1. Six months later, what held up

SaaS multiples have indeed not come back. The median SaaS company trades around 3–4x revenue vs the 30x we saw at the 2021 peak. We are seeing these multiples even for SaaS companies that are growing, with good gross margins, and profitable. Remember, the market is pricing terminal value. If the market believes long term growth is going to be impacted due to agentic AI, no amount of improved cash flow or even near term growth will bring back the old multiples.

2. The market has bifurcated

The good news: the market is re-rating SaaS companies that have figured out successful playbooks in this AI native world. For example, companies like Datadog, Shopify, and Samsara, have managed to achieve above average ARR multiples.

Many of these companies leveraged elements of the survival guide I outlined previously. In particular, the companies that fared the best managed to especially see the product opportunities around agentic AI and to tap into new AI demand.

3. Three paths are emerging

Path 1: Re-found around an AI-native outcome. E.g., Intercom -> Fin.ai. With growth stalled, Intercom built an AI support agent six weeks after ChatGPT launched. It moved to consumption based pricing. The new product went from $0 to $100M+ ARR in a few years. Salesforce agreed to buy the company for ~$3.6B.

Path 2: Put the company in the path of AI demand. E.g., MongoDB. AI Agents like Claude code are now making database choices. MongoDB has managed to become the default choice in many cases. The market recognizes they were able to capture AI demand, and they have benefited from multiple expansion.

Path 3: Turn the system of record into the system of action - Samsara - Samsara is a system of record for physical operations. Its new Agent Studio lets customers turn that context into action by deploying agents that automate operational workflows. Instead of being displaced by AI, Samsara is using its data and installed base to become the platform through which the work gets done, achieving above average multiples.

4. Updates to the SaaSpocalypse CEO survival guide

Much of the advice I provided back in February still stands today, but would I would emphasize is:

1) If I didn’t convince you before, accept that old multiples aren’t coming back

2) Inventory advantages, identify one of the success paths:

• Option 1 - Re-found around an AI-native outcome.

• Option 2 - Put the company in the path of AI demand.

• Option 3 - Turn the system of record into the system of action

3) Build an AI native development team, as a rule of thumb it’s probably 30% of your current team size

4) Build an AI Native GTM team, higher quotas, new pricing models, FDEs, etc

5) If you can’t stomach 1-4, sell the company

5. The status quo remains the worst option

The good news: There are options available to each SaaS CEO. Re-found the company around an AI-native outcome. Put the company directly in the path of AI demand. Or use the durability of the existing business to create liquidity and fund the next chapter. Each option does require a new playbook.

The hard news: It’s not easy.

Only option that really doesn’t work? It’s the status quo.

I’ve seen services from both sides — as a SaaS CEO who used professional services as a strategic differentiator at Gigya, and now as an investor watching AI collapse the old line between software and services. Here’s where I’ve landed.

Why services, why now

The first-order case is clear: AI can transform a historically difficult category.

Traditional services businesses suffer from low margins, inconsistent quality, limited scalability, and weak NPS. Labor is the product, and labor is hard to hire, train, and keep consistent. AI-enabled services can flip that model.

But the bigger unlock may actually be enterprise adoption.

Selling AI agents directly into an organization often requires multi-stakeholder buy-in, change management, and anxiety around job displacement. That friction is slowing growth for many agent startups.

Services markets are different. They are already outsourced.

The enterprise has already decided it doesn’t want to do this work in-house. It already buys the outcome, not the tool. That means faster sales cycles, quicker time to value, and far less internal change management. The AI can operate inside the vendor’s four walls, with FDE style customization handled in the background rather than pushed onto the customer.

For founders, the conclusion is a little counterintuitive: the best way to sell agents into the enterprise may be to not sell agents at all.

The 2x2 below is how I think about it. Internal AI agents deliver value fast but drag change management along with them. Traditional enterprise software is slow on both counts. Traditional outsourced services are easy to buy but slow to improve. AI-enabled services sit in the quadrant enterprises actually want: minimal change management, fast time to value.

Margins are not moats

Now the catch.

Many AI-enabled services startups will have better gross margins. Fewer will have real moats with compounding advantages: better product. Better data. A tractable market. Higher switching costs. Recurring revenue.

Margin improvement is what AI does out of the box. Point a model at a labor-heavy workflow and the cost line drops. But if that’s all that happens, your advantage lasts exactly as long as it takes the next founder (or the incumbent you’re disrupting) to apply the same models to the same workflow.

That’s the test in the 2x2 below. High AI impact with no compounding durability? Cost cutter. Durable business where AI doesn’t materially improve the service? Sticky, but not AI-native. The venture-scale outcomes live in the top right — where AI improves both the service and the durability of the business.

Founders should look for categories where AI improves both the service and the durability of the business.

The scorecard

To evaluate AI-enabled services companies — and entire categories — I use an 8-criteria scorecard, each scored 1 to 5:

1. Total addressable market. Is the current services market large enough to support a venture outcome?

2. Fragmentation. Is the market fragmented enough for a new entrant to wedge in?

3. Revenue quality. Is revenue recurring, or mostly one-time and project-based?

4. AI product impact. Does AI make the service materially better, not just cheaper or faster?

5. Data flywheel. Does every customer interaction improve the service?

6. Switching costs. Once utilized, is the service hard to rip out?

7. AI economics. Does AI structurally change the gross margin profile — a limited bump, or a massive one?

8. Competitiveness. Is the category already a red ocean of funded entrants, or still open water?

   

The first seven criteria tell you whether the business can compound. The last tells you whether the opportunity is still worth entering.

I ran every major services category through it

Using Claude’s Fable model, I ran the full services category list through the scorecard and ranked the results. Two things stood out.

