This was a full-circle moment for me — I started my career there in Chicago in 2010.

My now co-founder Joe Malandruccolo started around the same time, where we spent many hours in the Grand conference room overlooking the Chicago River, whiteboarding ideas and talking about maybe starting a company someday.

Joe eventually left to get his master’s in computer science. I worked at Mercer for four years before moving on to Juniper, then Workday.

A few years later, we met Taylor Cone and together started Compa.

And now we’re partnering with Mercer to change compensation forever with AI.

Why we partnered with Mercer

Earlier this year, when we announced our $35 million Series B, I wrote that our vision at Compa is to combine live market data with AI-powered analysis.

This partnership is one way we’re bringing that vision to life.

Compa customers can now use our Analyst Agent with Mercer compensation insights built in, dramatically expanding on our live market data with Mercer’s global survey dataset and expertise, and unlocking an incredible new level of analysis and scope of data.

What does that look like in practice?

Comp teams can use Analyst Agent to:

• Analyze Mercer survey data automatically

• Benchmark using both Compa live data and Mercer surveys to expand depth and breadth of analysis

• Build custom playbooks to automate workflows like analyzing retention risk or offer exceptions

• Model range updates and merit budgets with custom logic balancing multiple datasets

Instead of pulling data, building spreadsheets, and analyzing each question manually, comp teams can direct Analyst Agent to do the analysis and focus more time on strategically partnering with the business.

Share

The future of the comp industry

Every compensation team, and every enterprise function, is rapidly building AI workflows to increase automation and insight velocity.

I believe AI only succeeds in comp with deep, trusted expertise behind it — both built into the AI itself, and supported by human experts.

Compa’s partnership with Mercer underscores this belief by making it possible to fulfill an exceptionally high bar of trust in leveraging AI responsibly and accurate in high-stakes compensation decisions.

That’s the future we’re building at Compa with partners like Mercer and comp teams from Fortune 100 companies already putting it to work.

Congratulations to the Mercer team for their vision alignment and commitment to changing the industry with us, especially Stephanie Penner, Tauseef Rahman, Kevin Poff, and Michael Anderson. And thank you to many of my Companions who’ve worked hard to bring this partnership to life for our customers.

Join us on September 10 to learn more about our partnership! Register here.

Peer Group is a newsletter for comp leaders navigating fast-moving markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

What is happening in the market right now?

• Oil prices are up 90%+ YTD from rising geopolitical instability

• Over half of the S&P 500 tech companies are down YTD as rapid advances in AI breach software moats

• Layoffs are spiking again as companies rethink how work gets done

How does all of this change impact compensation?

At Compa, we attempt to answer a large (and growing) percentage of this question with our intelligence products. To answer the rest; we invest in our compensation community, ranging across webinars, roundtables, dinners, job boards, flash polls, market research and more.

And our biggest community event of the year, Compalluminati, is just one month away.

Here are the details.

Share

300+ comp professionals at Snowflake HQ

Compalluminati is a two-day exclusive summit held each year for compensation leaders at Fortune 500 companies.

We love to co-host the summit with our customers because Compa is built by comp for comp, and that includes building our community. Past years’ summits were held at Roblox, Stripe, and DoorDash.

This year we’re partnering with the phenomenal compensation team at Snowflake at their beautiful Menlo Park headquarters in the Bay Area.

• Dates: May 6-7, 2026

• Location: Snowflake HQ, Menlo Park, California

• Who: Free registration for compensation leaders

• Details: Go to compalluminati.com for agenda

This year’s theme is managing billions with AI in fast-changing markets.

I believe this is the most exciting time in history to be in compensation – the stakes have never been higher and opportunities to create impact have never been more democratized.

What you can expect:

• Learn from your peers, with 26+ leaders speaking from compensation and HR teams at Snowflake, Salesforce, NVIDIA, Cisco, and many more

• Get hands on with the latest AI in comp, including workshops and case studies

• Hear from Compa’s research team about the state of the market and latest trends

• Connect with old friends and make new ones over meals and drinks – community means having fun, too!

And of course, I will have some big announcements to share about our latest innovation at Compa.

We are nearly at capacity – you can register here to join us while spots last.

See you in May!

Peer Group is a newsletter for comp leaders navigating fast-moving markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

I wrote about the hourly worker compensation problem when we launched Compa Frontline last year, characterized by three fundamental market dynamics:

1. Massive, highly-fungible talent market

2. Rapid change driven by seasonality and local legislation

3. City-block level competition

Last week, Ashley Case on our team wrote an article diving much deeper into the location problem to understand why geography — down to the zip code — matters so much in the frontline market.

She is a seasoned comp expert and works closely with many teams who hire frontline workforces facing this challenge, and I found her article fascinating.

So I wanted to share it with you here, presented in full below, with Ashley as guest author.

If you don’t regularly work on comp strategy for hourly workers, I think you’ll find it surprisingly interesting how the challenges are equally great to the knowledge worker markets, but express differently.

And if you do work on hourly worker comp — this articulates some clear approaches for setting frontline pay in a very fast-moving market.

Share

Going Deeper on Hourly Compensation: Why Zip Codes and Proximity Matter

Within the same MSA, hourly wages can swing wildly by zip code.

Yet most pay strategies stop at the MSA level as if a metro area is one labor market.

It’s not.

That shortcut creates risk: overpay in some pockets, lose talent in others.

The trouble with MSAs

MSAs make benchmarking simple by bundling multiple labor markets into a single number and calling it “the market rate.”

Take the Boston–Cambridge–Newton MSA. Median base pay for frontline roles is $18.75.

At the zip code level, wages range from $18.41 to $25.92 but in central Boston, median base pay is $25.77.

Price that role at the MSA average and you’re nearly $7 below the local market.

That’s not benchmarking, it’s guessing.

Understanding talent movement

Years ago, I worked for a company with call centers.

Our starting wage was competitive, until a new center opened down the street with a banner advertising $0.50 more per hour.

Overnight, the MSA rate didn’t matter but the sign down the street did.

Talent doesn’t compare MSAs, it compares commutes.

People will travel for higher pay but only if the wage clears the cost of getting there.

Chicago is a great example of this.

If you’re hiring downtown, base pay typically lands between $18.50 and $19.00.

Steps outside the city center and neighboring zip codes can differ by more than $1.00 an hour.

That gap drives movement.

When you’re setting pay, what the next zip code offers and how far it is matters more than a blended downtown average.

Other local pay factors

Other factors that influence pay at the zip code or local level include,

• Commercial vs residential proximity: Locations near logistics hubs, retail and industrial corridors, and hospitals can command higher hourly rates than residential areas

• Talent availability: When employees can’t find or afford housing/transportation near their workplace, a premium may be required to support the longer commute.

