those investing in pricing infrastructure. Dedicated pricing directors, matter-level profitability analysis, AI-enhanced scoping tools. Most firms lack this almost entirely. The transition cost is not minor, and firms that wait too long to build these capabilities may find themselves structurally disadvantaged against competitors and ALSPs that already operate on outcome-based models.

And finally, the question that sits beneath all the others: are we witnessing a genuine structural transformation in how legal services are priced and delivered, or are we watching the latest iteration of a cycle in which temporary disruption pressures are eventually absorbed by an extraordinarily resilient institutional model? The honest answer is that the data supports both readings. Revenue and profits have never been higher. Client dissatisfaction has never been greater. AI productivity gains are real but unevenly distributed. The billable hour is under more pressure than at any point in its history, and it still generates 78% of all law firm revenue.

Disappointing ending, I know. What I can say with confidence is that the economics of modern law firms are more internally contradictory than the numbers suggest. Record profits coexist with declining realization rates. Rising partner billing rates coexist with stagnating junior compensation. Widespread AI adoption coexists with minimal billing impact. These contradictions will eventually have to resolve one way or another, and the resolution will determine whether the next decade looks like a managed evolution or a transformation.

The transformation, when it arrives, if it arrives, will more likely be imposed from outside. By clients with better data, by technology that makes the time-value equation too absurd to sustain, or by a new generation of lawyers who refuse to organize their professional lives around a metric that no longer measures what they actually produce.

These are not conclusions so much as observations about where the numbers point. I would rather update my thinking as the evidence changes than defend a prior position past its usefulness. You will probably see me change my mind more than once in these essays. I would encourage you to do the same.

The numbers are on the table. Make of them what you will.

If you like this essay, the best way to support it is to share it with your colleagues.

Share

Les me know what you think

Leave a comment

Sellers often think leverage is something negotiated into existence. A tighter closing condition. A stronger remedy. A cleaner walkaway standard. In reality, some of the most durable leverage in a private M&A process forms much earlier, long before the SPA is marked up and long before lawyers start arguing about who bears which risks.

It forms when the buyer starts spending real money. Not symbolic money. Not preliminary banker fees. Real, visible, board-approved diligence spend: legal, accounting, tax, regulatory, consultants, industry experts, IT diligence, cyber reviews, environmental reports. The kind of spend that shows up in budgets, requires internal approvals, and attaches names to line items. At that point, the transaction stops being an option and starts becoming a commitment, not legally, but institutionally.

From the seller’s perspective, this is one of the most reliable signals of buyer seriousness that exists. More reliable, in many cases, than expressions of interest, headline valuations, or even early exclusivity. Because once enough money has been spent, the cost of not closing shifts inside the buyer organization in ways that contracts rarely capture.

In a typical mid-to-large private equity transaction, total diligence spend can easily run into the high six or seven figures before signing. Legal fees alone for complex carve-outs, regulated targets, or cross-border deals routinely exceed $500,000. Add accounting, tax structuring, quality of earnings, commercial diligence, cyber, ESG, and regulatory work, and the number grows quickly. None of that money is discretionary once the process is underway. It has sponsors. It has internal owners. It has an audience. By the time that level of spend is authorized, the buyer is no longer deciding whether to pursue the deal. The buyer is deciding how to justify it.

Inside most PE firms and corporate acquirers, diligence spend is a signal of conviction. Someone took the deal to investment committee. Someone argued that this opportunity justified deploying scarce resources. Someone approved hiring advisors and allocating internal bandwidth. When a deal collapses after that point, the question is rarely “why didn’t it close?” It is “why did we spend all this money and get nothing?”

Senior deal professionals cannot move on from this. Failed deals with heavy spend accumulate quietly in performance reviews, partner discussions, and promotion decisions. One abandoned deal is explainable. Several start to form a pattern.

That is why sellers often feel a subtle but real shift in tone midway through diligence. Questions become narrower. Timelines compress. Pushback softens. Issues that once felt existential begin to look manageable. The risk is still there, but internal tolerance for a pens down has declined.