1. The top of the list is a red ocean.

Medical billing / RCM services tops the ranking at 4.9, followed by AI-native accounting & tax firms, finance back-office services, TPA / claims administration, and MDR / security operations.

But notice the overlap: accounting & tax, MDR, and MSPs all rank top 10 on score and top 5 on competition. The rest of the most-crowded list won’t surprise anyone reading fundraise announcements lately: AI-native law firms and ALSPs, SDR-as-a-service, recruiting-as-a-service, AI implementation services, back-office BPO. This is where founder and investor attention is concentrated today — rightfully so — and these crowded categories are likely great Series A/B+ opportunities, where execution and distribution will decide the winners.

2. Blue oceans still exist.

Good news for seed-stage founders: attractive, less-crowded opportunities are still out there.

Tier-1 support BPO, procurement as a service, and marketing operations all score high with few venture-backed incumbents. And categories that touch the physical world look underserved — customer operations beyond support, physical-world audit & verification, and logistics operators, for instance.

At the top of the blue-ocean ranking: customer operations beyond support (4.4), tier-1 support BPO (4.3), and physical-world audit & verification (4.0).

The bottom line

The most interesting AI-enabled services companies are not just replacing labor with AI. They’re turning service delivery into product, data, workflow depth, and customer lock-in.

Better gross margins are the starting point. Compounding advantages are the company.

One more thing: this scorecard evaluates categories. Whether a specific startup in a category survives contact with Claude, ChatGPT, and Gemini is a different question — that one gets the Defensibility Report Card treatment from my last post. Use them together.

If you’re building in one of these categories, I’d love to learn more. And if you’re operating in one of these markets and think the scorecard got something wrong — run your own category through it and report back.

Claude Code means code is becoming commoditized. Claude Cowork means agentic work is becoming a platform feature.

The question I've been thinking about is if both of those trends accelerate, what is defensible for startups?

For fun, I built an app for founders asking this question. Check out the Startup Defensibility Report Card

Based on a framework for evaluating AI startup defensibility - it scores any startup across 10 dimensions, including multiplayer usability, human accountability, proprietary workflows, data flywheels, trust, outcome ownership, etc

Below, I’ve included the report cards for a few leading AI startups

Harvey:

Cursor:

Lovable

Here is the 10-point moat checklist.

1. The Removal Test

Would the product still matter if you removed the LLM? If nothing meaningful remains, it is probably a prompt, wrapper, or Skill. If the workflow, orchestration, permissions, data, and outcomes still matter, the model is labor inside the product, not the product itself.

2. Multiplayer Work

Does the product coordinate multiple people? Single-player workflows are easy for better models to absorb. Multiplayer workflows need roles, handoffs, approvals, queues, access controls, escalation paths, and shared state. This is the most important moat.

3. Stateful Workflow

Does the system remember what happened over time? Defensible products do not just answer one request. They track decisions, exceptions, approvals, outcomes, and unresolved work. The more state the product owns, the harder it is to collapse into a generic agent.

4. Human-in-the-Loop Requirement

Is a human required because of risk, liability, or regulation? In legal, financial, healthcare, security, and compliance-heavy workflows, accountability cannot be fully delegated to an autonomous agent. The moat is not just automation. It is controlled automation with accountable humans in the chain.

5. Proprietary Workflow Logic

Does the product contain domain-specific rules that compound? Policies, exception taxonomies, deterministic checks, review standards, regression tests, and domain-specific operating procedures are harder to copy than prompts. As code gets cheaper, correctness under real-world variation becomes more valuable.

6. Context Graph

Does the product build a history of action, not just store data? Generic agents can read customer data. They cannot easily recreate the history of what decisions were made, why exceptions were handled a certain way, what worked, and what failed. That operational memory becomes a moat.

7. Regulatory or Compliance Moat

Does the product operate in an environment where certification, auditability, or legal accountability matters? Agents cannot be licensed, audited, insured, or held responsible in the way companies and humans can. If compliance requires a trusted chain of accountability, horizontal platforms cannot simply replace the product.

8. Proprietary Data Flywheel

Does usage generate valuable live operational data? Static data gets commoditized. Live data from repeated workflows, resolutions, edge cases, outcomes, and performance benchmarks can compound. The best data moat comes from the work being done through the system.

9. Distribution and Trust

Is the product sold into a specific buyer with domain credibility? Self-serve tools for everyone are fragile. A considered purchase in a specific vertical, sold with expertise, trust, and clear outcomes, is harder to displace. Distribution is stronger when the buyer believes the company understands the domain better than a generic agent platform.

10. Outcome Ownership

Does the company sell results, not software access? If the product promises throughput, accuracy, reduced risk, faster resolution, or business outcomes, it is not competing only on UI or model quality. It is taking responsibility for the result. That is much harder for a generic Skill to replace.

Try your own startup (or idea) and report back!

But to build a great company—an enduring company—my sense is it still comes down to leadership.

While metrics and product strategy evolve, the fundamentals of how people work together remain constant. Long before I was scaling Gigya or investing at Mayfield, my leadership style was shaped by a different set of mentors. I was fortunate to be surrounded by coaches who believed deeply in leadership and character: Coach Kunau, Coach Vies, Coach Sherman, Coach Cassidy, and Coach Wilson.

I listened to seminars by Zig Ziglar. I wrote book reports on sports leadership titles by John Wooden, Lou Holtz, and Coach K. At the time, it felt like sports leadership training. In retrospect, it was startup leadership training.

The Founder’s Reading List

In a similar vein, as a startup CEO, if you are looking to sharpen your leadership toolkit, this is the shortlist I recommend:

On Leadership

• Unusually Excellent: The Necessary Nine Skills to Practice Great Leadership by John Hamm. This is a practical guide on the skills that actually move the needle. I recommend reading this every year as a CEO or Founder.