• Municipality minimum wage: Variations in minimum wage at the city or county can impact wages

In Portland - Vancouver - Hillsboro MSA, which spans Oregon and Washington municipalities, median entry level (S1) wages vary from $16.32 to $19.30. At the low end, median hourly pay in an Oregon zip code falls below Washington’s minimum wage requirements, adding complexity to pay variance within MSAs.

Approaches for setting frontline pay

Understanding the nuance of local data, how should companies think about setting compensation for new and existing locations? Here’s a few ways:

In hourly hiring, local data makes or breaks your compensation strategy.

Talent movement and other local pay factors can create considerable wage differences from one zip code to the next - in the example of the Boston - Cambridge - Newton MSA, the highest paid zip code is 38% higher than the MSA median. By leveraging local wage data beyond the MSA-level, pay strategies can better attract and retain frontline staff.

Want to dig in deeper? Watch Ashley's recent webinar, Cracking the Code on Frontline Compensation to learn how top comp teams are solving frontline pay and fungibility problems that have held hourly roles back for years.

Data Source:
Compa Frontline, S1 and S2 level in-store retail roles. As of 11/12/2025.

Peer Group is a newsletter for comp leaders navigating fast-moving markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

Passions and predictions on software’s future aside, this sudden shift in market value massively impacts the talent markets through stock compensation, even if only temporarily.

Savvy compensation teams will move fast to analyze the market change and develop strategies to defend or compete.

Before we get into the math — are you taking action to compete or defend?

This is how to quantify the risk and opportunity across beleaguered software companies and high-growth AI companies by focusing on the value of unvested equity.

Measuring the market rift

Let’s take a basket of 20 software companies and 20 AI companies and look at how their stock prices changed over a three-month period from November 18, 2025, to this week closing February 18, 2026.

• The software stocks are down -30.5% on average, ranging from -13% to -49.7%. This is the meltdown we’re all reading about.

• Over the same period, the AI stocks are up +23.2% on average, ranging from -19.1% to 107.7%.

Now you can debate the exact composition of these comparator groups, and you should — in your own analysis, refine it to the peer companies you compete against, and sharpen the math with market cap or employee size weighting.

But make no mistake, the market is moving fast and non-uniformly, exposing risk and opportunity for talent competition.

Share

Impact to unvested equity value

Unvested equity holds help retain talent and defend against competitors picking off our employees. So when stock value declines, retention risk rises.

If everyone’s stock goes down, we all ride the market together as an industry.

But if my stock does down and yours goes way up, my risk is your opportunity. This is precisely what’s unfolding now between software and AI companies.

The relative change in unvested equity value is a massive +77% advantage to AI companies.

Let’s look at a P4 Software Engineer as an example. Using Compa we can get the market median unvested equity value at $181,630 in November 2025.

All else equal, this engineer’s unvested equity hold dropped on average to $126,128 at software companies and grew to $223,737 at the AI companies.

If the software company and AI company made the exact same target grant values, the AI company now has a $97,569 greater retention hold over the software company.

Harder to defend software, easier to compete in AI.

Going into the refresh grant cycle, comp teams who ignore the market change and age forward last year’s median survey grant value risk massively underpaying and hemorrhaging top talent, or massively overpaying and missing a market opportunity.

Here are some defensive and offensive strategies to compete for talent during the software meltdown.

Defensive and offensive strategies

What are the opportunities, imperatives, and constraints for defense (software) and offense (AI)?

Software defense

• Brain drain mitigation: Your top performers are most mobile and at-risk — concentrate expanding unvested equity holds here by focusing on a ratio >1 of unvested equity to the median AI company new hire grant value + sign-on bonus

• Voluntary attrition: You may quietly welcome voluntary attrition to manage costs and avoid layoffs — but protect key talent and teams that must weather the correction

• Governance limits: Flexibility is constrained by stockholder-approved share pools and burn rate limits, but SBC expense goes down with lower prices — negotiate where you have the most flexibility, or turn to alternatives like cash and benefits (but beware of attracting a different talent profile)

• Front-loading: Experiment with pulling equity grant value forward in the form of shorter vesting schedules and front-loaded grants, especially for current employee retention

• Talent marketing: Market to new hires and refresh grant recipients that they are “buying low” and if the correction is overblown they could experience historic upside

AI offense

• Talent discounts: Take advantage of the talent arbitrage; top performers at software companies are available at a steep discount — their target unvested equity hold could be trading closer to market median

• Aggressive offers: Shift stock compensation budgets from current employees with high unvested holds towards new hire grants to maximize top talent attraction

• Speed up growth: Pull forward annual hiring plans to Q1 to take advantage of market conditions, and ramp up marketing to generate talent market availability while software sentiment is down

• Software targets: Temporarily construct a public software company peer group distinct from private AI competitors where you could concentrate hiring on top technical talent

• Surgical tactics: Monitor software company layoffs and disclosure of exit package terms to selectively target top talent sources in your priority geographic markets

The cost of inaction

The Wall Street fears may be temporary or overcorrected — perhaps the software companies rebound and the AI stocks reprice, or perhaps not.

But either way, the next few weeks or months at a minimum represent a tactical opportunity for compensation teams to play smart defense and offense in the talent markets.

Inflexible teams relying on last year’s assumptions and stale data will make mispriced pay decisions.

Adaptive teams monitoring market conditions and real-time data will quickly and quietly accumulate advantage.

Peer Group is a newsletter for comp leaders navigating fast-moving markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

Yesterday, we announced Compa’s $35M Series B funding round — an incredible milestone for our company, and for the compensation industry.

I want to share my perspective on how we got here, my vision for the future of compensation, and my gratitude for the people behind Compa’s extraordinary story.

1. Why this fundraise happened

First, we founded Compa to bring real-time data to an industry starved for innovation and better insights as markets move faster and expectations grow.

Our team focused on rapidly shipping new ways of understanding the market, beginning with offer data to measure market change, then deep stock comp insights, skills-based market precision, map-based data for hourly workers, and now all-employee market data — a real-time alternative to traditional surveys.

Each of these data products focused on answering the same, simple question: what’s happening in the market right now?

As a result of this innovation, and commitment to the highest data quality standards for enterprise teams, the network has exploded since launch in 2023, and in just the past 12 months has grown over 8x. Today, Compa is the largest real-time market data platform in the world with over 7 million observations in 35 countries.

Second, we brought agentic AI to the compensation industry, focused on intelligence — synthesizing real-time market data, automating compensation analysis, and applying new insights to the thousands of decisions companies make each year.

In 2025, we launched our agents with an extraordinary group of leading compensation teams at the world’s best companies. We concluded that program with agents achieving 67% accuracy, up from just 38% only four months prior. This technology is accelerating rapidly, and it represents the potential to completely transform how we practice compensation.