Importantly, this dynamic does not rely on irrationality. It does not require the buyer to “fall in love” with the asset. It flows naturally from how organizations behave once resources are committed. Even disciplined investors face friction when reversing course after visible investment. The cost is not the money already spent; it is the explanation required for having spent it.

This is why broken deals cluster early. Empirically, transactions that fail tend to fail before diligence ramps up fully, not after it peaks. Once a buyer is deep into confirmatory diligence, abandonment rates drop materially. That does not mean all such deals close, but it does mean the bar for walking away rises.

It shows up in how hard the buyer pushes, how it frames issues internally, and how willing it is to absorb uncertainty. A buyer with minimal sunk cost negotiates defensively. A buyer with meaningful sunk cost negotiates pragmatically. A buyer with very high sunk cost negotiates to finish.

This does not mean sellers should ignore risk. Nor does it mean diligence spend guarantees closing. Deals still die late, sometimes spectacularly. Regulatory blocks, financing collapses, fraud discoveries, or fundamental business deterioration can overwhelm even the strongest internal incentives. But absent those events, heavy diligence spend materially reshapes the balance of power.

What sophisticated sellers do differently is recognize when that shift has occurred. They understand that once the buyer is deeply invested, leverage is no longer symmetric. At that stage, sellers can be firmer on issues that truly matter, while being pragmatic on those that do not. They can do this because they know the buyer is now doing the same.

This also explains why sellers sometimes tolerate extended diligence periods. Length alone is not the issue. Spend intensity is. A buyer that drags its feet without deploying resources is preserving optionality. A buyer that spends aggressively is doing the opposite.

None of this replaces good drafting. It sits alongside it. But it does reframe how leverage should be understood in modern dealmaking. Power does not suddenly flip at signing. It accumulates gradually, shaped by internal incentives long before lawyers finalize terms. Sellers who focus exclusively on legal mechanics often miss the quieter forces already working in their favor.

Due diligence is not just about information. It is about commitment. And commitment, once visible, is leverage.

For sellers, the lesson is not to exploit buyers recklessly or push deals past the point of sense. It is simply to recognize reality. By the time enough money has been spent, the buyer is no longer evaluating the deal in isolation. It is evaluating itself.

Leave a comment

Lately, most private M&A negotiations have felt stuck on price. Buyers and sellers have spent months debating value in a market shaped by higher rates and greater uncertainty. On the surface, this looks like a normal adjustment cycle, and many expected it to show up directly in lower multiples. In reality, price hasn’t moved as much as people think. Instead, valuation has adjusted in a less obvious way.

Across much of the private-company market, headline valuation has been surprisingly stable. Middle-market EBITDA multiples haven’t collapsed. According to GF Data, the average multiple in 2024 sat around 7.2× EBITDA, roughly in line with 2023. Data through late 2024 and into 2025 shows minor variations by size and quality, but no clean, market-wide reset. But what the averages hide is a change in where risk is being carried. The repricing is happening elsewhere.

Instead of forcing valuation gaps into the multiple, the market has pushed them into structure. What gets paid at signing, what is contingent, what rolls forward, and how post-closing adjustments actually play out.

Start with earnouts. They’re not new, but they’re being used differently. Once a tool for special situations, earnouts have become a standard way to deal with forecast uncertainty. Recent market studies show earnout usage well above historical norms, with shorter measurement periods and tighter metrics. On private equity exits in particular, a growing share of consideration is contingent, allowing buyers and sellers to bridge valuation gaps without touching headline enterprise value. The question is no longer about what the business is worth today, but about which assumptions the seller still has to prove before receiving full value. Risk moves into time and performance, not into the multiple. That also explains why practitioners have become more cautious. Earnouts don’t eliminate valuation disputes; they defer them.

What often gets missed is that earnouts are not just valuation tools; they are operating arrangements. Performance targets, budget control, integration decisions, and capital allocation during the earnout period all shape outcomes. Buyers typically control these levers, even when sellers remain involved in management. That creates a built-in tension. The seller’s payout depends on results, but the buyer controls many of the inputs. Over time, earnouts become less about accounting definitions and more about how much practical autonomy the seller retains while the clock is running.