• The Hard Thing About Hard Things by Ben Horowitz. The honest manual for founders. It serves as a necessary reminder that this journey isn’t supposed to be easy.

On Management

• Good to Great by Jim Collins. The seminal work on what separates flash-in-the-pan success from enduring greatness.

On Life & Perspective

• The Last Lecture by Randy Pausch. A vital reminder of time and priorities.

• How Will You Measure Your Life? by Clayton Christensen. A framework for ensuring your personal definitions of success align with your professional ones.

The list could be longer, but these five are enough to build the foundation. I suggest starting with the first two. The last two are a nod to the fact that CEOs have to play the long game and holistically prioritize all parts of their life.

My Learned Leadership Principles

While every CEO must develop their own leadership philosophy, I’ve spent time distilling 12 foundational values that stuck with me:

1. Show up early. Every time. Respecting time is the first step in respecting your team.

2. Look people in the eye—and listen actively.

3. Don’t make excuses. Own the miss.

4. Be accountable to your teammates.

5. Pick people up when they’re down. Culture is defined by how you treat people during the low points, not the high ones.

6. Put the team’s success over individual success.

7. Stay humble. Praise others when you succeed; look in the mirror when you fail.

8. Lead by example. You cannot demand what you are not willing to give.

9. Set clear goals and objectives. Ambiguity is the enemy of execution.

10. Success is built through preparation and process.

11. If you don’t work harder than the competition, you don’t deserve to win.

12. Have a positive mental attitude. (Admittedly, this is the hardest one for me, but it is non-negotiable for the team).

I can imagine that in this AI native era, people will need leadership now more than ever.

My feelings about the exit have whipsawed accordingly. First I felt OK. Then I felt like an idiot. Now I mostly feel empathy for the SaaS CEOs still in the arena.

It’s a strange moment. Much has been written about the death of SaaS. First, software was going to eat the world. Now AI is eating software—and maybe the world too.

Is SaaS dead? No. Software doesn’t disappear. But the premium valuations may be permanently gone. Growth has slowed, markets are more competitive, and the investor appetite that once rewarded ARR growth at any cost has shifted to a new category entirely: agentic AI.

I have no vested interest in the answer to the question “what should a SaaS CEO do right now?” other than camaraderie. I belonged to that lonely club of CEOs for a long time, and I know many who are still in it. So I put together a survival guide—ten things I’d do if I were running a SaaS company today.

1. Accept That the Old Multiples Aren’t Coming Back

Stop valuing your company off 10x revenue. The market has re-rated SaaS, and hoping for a return to peak multiples is not a strategy. Start planning around 2–3x revenue. That number changes every decision downstream—fundraising, M&A, burn, etc.

2. Recognize You Have Nothing to Lose

Once you internalize that your company’s value may sit below your liquidation preference stack (which it might), something shifts. You’re no longer protecting the status quo. You’re free to take extreme action—and you should.

3. Get Curious, Not Angry

There’s an agentic-native startup in your space that launched three years ago with one-tenth of your revenue. It’s valued much higher than you are right now. That stings. They probably have inbound funding interest, and you probably can’t raise. But instead of getting mad, get curious about why investors are willing to pay that premium. The answer is the opportunity.

4. See the 100x TAM Opportunity

The real unlock is that agentic AI doesn’t just improve existing workflows—it automates entire outcomes. These are greenfield markets, far bigger than the ones your SaaS product originally targeted. This is where the attention, the talent, and the capital are headed. The question isn’t whether to pursue it. It’s how.

5. Inventory Your Actual Advantages

You’re not starting from zero. You have distribution—getting contracts into enterprises is hard, and you already have them. You likely have data—your SaaS platform required accessing and organizing customer data that an agentic product can build on. And you have trust—your customers already work with you and don’t want to onboard a dozen new vendors. You have a brand. Build off these advantages deliberately.

6. Transform How You Build Software

It’s time to build an agentic product, and that probably isn’t happening with your current team using their current tools. Your competitor with five engineers each running ten agents on Claude Code is shipping faster than your team of fifty. Meanwhile, your team is building the one-thousandth feature to protect a renewal, lobbying to fix tech debt, and treating GitHub Copilot as the AI strategy.

A small change here won’t get you there. You need to transform how you code for 10x the output. Stop building new features for the legacy SaaS platform. You may need to acquire new AI-native talent. This is a step-change, not an iteration.

7. Learn to Build Agents—It’s a Different Discipline

Building an agent isn’t like building the SaaS product you’ve shipped for years. You need to solve for agentic orchestration, accuracy and consistency, personality, memory and context, human-in-the-loop control, tool integrations, multi-agent collaboration, and usage or outcome-based billing. How it gets sold and packaged are completely different from seat-based SaaS. All of this is still being figured out in real time—every new model release reshuffles the deck. This is not a part-time job.

8. Sell to New Customers at AI-Native Speed

You need to start landing new customers with the agentic product, and the old 0→1, 1→3, 3→9 growth playbook doesn’t cut it anymore. You need 5-10x+ year-over-year growth just to keep pace with AI-native competitors who are setting the benchmark for what investors want to fund.

9. Migrate Existing Customers—Even If It Hurts

Transitioning your installed base is tough. Your previous buyer may not be your new buyer. The packaging and pricing will be different. Your gross margins will shift. But this is where your advantages—data, contracts, trust—pay off.

Be willing to give the agentic product away to existing customers. Any new AI customer, whether net-new or an upgrade, is worth 30x what a legacy SaaS renewal is worth. Don’t optimize for the upsell—optimize for the transition. Make sure your team is incentivized accordingly, even if that means paying more commission to drive adoption.