Agentic AI has the potential to completely transform compensation, and our Series B investment positions Compa to build that future.

This is my vision.

Share

2. My vision for the future of compensation

Fundamentally, compensation teams are asset managers of multi-billion-dollar portfolios, stewarding a company’s largest investment: human capital.

The problem is that with surveys and spreadsheets, we attempt to match manual, slow analysis to one of the fastest-moving capital decisions in the enterprise. As a result, the business moves faster than comp, creating all the familiar problems across fairness, competitiveness, and cost.

Compensation teams need to manage human capital the way Wall Street manages financial capital — continuously, with low-latency analysis, predictive signals, controls, and intelligence.

Our vision for the future of comp — and what we’re building at Compa — is not just real-time data, but also real-time analysis and decision support.

Proliferating data powers automated analysis and just-in-time intelligence at the point of decision, driven by agents. Real-time data doesn’t just complement surveys; it can change how we work.

Modern compensation intelligence exactly reverses the flow of work.

Instead of responding to problems raised by the business, we automatically detect risk as the market changes. Instead of pulling market data to manually analyze, agents push novel insights and automate analyses we humans orchestrate.

It’s a shift from reactive to proactive. The science automates with AI, and the art gets smarter with humans.

This is the future we’re building at Compa, with many of the world’s best companies already using this technology today.

3. Gratitude

I spent the first 10 years of my career in large enterprise compensation. In building Compa, I’m doing my life’s work to transform this industry. We succeed when compensation is fair and competitive for everyone.

Thank you to our customers who share this vision — we spend every day focused on making your lives better.

Thank you to our extraordinary team at Compa, including my co-founders Joe and Taylor, and all our employees — you deserve to celebrate this milestone!

Thank you to our investors who continue to underwrite this vision, and to our partners who co-innovate with us.

And finally, thank you to my wife and (now three!) children for supporting me as I build two startups at once: my family and Compa.

PS: Join us at Compalluminati 2026

We are gathering leading compensation teams for our third annual summit, Compalluminati, co-hosted this year with Snowflake at their headquarters in Menlo Park.

Join us to network with your peers, learn from expert panels and workshops, and be the first to learn about Compa’s newest innovation and product announcements!

You can register early here. More details coming soon!

Peer Group is a newsletter for comp leaders navigating fast-moving markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

Here are my comp predictions for 2026:

1. AI begins to eat comp as we know it. Analysis moves from spreadsheets and Tableau to agents and ChatGPT. Mega-manual processes are not optimized with AI, but deleted. This trend concentrates in the tech industry, but becomes highly visible with real success stories, not just hand-waving.

2. Merit increases shift from discretionary to formulaic. More companies lose patience with manager peanut-buttering leading to bad outcomes on pay equity and retention and opt for comp-controlled, formulaic pay changes — prevalence of this approach at least doubles from 2025. Pay-for-performance focuses on bonus and stock, if at all. Shoutout to Ashley on my team for underscoring my conviction here.

3. AI premiums cool off. No, I’m not predicting an AI winter and Mag Seven crash, but downward trending premiums over SWE (especially at higher percentiles), and less spend on sign-on bonuses as the market absorbs this new engineering discipline. I noticed early signs of this last month and wrote about it here.

4. “Non-standard” vesting becomes majority practice. The multi-year trend continues and the standard 4-year vest loses its majority position as companies rethink goals on dilution, retention, and SBC expense. In 2025, 54% of companies used the standard 4-year vest, down from 60% in 2024 and 62% in 2023.

5. We see fewer secondaries for large private companies. In 2025, companies like Stripe, Databricks, OpenAI, and Anduril completed tender offers, providing liquidity for employees while staying private. The IPO window opens further and this approach wanes as more companies go public.

   Share

Grading last year’s predictions:

1. Skills-based compensation has its breakout year. Usage of skills data in comp spiked in 2025 for one particular use case: precision in hot jobs. Deciphering what “AI Engineering” really means or isolating the semiconductor market to analog design talent was an unlock for skills in comp, but many teams are only experimenting at the surface level. Mixed bag, so we’ll score 0.5.

2. Comp teams upskill on finance. Every comp team we work with got closer to, not further from, their finance functions in 2025 as they navigated stock comp pressure, M&A, and massive offer exceptions. Score: 1.

3. Pay equity and pay transparency focuses on Europe. Yep, very quiet on US pay equity initiatives, lots of compliance preparation for European legislation, including paying off some job architecture debt. Score: 1.

4. Stock comp declines in target value and participation for most roles. Sales & Customer domain median new hire grant up 5%, prevalence up 7 points; Corporate Functions down 8%, prevalence down 7 points; Technical up 10%, prevalence down 10 points. This was a mixed result — I was right that G&A would take a big hit, but Sales & Customer went up, and while Technical comp rose fastest we saw a drop in prevalence, perhaps because it’s the biggest pot you can pull from to reduce equity spend. But I’ll take the loss, score: 0.

5. Compensation dips a toe in AI. This was spot on, AI is exploding into comp, but unevenly. We increasingly see barbell-shaped AI adoption — you’re either all in, or falling far behind. Score: 1.

Overall, 3.5 / 5.0. Looking at our poll from last year, we all got EU pay transparency and AI right (maybe it was a lay-up?), but I beat you (qualitatively and anecdotally) on comp teams learning finance!

Peer Group is a newsletter for comp leaders navigating fast-moving markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

Together these three metros account for 76% of AI offers in 2025 year-to-date.

Here’s how it’s trended over the past two years:

Share

The Bay Area leads with 46% of all AI offers in 2025, but has steadily trended downward over the past two years from 57% in Q1 2024.

Seattle accounted for 24% of AI offers in 2025, rocketing up from 9% in 2024, solidly in second place and well ahead of New York.

Focusing on the Bay Area, we see an even stronger lead in AI Research roles, 78% of the entire market, and this lead grew in 2025 while AI / ML Engineering shrank to 45%:

The above charts look at individual contributors, but SF similarly leads management (M) roles more than it does professionals (P):

Looking at compensation also shows SF leadership, but interestingly only in 2025:

All three of the top markets trended very closely in median base salary for P3-P4 AI offers from 2022 until Q1 2025 when New York flatlined and Seattle dropped over 20% and hasn’t recovered since.

This pattern persists no matter how I slice the methodology — median versus average, base versus total comp, weight by offers versus weight by companies, etc.

Seattle and New York are major centers of AI innovation, but SF dominates in research roles, management levels, and compensation.

Share

Looking ahead to 2026

If SF Bay Area, Seattle, and New York lead the United States in AI Engineering & Research, we can look at two signals from Compa’s offer data to forecast where the market heads in 2026: offer volumes and offer acceptance rates.