Rollover equity has evolved in much the same way. In PE-backed transactions, it is no longer a signaling device or a soft alignment tool. It has become a core feature of deal economics. GF Data reports seller rollovers in roughly two-thirds of middle-market transactions in 2024, continuing a trend that began during the pandemic and persisted as leverage remained constrained. Typical rollover stakes now fall in the 10% to 40% range of equity value, frequently structured on a tax-deferred basis.

At a basic level, rollover equity allows buyers to preserve headline valuation while shifting part of the consideration into future performance. Sellers avoid conceding value at signing, but accept that full realization depends on a later exit. What looks like agreement on price is really an agreement to defer part of the valuation question. The debate is not resolved at close; it is carried forward into the ownership period.

Where rollover equity becomes more complex is in how it reallocates risk and control. A seller who rolls equity becomes a minority investor in a governance and capital structure designed primarily around the sponsor’s incentives. Information rights, board representation, consent thresholds, exit mechanics, and distribution waterfalls determine whether the rollover functions as genuine participation or as deferred consideration with limited influence. In platform strategies or leveraged structures, dilution, recapitalizations, and exit timing can materially reshape outcomes. In that sense, rollover equity is not just an alignment tool. It embeds valuation risk into ownership, governance, and time, leaving much of the real economics to be determined well after signing.

Seller notes and other deferred consideration serve a different purpose. As acquisition financing became more expensive, notes reappeared as a way to smooth cash at closing without cutting enterprise value. Typical structures involve 10% to 20% of EV, often unsecured, paying mid-single-digit interest and sitting behind acquisition debt. From a pricing perspective, notes spread value over time rather than compressing the multiple. From a risk perspective, they bring familiar complications: offset rights, structural subordination, and the reality that enforcement is rarely easy.

Seller notes shift valuation risk into credit exposure. Sellers are no longer just agreeing on price; they are underwriting the buyer’s capital structure and execution risk. Payment depends not only on business performance, but on leverage levels, refinancing conditions, and sponsor behavior over time. In practice, sellers holding notes are exposed to risks they neither control nor can easily hedge. What looks like deferred consideration is often junior risk capital, priced more like patience than debt.

Purchase price adjustments have undergone a similar evolution. What once operated as a mechanical true‑up, an accounting safeguard to ensure “normal” working capital at closing, now functions as a meaningful valuation lever. Recent PPA studies covering more than 1,250 private‑target acquisitions between 2020 and Q3 2024 show that adjustments routinely move millions of dollars post‑close, with outcomes split across no‑adjustment, surplus, and buyer claims, underscoring that PPAs now influence realized value rather than merely tidying up balance‑sheet timing differences. At the same time, both the 2024 and 2026 SRS Acquiom PPA studies highlight rising complexity in defining target working capital, broader use of customized escrows, and a continued increase in PPA-related disputes, signs that parties increasingly negotiate PPAs as a way of embedding risk allocation into the deal rather than treating them as a technical afterthought. In effect, PPAs have shifted from a neutral post‑closing adjustment to an integral component of valuation structure: a mechanism through which buyers and sellers test assumptions, reprice short‑term performance, and push part of the valuation debate into the post‑close period.

So why didn’t the market simply cut multiples?

Because, in many segments, it didn’t have to. Valuations didn’t collapse; they spread out. High‑quality assets often cleared at familiar levels, while riskier or smaller deals adjusted through structure rather than headline price. Quarter‑to‑quarter changes in average multiples frequently reflected changes in deal mix rather than a broad repricing. In that environment, structure became the outlet for uncertainty.

The market effectively decided where valuation disagreements should live. Instead of forcing agreement at signing, private M&A has shifted uncertainty into mechanisms that activate only if certain assumptions don’t hold. The multiple remains visible; the real economics unfold over time.

As a result, drafting matters more. Post-closing management matters more. Disputes become more likely, not because parties are more aggressive, but because the contract now spreads pricing across multiple decision points instead of one. The contract isn’t just a record of value; it’s the place where value is tested.

Valuation expertise no longer lives only in banker models or fairness opinions. It lives in earnout definitions, rollover governance, note subordination, and adjustment mechanics. Deals that appear “fully priced” at signing often aren’t. They’re provisionally priced, with real economics spread across structure, performance, and time.

Leave a comment