10. Accept That You’re Starting Over—and Go

No one is going to hand you credit for what you built before. You’re essentially re-founding the company. Suck it up, and go.

The Bottom Line

If this sounds like a lot, it’s because it is. As someone who had to pivot multiple times as a CEO, I can tell you: Is it possible? Absolutely. Is it worth it? Absolutely. Is it easy? Well…there is a reason for the metaphor “burn the boats”

.

I work with founders who’ve cracked PMF and are ready to scale. The biggest trap they face isn’t strategy or market timing—it’s that they don’t realize their role as CEO has to change.

In the early days ($0–$1M ARR), you’re the ultimate individual contributor. You set up the infrastructure. You build the product. You create the pipeline. You schedule the meetings. You create the sales deck. You give the pitch. You close the deals. You complete implementation. You handle support.

Your personal output drives the company’s survival.

But as you scale, that same instinct becomes a liability. What made you fast early can make the company slow later—because everything routes through you.

The shift: from “doing” to “designing”

Scaling changes the nature of the CEO job. The role shifts from doing the work to designing the system that produces the work.

Instead of constantly asking, “What can I do myself?” your job becomes, “Who—or what AI agent—can do this reliably without me?”

Early on, your responsibility is to make sure things are done right. As you scale, your responsibility is to make sure you’re doing the right things—and that the organization can execute them without you being in every loop.

That’s the difference between being the hero and building the machine.

The CEO scorecard for scaling

Here’s a scorecard for scaling-stage CEOs:

1. Build the team

Hire great people to do the jobs you used to do.

This is the hard part emotionally. It requires admitting that the company no longer wins because you’re the best at everything. The company wins because you’ve built a team that can out-execute what any one person—no matter how talented—could do alone.

2. Prioritize relentlessly

Ensure the team is doing the right things.

When you’re small, effort is the constraint. When you’re scaling, focus is the constraint. You’ll have smart people who can ship a lot of work—so the CEO’s job is to keep the organization pointed at the few things that actually move the business.

3. Over-communicate

Repeat the vision and priorities until you’re blue in the face.

You may feel like you’re saying the same thing over and over. Good. In a scaling company, alignment doesn’t happen because you said it once. It happens because you’ve repeated it enough that it becomes the team’s default operating system.

4. Focus your personal time

Own 1–3 critical initiatives per quarter. Delegate the rest.

This is still important. The CEO should remain deeply engaged in the highest-leverage work. But the definition of “leverage” changes. Your goal isn’t to be involved in everything—it’s to be involved in the few things that unlock the next phase of growth.

What you can do differently next week

If you’ve hit PMF and you’re trying to scale, here are a few concrete moves to make immediately:

• Make a list of everything that currently depends on you. Pick one and redesign it so it no longer does.

• For recurring work you do every week, decide: hire it, delegate it, or automate it (including with AI agents where it makes sense).

• Write down the top 3 priorities for the quarter in plain language. Share them, then repeat them in every forum until there’s no ambiguity.

• Pick the 1–3 initiatives you will personally own—and explicitly delegate everything else.

The unlock

This is a big unlock for founders who get it.

Don’t be the bottleneck. Build the machine that builds the business

.

Not just about next year’s budget, but about alignment:

• Is the company clear on where we’re going?

• Do we agree on what matters most?

• Does the plan translate into concrete execution, quarter by quarter?

Here’s the simple planning flow I used at Gigya the year we achieved $100M ARR run rate.

Start with mission, objective, and strategy

Annual planning begins with revisiting the mission, objective, and strategy. This is the CEO’s job.

Before you talk numbers, you should be able to answer:

• Mission: Why do we exist?

• Objective: What are we trying to achieve over the next 12–24 months?

• Strategy: At a high level, how are we going to win?

Take the time to restate and pressure-test with your leadership team.

Then align on the financial plan

Next comes the financial plan.

That’s a post for another day, but the short version is: this works best as a top-down collaboration between:

• CEO

• Head of Finance

• Sales, Marketing, and Customer Success leaders

You want a model that reflects reality (pipeline, conversion, churn, margins) and constraints (cash, runway), supports your strategy, and is aspirational.

Once you’ve aligned on the financial envelope, you can translate it into priorities and initiatives.

Turn strategy into clear annual priorities

Founders should use the planning process to define the 3–5 key priorities for the business next year, along with the key initiatives that support them.

Doing this as a team exercise not only improves the thinking; it also creates buy-in. People execute harder on a plan they helped build.

Good priorities should:

• Support your mission, objective, strategy, and financial goals

• Be aspirational but specific – big enough to matter, clear enough to act on

• Ideally should begin with a verb and finish in the form of from X to Y by Z

• (If possible) Include a Lead Measure that is:

  • Predictive (moves before the lagging result)

  • Influenceable (the team can directly impact it)

  • Ongoing (tracked weekly or monthly)

By the end of this step, you should have a short, sharp list of what must happen this year for the business to move meaningfully forward.

Translate priorities into initiatives

Once you’ve defined priorities, you need initiatives starting with Q1 – the concrete efforts that will move the needle.

Strong initiatives should:

• Clearly impact a specific priority in a given quarter

• Have a single executive owner (one throat to choke / one leader to unblock)

• Be written in the same “from X to Y by Z” form whenever possible

• (If possible) Include a Lead Measure for each initiative that is:

  • Predictive

  • Influenceable

  • Ongoing

If you can’t articulate an initiative in that way, it’s probably too vague.

How to run the team planning exercise

Here’s a concrete way to run the annual planning workshop with your team.

1. Determine what’s working and what’s not

Start with reality.