Both metrics show that AI hiring may cool off next year.

Analyzing offer volumes in the top three metros versus all others, we see a huge pulse of hiring in the first half of 2025 followed by a sharp decline over the past six months:

Meanwhile offer acceptance rates have been rising since May to end the year at 85%, with the top AI metros converging with the rest of the US:

If offer volumes measure level of demand, then demand is falling. And if acceptance rates measure employee versus employer favorability, the markets are shifting to employer-friendly. Both are leading indicators of slowing wage growth.

These early signs suggest the AI boom may slow down going into 2026 — something to consider ahead of Q1 comp planning cycles.

Peer Group is a newsletter for comp leaders navigating fast-moving markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

Every comp team has three archetypal roles: analyst, partner, and operations. And AI is impacting each of them.

Some teams literally organize jobs like this, others are blended together; regardless, the underlying work ultimately fits these building blocks.

I think of each role like a part of the P&L driving the comp portfolio (“top line”):

1. Analyst → R&D: produces insights

2. Partner → S&M: brings insights to the business

3. Operations → COGS: delivers programs

4. (And leadership is the equivalent of G&A)

The P&L metaphor is useful to understand the unit of value each comp role provides:

1. Analyst → Productivity: how much high-quality insight can you produce?

2. Partner → Outcomes: how many strategy-aligned decisions can you manage?

3. Operations → Efficiency: how can you deliver the best service at the lowest cost?

The units of value for each role reveal how AI is changing work in compensation.

Share

Analyst

Compensation Analysts work with lots of data to shape insights, synthesizing inputs like cost, competitiveness, fairness, and performance. Their insights, in turn, are used to guide strategy, set policy, and identify risk.

If Analysts are like an engineering team, then the successful Analysts are more productive — and that’s how AI is changing their work.

AI massively expands analytical capacity. Faster analysis with more data and more complexity to produce more insights. Manual spreadsheets move to automated analysis. Old surveys move to connected, real-time data. VLOOKUPs become agentic prompts and playbooks.

What makes AI exciting for Comp Analysts is that the analytical expansion isn’t incremental, but an order of magnitude — they can finally achieve escape velocity, slip beyond the event horizon, and shift from reactive to proactive, sourcing compensation insights automatically, instead of only inbound problems originating from the business.

Partner

Compensation Partners apply strategy to the business to get smart decisions made. They are the interface between the compensation team and leaders in engineering, GTM, the new business unit, APAC, and so on.

If Comp Partners are like a sales team, then successful Partners get bigger better outcomes — and again, that’s how AI is changing their work.

AI massively expands compensation coverage. Emails, phone calls, and meetings become a low-latency agentic chat interface. Broad salary ranges become highly-curated recommendations at scale. Exceptions and accidental “acts of cash” are eliminated or managed by automated, dynamic rules and goal-seeking.

The big change for Partners is a shift from triage to coverage — more pay decisions touch comp and get smarter, driving more strategy-aligned outcomes and less mess.

Operations

Operational comp roles deliver programs, manage technology and systems, and maintain compliance.

Like a support team, it’s all about delivering programs smoothly at minimum cost and effort — and AI has an important impact here, too.

AI massively improves operational efficiency. This benefit is well-known in support-heavy company functions like customer service and help desks, and it applies to Operations roles on compensation teams, too. Delivering comp planning, answering program questions, and completing high-volume tasks get automated with AI.

So the big gain for Operations is lower administrative costs.

While operational efficiency is always important, comp teams are not large and expensive like major operational teams in the company; efficiency will improve with AI, but I see the bigger impact in compensation with Analysts and Partners, where intelligence dramatically expands to achieve smarter outcomes on the company’s largest investment — its people.

What should comp teams do?

Short answer — begin experimenting.

If you haven’t already, you’re behind. Many, many comp teams are implementing AI capabilities to gain competitive advantage. And as one leader told me last week, real-time data and agents will shift from a competitive edge to handicapped if not deployed in the next 12 months.

Another place to start — later today I’m talking with Nick Larson at T-Mobile about using AI in compensation. Join us live to learn and ask questions about how this phenomenal comp team is innovating with AI.

On a final personal note, I’m back from a writing break since my third child was born in August. Thank you kind readers your patience while I balance Compa and my growing family.

Lots of market trends to discuss, and exciting news coming soon. I’m glad to be back!

Peer Group is a newsletter for comp leaders navigating fast-moving markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

Adoption of AI in compensation is inevitable.

Why? Economics.

Supporting comp decisions with AI agents is 90% lower cost than humans. That, in turn, unlocks new levels of coverage and precision to manage billions in pay each year.

Comp teams working alongside agents will be smarter and more productive. And I believe (per Parkinson’s Law’s induced demand) we will observe more expansion of impact than job displacement — there is simply too much ROI at stake.

In this post, the third and final in my 3-part series, I lay out the math — buckle up — to show why AI can’t be ignored.

And how AI will change compensation forever.

TL;DR on my first two posts

The economics of AI in compensation builds on the core concepts I laid out in my first two posts in this series:

Comp portfolio size and turnover (Part 1)

• Comp teams are like Wall Street firms managing billions in assets, their comp portfolio

• The comp portfolio size is the value of all employee compensation each year

• The comp turnover is the value of all the pay decisions we make each year

Comp team expense ratio (Part 2)

• Comp turnover is high, requiring active management, more like a Wall Street mutual fund than a passively-managed index fund

• The comp expense ratio is the size and cost of your comp team as a percentage of the total comp portfolio

• Most comp teams rationally triage a small percentage of high-value decisions and passively managing the rest, resulting in a lower-than-expected expense ratio

Tech company example

Putting these concepts together for an example 10,000-person tech company:

• With an average TDC of $200,000, the comp portfolio is $2 billion

• Offers, pay increases, bonus payouts, etc., are worth $950 million, for 48% turnover

• The 10-person comp team that costs $3.7 million to manage the $2 billion portfolio has a 0.19% comp expense ratio

• This team has the capacity to actively manage about 5k of its ~27k annual pay decisions covering half the portfolio value, leaving 22k decisions worth $1 billion passively managed

Agent unit costs

Let’s start by positioning agentic AI conservatively as a lower-cost alternative to humans. This understates its value because AI makes the humans way better - we’ll come back to that.

Human costs

We estimated that a human comp expert has capacity to actively manage 500 pay decisions per year based on a 10-person team focusing on the top 5,000 most valuable decisions.

At a fully loaded cost of $250,000, the human comp expert has a unit cost of $500 per decision.