Individually:

• Ask each leader to brainstorm:

  • 3 things that are working in achieving your mission, objective, strategy, and financial goals

  • 3 things that are not working

As a group:

• Synthesize the lists into a single view of:

  • What’s working

  • What’s not working

This grounds the conversation and surfaces disagreements early.

2. Brainstorm and choose priorities

Next, move from diagnosis to direction.

Individually:

• Ask each person to brainstorm 3 priorities for next year

  • Start aspirational: they should drive your mission, objective, strategy, and financial goals

  • Encourage them to incorporate what’s working and what’s not

Then, individually:

• Have everyone rank their top priorities by distributing 10 votes across the combined list

  • They can put all 10 on one, or spread them out, but they must use all 10

As a group:

• Combine and discuss the results

• Use the votes as input (not the only input) to converge on the company-wide priorities

3. Choose initiatives and lead measures

Once you’ve agreed on the priorities, design how you’ll actually execute.

As a group:

• Brainstorm at least one Q1 initiative for each priority that includes:

  • The current result (where we are now)

  • The future result (where we want to be)

  • The timeline (by when)

• For each initiative, brainstorm a Lead Measure that:

  • Is predictive of the result you want

  • Is influenceable by the team

  • Is measured on an ongoing basis

Then have a group discussion to challenge and refine:

• Are these the right initiatives?

• Are the owners clear?

• Are the lead measures realistic and trackable?

4. Agree on a cadence of accountability

A plan without a cadence is just a document.

As a final step, agree on a cadence of accountability:

• How often will you review priorities and initiatives? (e.g., weekly, biweekly, monthly)

• Which metrics (lead and lag) will you look at?

• How will you track ownership and progress?

This is where execution actually happens. The annual plan sets the direction; the cadence keeps you honest.

Annual planning is one of the highest-leverage moments in a CEO’s year.

If you use it to clarify mission, align on a realistic financial plan, define a small set of priorities, and commit to concrete initiatives with a real cadence, you’ll give your team something far more powerful than a budget: a clear path to run on.

What Are Professional Services, Exactly?

Professional Services (PS) encompass the project management and technical resources that help customers successfully implement your software. Think of it as white-glove implementation support. This is distinctly different from Customer Success, which focuses on ensuring customer happiness, driving renewals, and maximizing long-term customer value.

The Paradox of AI Agents and Human Assistance

It might seem paradoxical: AI agents are supposed to eliminate manual work, right? So why is professional services (PS) so important for AI agent companies? It turns out the final 10% of AI agent implementation is challenging.

AI agents exist to automate complex workflows, and that often requires deep data integration, seamless interaction with existing applications, and nuanced "human-in-the-loop" considerations. Moreover, LLMs are inherently non-deterministic, requiring robust evaluation and ongoing performance monitoring.

Finally, the landscape is constantly shifting. New LLMs emerge weekly, and the pace of technological advancement of the broader LLM infra stack is relentless. Given these factors, AI Agent startups are quickly discovering that PS is a critical ingredient for success.

The Multifaceted Benefits of Professional Services

This isn't new territory for enterprise startup founders. While some view professional services as a necessary evil, many experienced enterprise software company builders see PS as a strategic advantage – going beyond mere project support to enhance customer relationships, accelerate product adoption, and differentiate their offerings in a crowded market.

Here’s a breakdown of the key benefits of offering Professional Services:

A Case Study in Professional Services: GigyaWorks

At my startup, Gigya, as we moved upmarket to the enterprise, we discovered the power of professional services, branding our implementation methodology as "GigyaWorks." This wasn't just a set of best practices; it was a key selling point.

During the sales process, we showcased GigyaWorks as our proprietary approach to ensure customers received maximum value from our solution. Some might think this slows down deals, but we found the opposite to be true. GigyaWorks offered a non-technical differentiator that set us apart from the competition.

GigyaWorks promised to "Start Fast, Stay Fast," with a realistic 90-day go-live timeline. This was for a product that required deep integration within a website's login/registration experience – a mission-critical area involving data, security, and UX. Setting this expectation upfront didn't deter customers; it built credibility and demonstrated our understanding of the implementation challenges.

Packaging and Pricing Your Professional Services

Professional services are typically billed on a "time & materials" or "fixed-bid" basis. Some companies also offer recurring professional services, which can be bundled into subscription bookings.

Here's a quick rundown of pricing models:

Enterprises typically expect to spend anywhere from 50-200% of their first-year subscription license cost on professional services, billed at rates ranging from 150−350+ per hour for U.S.-based resources. Rates are often lower (30-50%) for offshore or nearshore resources.

Some startup CEOs fear that professional services will cannibalize software sales. In my experience, this is rarely the case. Enterprises usually have separate budgets for software and services, and accessing the services budget can often be easier.

Here's a general guideline for typical hourly rates:

Profitability and Utilization: Finding the Right Balance

In the early days, PS doesn't need to be a profit center. It's perfectly acceptable to run it at break-even or even as a loss leader to reduce churn and boost product stickiness.

Longer-term, successful software companies often aim for a gross margin of 15-30% on services – enough to cover costs and contribute, but not so high as to detract from the core mission of driving software adoption. High utilization rates (~75%), and disciplined pricing are crucial for achieving these margins.

Here's how utilization rates impact profitability:

The Valuation Question: Does Services Revenue Hurt Your Multiple?

Another common concern among startup CEOs is whether services revenue will negatively impact their valuation multiple. The general rule of thumb is that as long as services revenue remains roughly 25-30% of total revenue or less, it won't significantly drag down your multiple.

Here's a look at how the revenue mix impacts valuation:

The Bottom Line: Embrace Professional Services

Professional Services are a powerful tool for AI Agent startups targeting mid-market and enterprise customers. Done right, this can be a differentiator during the sales process, ensure a successful implementation, and create deep stickiness, both from creating a deep product integration but also from a customer relationship management perspective.

Note, at Gigya (where I was the CEO for a decade), we increased average contract values (ACVs) from $10K → $40K → $75K → $125K → $250K, each subsequent year, with largely the same core product offering. The gains we achieved were largely due to improving the way we priced and packaged the product.

In this post, I’ll share the five core pricing models (Platform, Seat-based, Consumption-based, Add-on, and Outcome-based) and specific recommendations on how to package these models based on your stage and what your (AI) product actually does.

Let’s dive in.

The Big Picture

You can think of pricing for AI software as a balancing act between (a) how your customers perceive value, and (b) how you want to capture that value. Traditional SaaS had it easy: seat-based pricing with feature-tier gating (Basic/Pro/Enterprise). AI-based software flips the script because you often have real usage costs, plus the potential to deliver (and measure) actual outcomes that your customers really value—like driving more sales or saving them real dollars on operational costs.

Here’s a quick reference chart summarizing the five main pricing models:

Use this matrix as a starting point to figure out which approach might resonate with your buyers and your product’s unique value prop.

1. Platform Pricing (Flat or Tiered)

Platform pricing is essentially a subscription fee for broad access to your AI’s capabilities. Think: “$X per month (or year) for unlimited usage,” or tiered packages (Basic/Pro/Enterprise).

• Pros

  • Dead simple. Finance teams at big companies love a predictable budget line item.

  • Encourages organization-wide adoption without fear of usage overages.

• Cons

  • You might get hammered by outlier usage if you don’t have a usage policy.

  • Don’t get to take advantage of as many upsell opportunities as the product gets more usage

• Best For

  • A starting point for offering additional pricing modules.

  • Early-stage deals when you just want the big logo on your site.

  • Enterprise deals where the buyer demands cost certainty or hates the idea of pay-per-use.

Tip: If your AI agent has real variable costs, consider capping usage or layering in fair-use language.

2. Seat-Based Pricing

Seat-based (per user) is the classic SaaS approach. You charge, for example, $30/user/month, and if the customer adds five new reps, you get expansion in monthly revenue.

• Pros

  • Familiar: Everyone from Slack to Salesforce uses it, so it’s easy to explain in the sales process.

  • Works if each user truly experiences separate, individual value—like AI coding assistants for devs.

• Cons

  • Value misalignment for automation-based AI. If your AI takes over tasks that used to require a whole team, then your seat count drops—so you lose revenue as your product gets more successful!

  • Doesn’t account for usage intensity. One user could be hammering the AI, while another barely logs in.

• Best For

  • Tools where human usage is still the main driver of ROI (e.g. dev assistants, sales writing tools).

  • High collaboration or user adoption contexts. You grow revenue as they roll it out to more employees.

Tip: If your AI replaces people, or if it’s more of an agent than a user tool, you might consider charging per “AI agent” seat. That keeps the seat logic but doesn’t punish you if they automate away real humans.

3. Consumption-Based Pricing

Consumption-based or “pay-as-you-go” means you charge per API call, per message, per CPU-hour, per thousand tokens—whatever usage metric fits. This approach is super common in cloud infrastructure. With the rise of generative AI, it’s everywhere.

• Pros

  • Ties revenue to actual usage, so if customers scale up, you make more.

  • Can attract new customers with a low or even free entry tier, then let them scale.

• Cons

  • Bill shock is real. If a big marketing campaign suddenly triggers an explosion of usage, your customer might freak out at the next invoice.

  • Hard to forecast your own revenue. Investors like predictable recurring revenue, so you might have to get creative with how you present your MRR.

• Best For

  • Developer-facing AI products, where usage is easy to measure and devs expect usage-based pricing (OpenAI, Amazon Rekognition, speech-to-text services, etc.).

  • Use cases where value is clearly tied and aligned to usage

Tip: Provide usage dashboards and notifications. The more you can help customers manage their usage (and thus their bill), the better your relationship.

4. Add-On (Feature) Pricing

Add-on pricing is basically “À La Carte.” You’ve got a base subscription (e.g., bronze, silver, gold), but advanced modules or specialized AI capabilities cost extra. Think: “$30/user/month for the base app, plus $10/user/month if you want the advanced generative AI feature pack.”

• Pros

  • Great for upselling. Start small, then add more modules as the customer sees ROI.

  • Customers only pay for what they need, so it can be a compelling pitch.

• Cons

  • Can quickly create packaging confusion if you slice features too finely.

  • Customers might feel nickeled-and-dimed if you label core features as paid add-ons.

• Best For

  • Broad platforms with distinct modules. E.g., “core CRM” plus “AI analytics,” plus “AI lead-scoring.”

  • Verticals with specialized compliance or domain features that not everyone needs.

Tip: Keep the menu limited. One to three well-defined add-ons is usually enough. If you have dozens of little add-ons, it starts to feel like you’re running an App Store inside your product.

5. Outcome-Based Pricing

Outcome-based models tie your fees to specific results. For instance, you charge $2 per successfully resolved customer support ticket, or a percentage of new revenue your AI-driven ads generate.

• Pros

  • Easy ROI pitch: “You only pay when you actually get value.”

  • Potentially big upside if you take a slice of significant revenue or cost savings.

• Cons

  • Defining the “outcome” can be messy. Who decides if a support ticket was “resolved” by the bot vs. a human agent?

  • Can lack predictability that is usually the foundation for business models for enterprise software vendors.

• Best For

  • Customer Support: Many new AI support bots charge per ticket resolved. If the bot fails, the ticket goes to a human, and the company doesn’t pay for that.

  • Sales: Pay per qualified lead, or a percentage of revenue gain from your AI’s leads.