$250,000 ÷ 5,000 decisions = $500 / decision

Yes, of course we could try to find lower-cost people — more junior, different geos, etc. But notice the cost is still high even on the low end: a $60,000 comp analyst in India still carries a $120 unit cost, and that’s assuming decision capacity is just as high.

Adding in management and discretionary costs, we’re spending $3.7 million to manage ~20% of our decisions worth 50% of our portfolio turnover value. The rest are passively managed, meaning comp people don’t review them.

Share

Agent costs are 90% lower

An agent, on the other hand, can manage thousands of decisions simultaneously.

While an AI agent is technically unconstrained, we should characterize the agent as contextually constrained. That is, we may deploy multiple agents focused on different areas, like different people on the comp team*:

• Comp partnering domains: R&D, GTM, and G&A

• Geographic or business units: AMER, EMEA, APAC, LATAM

• Pay decision types: offers, merit cycles, and retention

*How these agents are trained and deployed is critical to success, and was the topic of deep discussion at our most recent agents product workshop; but it is beyond the scope of today’s post.

Let’s assume we train and deploy five comp agents, and each costs $150,000/year, plus $100,000/year in a person’s time to manage them, for a total of $850,000 spend at our company.

What can these agents accomplish, and is it worth it?

In our example company, we have 22,000 decisions and $475 million in portfolio turnover value passively managed by the human team.

With our new team of five agents, we can actively support 90% of the long-tail decisions (some will always be made without comp looking).

That’s 20,000 decisions supported by $850,000 in agent spend, for a unit cost of only $42.50 per decision, compared to $500 per decision for humans.

Agent unit costs are 90% lower than people.

New technologies are economically disruptive when they fundamentally change unit costs. Agentic AI lowers the cost of compensation analysis by an order of magnitude.

With AI, a comp team can radically expand its reach to nearly every pay decision in the company with rational unit costs, unlocking tens of millions in ROI for the first time.

Agent ROI

Our $850,000 agents investment dramatically expands comp team coverage, and the incremental spend grows our annual comp team costs to $4.6 million, a 0.23% expense ratio of the $2 billion portfolio — still inexpensive compared to mutual fund fees.

Great, the math works. So what do we get in return?

Well… what do you get when a comp expert supports a pay decision?

The answer is smarter outcomes, which is to say, achievement of your compensation goals.

These goals vary over time and across the business. Sometimes we’re explicitly reducing cost. We are stewards of our company’s largest line-item investment, after all.

But usually cost reductions are balanced with increased competitiveness — spend less here so we can spend more there to get the best talent. Or we’re moving comp to manage the P&L, from cash to equity, or to manage performance, from fixed to at-risk.

Ultimately, the objective of a comp expert supporting pay decisions is to optimize your portfolio investment to achieve your goals.

The math

Let’s make a modest assumption that when a comp expert analyzes a comp decision, they make it 3% smarter. It’s not that much, on the margin:

• A salary of $200,000 moves $6,000

• A sign-on bonus of $30,000 moves $900

• A starting hourly wage of $23.80 moves $0.72

In our example company with a $2 billion portfolio and $950 million in turnover, the comp team can generate $29 million in value each year. That’s a lot.

Our new agents team actively manages 48% of portfolio turnover value, or $456 million, for the first time. 3% smarter is $13.7 million in value, and the agents cost $850k.

Our agents investment yields an ROI of 16.1x.

Share

But every comp team needs to run its own math based on its own goals and assumptions.

• How much of that value is genuine, make-your-CFO-happy reduction in cost, versus “smarter” fluff?

• Can you really address 20,000 pay decisions with agents, or do you start with a pilot of one agent to take on 3,000?

• Can agents make decisions only 3% smarter? Can some areas like stock comp get 15% smarter?

This leads us to the natural question all of us should be asking — is this too good to be true?

AI in comp is inevitable… if it works

History doesn’t repeat itself, but it often rhymes.

— Mark Twain

If agentic AI works, the unit economics are too good not to adopt agents.

Every comp team with a portfolio of more than, say, $100 million, can achieve a massive efficiency gain in how agents support smarter pay decisions.

But will agentic AI work?

While undoubtedly some amount of the promise of agents is overhyped, we should have every reason to believe the technology will rapidly progress — look how much it’s already changed in the two years since ChatGPT arrived.

And companies like Compa are speeding through learning cycles to make agents smarter comp analysts that understand more and more scenarios.

Betting against AI feels like betting against the early digital cameras with grainy, pixelated images. Or against the first iPhone with no app store.

It’s a failure of imagination, both for anticipating the inevitable, but also for getting excited about what it makes possible in compensation for the first time.

Peer Group is a newsletter for comp leaders navigating competitive talent markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

We announced a new product last week: Compa Frontline, a new way to benchmark hourly worker pay with real-time data down to the zip code.

I rarely paid attention to hourly worker (“frontline”) issues when I was in comp — it wasn’t very relevant in tech, and seemed uninteresting.

I had no idea.

Over the past year I have learned how overwhelmingly big the frontline comp problem is, driven by three fundamental challenges:

1. City block-level competition

2. Highly-fungible talent

3. Rapid market changes

25 cents becomes $50 million

Frontline workers are typically paid less than knowledge workers, but the companies are bigger by headcount, and there are way more people scattered across the world in storefronts, distribution centers, field sites, and manufacturing facilities.

As a result, seemingly small comp decisions drive millions in spend — it’s a different type of complexity.

Consider a 10,000-person tech company versus a 100,000-person retailer.

• If tech market salaries grow $1,000 per person, your annual costs increase by $10 million

• If frontline worker wages grow $0.25/hour per person, your annual costs increase by $52 million

Getting compensation correct down to 25 cents, distributed across 100,000 people, is the impossible frontline comp challenge.

And it starts with nailing location.

Share

Location, location, location

Location drives everything in frontline talent markets.

• In Las Vegas, Nevada, a coffee shop on the Strip pays far more than a coffee shop less than one mile away

• In Bozeman, Montana, the spiking real estate market is fueling a spiral in frontline worker shortages and rapid pay increases as new retailers scramble to establish in the area. Meanwhile, 20 miles away, nothing has changed

In knowledge worker comp we think about talent markets as broad metropolitan areas, like the SF Bay Area, and might have 3 or 4 geo tiers in the U.S.

In frontline comp, talent markets can vary down to the zip code, if not city block, and a company might define 200+ distinct markets in the U.S. alone.

Highly-fungible talent

The frontline workforce is highly fungible, meaning:

1. The talent market is massive with low switching costs

2. Job matching must flex between specific and broad

Unlike in tech where a small number of people know how to, say, engineer a fiberoptic network, low-skilled workers can easily switch between different roles and industries. The same person could be a cashier, a barista, or a concierge — all in the same year.