  • Cost-Saving Automations: E.g., “We take 20% of the cost savings we generate.”

Tip: If you go outcome-based, consider a hybrid: a small base fee plus a success fee. That way, you’re not on the hook for indefinite costs if it takes longer to see results.

Hybrid Approaches & Final Advice

In reality, the best approach is likely a hybrid approach where you use multiple pricing strategies to maximize value, for example:

• Base subscription (platform or seat) + usage overages: You get baseline revenue but still capture upside if they use alot of the AI.

• Seat + consumption: Each seat pays a monthly fee, but heavy usage or growth over time triggers an additional per-use charge.

• Add-on modules + outcome-based: The advanced AI module might have a success-based component.

A few final pointers:

1. Start Simple: If you’re an early-stage startup, sales velocity is super important so your best bet is to pick the model that is relatively easy for your target customer to grasp and for you to manage. You can always refine or expand your pricing later.

2. Optimize Later: As your company scales, experiment with pricing models that will help you maximize ACV and Net Dollar Retention per customer. This later point is super important - you want to see each customer have a chance to grow 25-50% per year in upsells.

3. Speak the Language of ROI: Whether it’s seat-based or outcome-based, always connect the dots for your buyer: “Here’s how we save/make you money.” It makes higher price tags more palatable.

I’ve seen all these models in the wild. The best approach boils down to the fundamentals: Which pricing mechanic best reflects caturing the value your product delivers?

Why It Matters

Focusing on PGMF from the start helps you identify:

1. The best customer segment for your product (Enterprise, Mid Market, SMB, or Startup).

2. A viable sales model (field sales, inside sales, self-service, etc.).

3. A path to predictable and efficient revenue growth.

Many successful companies have built billion-dollar businesses targeting different segments. Salesforce, ServiceNow, and SAP went after enterprise, while Intuit, Shopify, and Mailchimp focused on SMBs. Others like Stripe, Slack, and Zendesk initially zeroed in on startups. Over time, some companies move upmarket or downmarket (though downmarket is less common).

The challenge for early-stage founders is deciding which segment to target first. Below is a quick-reference matrix to help you choose your ideal customer segment to find PGMF:

Enterprise

Many founders try to avoid enterprise sales due to its long, complex sales cycles, extensive security and procurement requirements, and frequent travel. However, certain products and markets require an enterprise approach—especially if:

• The product is highly complex.

• Multiple stakeholders must sign off.

• There are extensive integration requirements.

• High-touch implementation is necessary.

If you check these boxes, enterprise may be your best bet. Be prepared for 6–12+ month sales cycles and field reps earning $250K–$350K on-target earnings (OTE), with quotas in the $1.2M–$1.5M range. Contract values typically land in the six-figure range to justify these sales costs. The pitfalls to this approach are well known:

• Slow sales velocity, which hurts ability to iterate on the product and achieve growth milestones

• High initial cost of sales, which is challenging especially early on in the sales cycle

• Often requires experience having sold into enterprise

Mid Market

Targeting mid-market customers, especially initially, can be a compelling alternative to going enterprise. Sales cycles are generally shorter, with fewer stakeholders and streamlined procurement processes. You can often succeed with inside sales reps who earn $150K–$250K OTE and carry quotas in the $600K–$1M range. These reps don’t necessarily need extensive rolodexes, typically have 3–5 years of sales experience, and can operate from a “hub” office. Sales cycles are typically 3-4 months and deal sizes here usually fall in the mid- to high five-figure range to justify the sales costs.

The mid market segment is probably the best choice for most startups, yet there are still pitfalls:

• Low contract values that require too many transactions to hit quota

• Multi stakeholder sales that extend sales cycles hurting deal velocity

Startup

Often referred to as the “YC playbook,” selling to other startups appeals to founders because it’s relatively frictionless. Sales cycles are short, usually involve a single decision-maker, and come with straightforward buying processes. Many young companies are also more willing to take a chance on new solutions, and often have “greenfield” opportunities (no incumbent product in place).

For this segment, founder led sales is the most common, followed by inside sales reps earning $100K–$200K OTE with quotas in the $400K–$600K range are common. Contract values can range from $10K+. This can be a seemingly attractive market to start with, but there are pitfalls:

• Startups’ spending heavily depends on the funding environment (booming in 2021, tougher in 2023), and churn can be high.

• As you grow and inevitably move upmarket, you may have to reestablish product-market fit for larger customers.

SMB

Selling to SMBs requires an ultra-low-touch or no-touch approach to remain profitable. Ideally, the end user is also the buyer, with enough budget authority to put the purchase on a credit card. Aim for as much self-service as possible—maybe just one or two calls, and a deal cycle of a few weeks at most. Success here hinges on a frictionless product experience and onboarding process.

Pitfalls include:

• Sales that require multi-stakeholder approvals or extensive customization—both of which drive up cost and complexity.

• High churn, as many SMBs go out of business each year.

Despite these challenges, well-designed SMB plays can scale enormously (think Intuit or Shopify).

Conclusion

It’s not enough to achieve product market fit - you must figure out your product-go to market fit (PGMF). Getting this right upfront - seed or series A stage - can save you from a pivot and/or team rebuild after you are millions in sales.

The most common way to analyze market size is by a 3rd party market research report that provides a top down market size analysis. I find those reports to be misleading especially for startups which are often creating new markets. Instead, I have a simpler mental model.

First, what is an IPO-sized market opportunity? Today, to go public you need $200M+ in revenue. Further, market leading companies often only get to 10-15% market share. So that means, for example, if you have 2,000 companies in your total addressable market and execute really well, you might land 200 of these companies in 8-10 years. Therefore, in order to go public, you need $1M+ deals.

Following this logic, I’ve outlined a “rule of thumb” approach to evaluate what average contract values (ACVs), total addressable markets (TAM), and go to market (GTM) is required to build an IPO-able company. By understanding these “physics”, you can make informed decisions about your startup’s eventual IPO-able potential.

When assessing the potential of your B2B startup idea, consider the following guidelines:

1. Whales:

   • TAM: 1000s

   • ACV: $1M+

   • GTM: Enterprise Field AEs w/ $300-350K OTE and $1.5-$2.0M quotas

2. Elephants:

   • TAM: 10,000s

   • ACV: $100K+

   • GTM: Enterprise Hybrid Inside / Field AEs w/ $250-$325K OTE and $1.0-$1.2M quotas

3. Deer:

   • TAM: 100,000s

   • ACV: $10K+

   • GTM: Inside AEs w/ $150-$250K OTE and $500-$800K quotas

4. Rabbits:

   • TAM: 1,000,000s

   • ACV: $1K+

   • GTM: Self Serve w/ Inside AEs w/ $100K OTE and $300K quotas

5. Mice:

   • TAM: 10,000,000s

   • ACV: $100+

   • GTM: Self Serve w/ No AEs

By understanding the “physics” of your startup, including the TAM, ACVs, and GTM approach, you can better assess whether your company has the potential to IPO one day. Note, there are many, many caveats to this rule of thumb, but my hope is it can provide a useful framework for founders.

Let's break it down into three phases of scaling sales in an enterprise software company:

• Discovery: Reaching Product-Market Fit

• Blueprint: Building a Go-To-Market (GTM) Motion

• Scale: Perfecting the Ramp Model

Discovery: Reaching Product-Market Fit

During the discovery phase, the main goal for founders is to achieve Product-Market Fit (PMF). We all know what PMF is about: making sure a startup's product or service meets the needs and demands of its target market. For an enterprise software company, this typically happens when you're at the $0-$1M ARR range. You'll know you're on the right track when you have a handful of referenceable customers (usually 5-10, depending on the Go to market model and target customer) and a "painkiller" solution.

Blueprint: Building a Go-To-Market (GTM) Motion

In the blueprint phase, the goal is to create a repeatable GTM motion. This covers everything from promoting and distributing to selling and implementing your product or service. You'll likely be in the $1-$10M ARR range during this phase. Success here looks like a standardized sales playbook and reps bringing in 3x their On-Target Earnings (OTE) in new annual bookings. So, if a rep's OTE is $250K a year ($125K base and $125K commission), you'd want them consistently generating $750K a year in annual bookings.  The Blueprint phase often is overlooked and involves building out the sales playbook like the one I shared when I was the CEO at Gigya.   

Scale: Perfecting the Ramp Model

The scale phase is all about replicating your GTM motion. It may sound simple, but it gets complex with factors like org design, capacity planning, sales enablement, pipeline plans, and forecasting. To succeed, aim for a repeatable sales model with reps consistently hitting around 80% of their quotas and 2-3x new ACV growth. This phase is what propels a company to scale and reach that $100M+ milestone.

In upcoming posts, I'll dive deeper into each of these phases, discussing the aspects to focus on and the metrics you should be tracking. Stay tuned!

Unlike traditional sales playbooks that focus on qualifying prospects out of a sales process, modern sales playbooks aim to teach buyers to see the world in a new way and create demand where it might not have existed. This is especially crucial when creating a new category or asking a customer to replace an existing product.

The sales playbook can be divided into three steps: 1) Confirm Need; 2) Why New Product; 3) Why Company’s Product.

A) Confirming Need: As a startup, you should clearly understand the need(s) your product or service addresses for potential prospects. There may be a single need, several overarching needs, or even sub-needs. During the sales cycle, a seller should have a point of view based on research that the prospect likely shares. The goal of the conversation is to use insight and teaching to make the buyer recognize and agree that they have one of those needs.

For example, at my last company, we helped customers with three overarching needs: 1) Improved customer experience; 2) Know your customer; 3) Consumer Privacy Compliance. This wasn't about discovery, but alignment and teaching. When selling to a major CPG company like Nestle, I might say something like, "In conversations with other CPGs like Unilever and Colgate, we've observed that building direct digital relationships and centralizing customer understanding across their brands for better lifecycle marketing is a top priority. They've discovered how 'customer identity' can be a unifying thread to make this happen." This approach sparks a response, leading to a productive back-and-forth conversation.

B) Why New Product: Once the buyer agrees they have a need, the goal is to convince them that your category of product (not specifically your company's product) can solve their problems. This is crucial because there are often multiple ways to address a need, and without aligning on the category, your product might get lost in the mix. To achieve this alignment, provide valuable analysis the customer can use, like scorecards and benchmarks.

C) Why Company’s Product: Most companies want to focus on this part of the playbook, but it should only come after aligning with the need and category. At this stage, the prospect agrees to build or buy the product within a specific timeframe. Typically, this part of the playbook includes a "review plan" with a 1) Functional Review; 2) Technical Review; 3) Implementation review. The goal is to emphasize why your product is the best fit.

The final stages of the sales playbook are procurement and signature, which, as a side note, often take longer than expected when selling to enterprises.

The key takeaway is that it's crucial for founders to be involved in establishing a sales playbook. This playbook should foster a culture of insight and teaching, initially focusing on aligning with the customer's needs and demonstrating how the product category addresses those needs. Only then should the focus shift to a technical review of the product itself. This approach helped me scale Gigya to a $100M ARR run rate, and I believe it holds true for many companies today.

Appendix:  Here is the playbook we leveraged at Gigya, my last company:

1. Confirming Need

2. Why New Product

3. Why Company’s Product

4. Procurement, Signature, Win