This means your peer group is as much a function of location as it is the company’s specific domain, and it constantly changes.

Even when you get your competition defined, job matching is still a nightmare.

Consider a real example from two nearly identical retailers:

1. The first company differentiates pay between cashiers who scan items over a counter and cashiers who scan items with a scan gun

2. The second company hires for a single job where the same person checks out customers, sells electronics, and unloads the truck in the back

How could a survey possibly match jobs accurately between these two companies?

Rapid change: competition, legislation, and seasonality

Talent markets change constantly and non-uniformly.

I recently asked the head of comp at a major retailer who their compensation peer group is. She said:

Step outside one of our stores and look across the parking lot at every other store you see in the strip mall — that’s our peer group.

When a new retailer moves in, or a new distribution center opens, or when a manufacturing plant closes, that talent market changes over night, while others simultaneously move the opposite direction.

Add to that minimum wage laws and pay transparency laws, which have exploded across the world in the past few years, varying by state, county, and city, and change year-to-year.

And of course seasonality drives changes in pay, whether it’s the holiday rush in retail or the summertime vacationers in hospitality.

A new way to benchmark frontline comp

I became excited about the frontline comp problem about a year ago.

One of our customers who spans the tech and retail worlds began partnering with us on the tech side, but urged us to learn more about the retail side and its accompanying comp challenges.

As we dug in, I observed that:

(a) traditional surveys terribly underserve this market, and

(b) real-time market data has the potential to completely transform frontline comp strategy with better data.

So we set about developing a new benchmarking solution, and recruited a group of visionary comp teams at the world’s best retailers to join us as launch partners.

The result is our new product, Compa Frontline, and it’s now live with many of the world’s best retailers.

It aims to solve each of the three big challenges:

1. A map-based benchmarking experience with zip code-level market insights

2. Flexible matching based on job and level, but also tenure and skills to solve the coordination problem

3. Software-delivered, validated, real-time data powered in a give-to-get model by the network of participating companies

Share

Where this goes next

Last week’s launch of Compa Frontline represents a powerful new way for enterprise comp teams to compete in the massive, volatile frontline markets.

We are initially focused on serving the retail industry, alongside tech and life sciences companies with frontline worker populations, and will add more soon.

We will also continue expanding the platform capabilities as more companies enter the network, tackling complicated and unique pay elements like shift differentials and bringing in more market context.

Want to see it live? DM me on LinkedIn and I’ll show you how it works.

Our mission at Compa is to make compensation fair and competitive for everyone by powering smarter pay decisions. We’re now applying that mission to the hardest problem in comp: the frontline.

Peer Group is a newsletter for comp leaders navigating competitive talent markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

A key insight from this analysis: the turnover of compensation investment is high and requires active management, like a mutual fund.

A mutual fund has more expensive management fees, or expense ratio, because it trades more actively.

So you might expect the comp team to be sized similarly to a mutual fund as a percentage of its portfolio.

How big should your comp team be?

That’s exactly what we dig into in this post, the second in a three-part series.

And no, we’re not using the usual headcount math; this time, we’re using money.

Comp teams are like mutual funds

We estimated that a 10,000 person tech company manages a $2 billion compensation portfolio with ~$950 million in annual turnover, or 47.5%.

Turnover is a useful metric to understand how actively the investment team manages a portfolio, i.e., how expensive is it to manage the money.

Index funds are passively managed and turnover, or trade, very little each year, 2-6%.

Mutual funds are actively managed and turnover a much larger percentage of the fund each year, 40-60% — similar to compensation teams.

Turnover correlates with management fees, or expense ratio, expressed as a percentage of investments managed:

Index funds charge very little, ~0.02%-0.10% expense ratios, whereas mutual funds charge ten times more, ~0.70%-0.90%.

And this makes sense: the more activity — buying and selling — the more people and cost needed to manage those transactions.

Comp very closely resembles a mutual fund profile in turnover.

What about management fees?

That is, do we size our compensation teams similarly to a mutual fund, 0.70%-0.90% expense ratio?

Share

Sizing your comp team, “expense ratio”

Comp team “expense ratio” consists of headcount and discretionary spend. Let’s size it.

Our 10,000 person tech company with a $2 billion compensation portfolio and 47.5% turnover has the following 10-person comp team:

• VP Total Rewards

• Exec Comp Director

• 4 Comp Partners (R&D, GTM, G&A, International)

• Comp Programs Manager

• 3 pooled Analysts

Using the market 50th percentile total direct comp using data from Compa, this team costs ~$3.2 million.

Adding $500k discretionary budget for market data, software, and other expenses like consulting projects and team travel, the total cost is $3.7 million, or 0.19% of the $2 billion compensation portfolio.

This is a hypothetical model for our tech company. Looking at an example set of 10 similar tech companies in Compa ranging 5,000-10,000 employees, we see a similar estimate, 0.14%-0.29% expense ratio.

But with the mutual fund comparison, we expected expenses closer to 0.70%-0.90%, or $14-$18 million each year. Instead we’re seeing a third of that.

Why are we spending so much less than expected?

…And should your comp team be three times bigger?

Comp triage

I think the answer is coverage.

Realistically, most teams actively manage a small (and noisy) fraction of all pay decisions across all comp programs — exceptions and outliers and exec comp — and leave the rest to discretion by recruiters and managers within a broad set of policies.

If we estimate comp teams actively manage only 20% of all pay decisions annually, with the balance passively managed via decision-makers out in the business, we can weight the predicted “management fees” to find the mutual fund comparison is very close to our 0.19% estimate.

20% active management:  0.80% mutual fund fees
80% passive management: 0.05% index fund fees

(20% * 0.80%) + (80% * 0.05%) = 0.20% comp fees

Cost per pay decision

So why not have a much bigger comp team?

Like a mutual fund aims for outsized returns with more active management, should comp do the same?

The underlying problem is that, unlike mutual funds, our compensation portfolios are fragmented into thousands of decisions, and humans are too expensive to manage all of them.

Looking back at our comp programs breakdown from my first post, we can add up the number of decisions involved for each program’s turnover based on the employee population.

Comp teams will triage these 27,000 decisions on a power curve, actively managing the most valuable decisions and passively managing the rest. If the top 20% of decisions are worth 50% of all turnover, we can calculate the per decision cost and value to reveal our unit economics problem.

(Buckle in for a some economics math, it’s worth it.)

Starting with the decision value: our top 5,000 decisions are worth way more than the remaining 22,000:

What about the per-decision cost?

If our 10-person comp team is actively managing only the top 5,000 pay decisions, then each person has capacity for 500 pay decisions. Assuming a TDC cost of $250k/comp person, each actively-managed comp decision costs $500.

Now the problem is emerging — comp people only have so much time in the day, and they cost a lot. The unit economics look good for the most expensive decisions, but look terrible for the rest.

The “management fees” for the most expensive comp decisions are rational, 0.53%, similar to a mutual fund.

But for everything else, it’s cost-prohibitive.

Actively managing the remaining 22,000 decisions would cost 2.32%, off the charts from our mutual fund comparison.

Your comp team size is a function of the value of your portfolio and the quantity and value spread of its underlying decisions.

Share

The $1 billion dilemma

Our 10,000-person tech company actively manages only half its portfolio value, leaving $1 billion to passive management — discretion within broad ranges, managers and recruiters with little training and context, high-level philosophies.

Your instincts might say, “let’s add one more comp person, if they optimized even 1% of that $1 billion, they could create $10 million in value.”

But herein lies the dilemma.

There are too many thousands of decisions to wrangle — one more person tackling 500 decisions worth $22k each can optimize only $11 million in value. You’d need 40 more comp people costing $10 million to get to every remaining decision, erasing all of your ROI.

AI breaks the impasse

The bottleneck is human cost and capacity to manage thousands more decisions. So we keep our comp teams small and only manage the hardest, most valuable decisions.

But AI agents introduce something new: the potential to manage thousands of decisions simultaneously at a fraction of the cost.

With AI, a comp team can radically expand its reach to every pay decision in the company with rational unit costs, unlocking tens of millions in ROI for the first time.

This, in my view, makes AI transformation inevitable.

The economics of AI will change compensation management forever.

And this will be the topic of my next and final post in this series.

Peer Group is a newsletter for comp leaders navigating competitive talent markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

Compensation teams are like Wall Street firms managing billions in assets.

It’s not just a metaphor.

The best teams really do consider how human capital cost hits the P&L, and search for new ways to identify market changes and take advantage to help their companies win.

It’s a useful framework understand the economics of compensation management:

1. The size and shape of annual comp spend, your “portfolio” and the decisions to make, your “turnover”

2. Rationally sizing the comp team, your “management expenses”

3. The resulting economically inevitable adoption of AI to optimize thousands of decisions

This is the first post in a three-part series, beginning with #1: defining the size and turnover of your compensation portfolio.

And I do love a good extended metaphor — if comp teams are like Wall Street investment firms, then yes, using old data from surveys is like making trades using stock prices printed in a newspaper 12 months ago. 😉

Sizing our comp portfolio

A tech company has 10,000 employees making an average of $200,000 in total direct compensation.

This company has a $2 billion compensation portfolio.

10,000 * $200,000 = $2,000,000,000 

How is this $2 billion portfolio invested each year?

This tech company invests its portfolio across four compensation programs:

1. Base salary

2. Commissions

3. Annual bonus

4. Stock comp

Base salary

All 10,000 employees receive a base salary, and the global average is $130,000, for a total base salary investment of $1.3 billion.

10,000 * $130,000 = $1,300,000,000

Commissions

Cash incentives are split between commissions (on-target incentives, “OTI”) and an annual bonus program.

30% of the workforce, or 3,000 employees, are commissioned with an average OTI of $60,000, for a total OTI investment of $180 million.

10,000 * 30% * $60,000 = $180,000,000

Annual bonus

Of the remaining employees, about 70%, or 4,850, are eligible for the annual bonus program with an average bonus of $20,000 (~15% of base salary), for a total bonus investment of $97 million.

7,000 * ~69.3% * $20,000 = $97,000,000

Stock comp

100% of employees are eligible for stock compensation, and 94% of employees receive grants each year worth an average of $45,000, for a total of $423 million. (Note we’ll break this down between new hire and refresh grants later.)

10,000 * 94% * $45,000 = $423,000,000

Total compensation

Our tech company’s total $2 billion compensation portfolio is comprised of:

Share

Calculating comp turnover

A $2 billion fund on Wall Street does not turn over its entire investment each year.

Rather, it invests a portion of the fund annually, using reserves or selling some investments to buy new ones.

Similarly, in compensation, we turn over a portion of our investment each year in the form of making pay decisions spread across each of our programs for new hire offers and current employees.

Let’s define these decisions as our annual compensation turnover.

The size of this turnover matters because it helps us estimate how big our compensation team should be.

But first, let’s calculate this tech company’s annual compensation turnover.

Offers

When looking at turnover, it’s useful to distinguish between offers and employees.

When we make offers, we turnover, or decide, 100% of the compensation package. For employee decisions, it varies based on the program.

Our tech company has a 15% hire rate, combining attrition backfills and growth, or 1,500 accepted offers with an average $200,000 total compensation package for a total of $300 million.

Base  1,500 * 100% * $130,000 = $195,000,000
OTI   1,500 *  30% *  $60,000 =  $27,000,000
Bonus 1,500 * ~49% *  $20,000 =  $14,550,000
Stock 1,500 *  94% *  $45,000 =  $63,450,000

Total                         = $300,000,000

(For simplicity, we’re assuming the same % receiving and $ averages as current employees, although this varies for most companies.)

Employees

For the remaining 8,500 employees, comp turnover varies by program:

• Base salaries have a 5% pool for merit and promotions each year

• OTI and bonuses are at-risk and discretionary each year, so 100% of investment turns over annually

• Stock refresh grants are also at-risk and discretionary with 100% turnover annually

Putting this together, our employee turnover is ~$650 million annually.

Base   8,500 * 100% * $130,000 *   5% =  $55,250,000
OTI    8,500 *  30% *  $60,000 * 100% = $153,000,000
Bonus  8,500 * ~49% *  $20,000 * 100% =  $82,450,000
Stock  8,500 *  94% *  $45,000 * 100% = $359,550,000

Total                                 = $650,250,000

Total comp turnover

Our tech company with a $2 billion compensation portfolio makes $300 million in offer decisions and ~$650 million in employee decisions annually, for a total annual compensation turnover of ~$950 million.

This turnover is spread across each comp program:

The resulting annual compensation turnover is 47.5%.

Notice where the turnover is concentrated (and where it’s not):

• Incentive pay is 74% of turnover at $700 million

• Offers total 32% of turnover at $300 million

• Stock comp is the most expensive program measured by turnover at $423 million

• Merit cycles account for $55 million, only 6% of turnover

Share

Why it matters

Two major insights:

1. The size of compensation investment, $ billions, is non-trivial to a company’s financial performance

2. The turnover of compensation investment is high and requires active management

Point #2 is essential to understanding the economics of compensation management.

Comparing to Wall Street, the compensation team is more like an actively-managed mutual fund than a passively-managed index fund.

This has profound implications for determining the right size of our compensation team.

And that is the next topic we’ll explore in part two of this series later this month.

Peer Group is a newsletter for comp leaders navigating competitive talent markets and AI transformation! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

Today I am proud to announce Compa Agents, building on our sneak preview at Compalluminati 2025, and marking the beginning of the AI era for compensation.

Compa is launching two agents:

1. Analyst AI — automated compensation market intelligence

2. Partner AI — comp-controlled agent for recruiters to make offers

In this post I will share my vision for the future of compensation, our AI strategy at Compa, and how Compa’s first agents work.

Over the next few months, I will write more posts diving deeper into the impact of AI on compensation, including the economic potential, impact on compensation teams, and the potential pitfalls our industry will confront with this disruptive new technology.

Share

How AI changes compensation

Every compensation team faces an impossible task — how do we manage our company’s largest investment, balancing cost, competitiveness, and fairness?

Our goal is to design an integrated set of compensation principles, policies, workflows, and data — in aggregate, our compensation strategy — to cost-effectively compete in talent markets.

Two big problems

Designing comp strategy is hard, but implementing it at scale is harder, for two major reasons:

1. Compensation expense is fragmented, not monolithic — billions in investments each year are made across thousands of decisions

2. Market conditions change constantly and non-uniformly — in the same short timeframe we may observe spiking engineering salaries in India, layoffs in G&A management, and emerging skill premiums in quantum computing

Our compensation strategy, implemented to scale, is executed by hundreds or thousands managers and recruiters who often don’t understand why or how it works.

And conditions on the ground across dozens of talent markets change faster than we can detect them and respond.

As a result, we risk overspending (and underspending) our companies’ biggest investment each year.

And so of course, every small-but-mighty compensation team is on the hunt for greater precision and controlled outcomes.

Agents 10x comp productivity

AI technology, specifically agentic AI, is poised to transform compensation because it scales intelligence to thousands of decisions in dynamic market conditions.

Agents can perceive data and learn, reason based on rules and goals, make decisions, and even act autonomously.

A 10-person compensation team cannot advise on 40,000 pay decisions each year. Instead, the team triages, focusing human review on maybe 10% (4,000) of those decisions: a small number of exceptions and big-ticket packages and relying on broad policies, discretion, and reactive data to manage the rest.

With AI, however, that same 10-person team can develop agents to guide every decision, applying not just its compensation strategy, but also real-time market data, changing business objectives, and feedback learning from the agentic interaction itself.

Agentic AI changes compensation because it unlocks precision and control at scale.

And its adoption is inevitable because the numbers are so big:

A compensation team managing $1 billion in investment each year that optimizes spend by 5% produces $50 million in value.

Share

Compa’s AI strategy

Compa is a compensation intelligence company.

What does this mean?

We can define intelligence generally, and understand our innovation roadmap, by studying the anatomy of a decision.

The OODA Loop

Last year I read Boyd: The Fighter Pilot Who Changed the Art of War, the autobiography of John Boyd, a US Air Force pilot who developed the OODA Loop decision-making model, an acronym that stands for Observe, Orient, Decide, Act.

This concept has been successfully applied in military operations, most famously in Operation Desert Storm, and subsequently in corporate strategy, litigation, and myriad other areas.

The OODA Loop consists of four stages

• Observe: detect new information

• Orient: contextualize the new information

• Decide: choose a course of action

• Act: execute the decision, then re-enter the loop

Here’s a simple example in business, courtesy of asking ChatGPT:

A company sees a drop in customer satisfaction (observe), realizes it’s due to support delays (orient), decides to deploy a support chatbot (decide), implements the bot (act), and then re-evaluates based on new data (back to observe).

Our Roadmap

As a compensation intelligence company, Compa’s innovation roadmap is a progression through the anatomy of a decision.

• ✅ Observe: Real-time market data
  Compa started with real-time market data to provide visibility to new information. Instead of waiting 12 months to know what’s happening in the market, you can discover it now

• ✅ Orient: Comp analysis tools
  Compa’s analysis tools help contextualize real-time market information, including peer groups, geo tiers, unvested equity risk, trend analysis, and more

• 📣 Decide: Compa Agents (new)
  Compa Agents mark entry into the back-half of the OODA Loop, bringing automation and scaled decision-making to comp teams and their stakeholders

• 🤫 Act: Coming Soon
  Compa Agents can automate a limited set of high-impact workflows, with more coming soon

Compa is now the largest real-time market data network in the world, and growing faster than ever.

With Compa Agents, compensation teams can apply this intelligence to comp decisions at scale to achieve massive ROI.

Subscribe now

How Compa Agents work

Compa is launching two agents, Analyst AI and Partner AI.

These agents are designed to solve the two big problems with implementing compensation strategy at scale:

1. Analyst AI — market conditions change constantly and non-uniformly

2. Partner AI — compensation expense is fragmented, not monolithic

Here’s how they each work.

Analyst AI is for comp teams

Analyst AI is automated compensation intelligence for comp teams.

This agent accelerates the work of a comp analyst. Market research is fully automated by a configurable agent that monitors the market 24/7.

Analyst AI compresses days of manual digging and research into instant real-time analysis:

• Analyze retention risk for a key engineering team

• Recommend ranges for entering a new market

• Understand market shifts for hot jobs and skills

• Evaluate geo tiers based on recent market trends

This is a big shift to move from doing the analysis to directing the analysis and processing available decisions instantaneously.

Partner AI is for recruiters, controlled by comp

Partner AI is a comp-controlled agent for recruiters to make offers.

This agent is natively integrated with Compa’s real-time offers-based market data network, and comp teams can add any data, structured or unstructured, to provide more context and configure rules and access for guiding recruiters to better decisions.

Partner AI embeds comp expertise into each one of your thousands of offer decisions, enabling you to automatically guide hundreds of recruiters toward controlled outcomes with measurable ROI. Examples:

• Limit sign-on bonus usage to targeted scenarios

• Guide higher offers for hot skills

• Navigate offer negotiations accurately

• Review exception requests automatically

• Remind recruiters of key talking points on your comp programs

Fascinatingly, recruiter interaction with the Partner AI agent becomes its own source of feedback data at scale, enabling you to analyze competing offers, detect emerging market trends, and pinpoint pressure on your ranges.

Share

Get involved

A special group of customers have entered our Design Partner Program and are currently onboarding Compa Agents and working directly with my product team to provide feedback and influence our roadmap.

If you are interested in getting involved, shoot me a note and let’s discuss availability in the next cohort of companies onboarding into Agents later this summer.

In the meantime, we sat down with our Design Partners to talk about how top comp leaders are thinking about AI Agents, including the value proposition, risks, and how configurability should work — you can read the report here.

One more thing

The AI era represents a bright future for compensation.

Fundamentally, we are small, highly-leveraged teams managing billions in investment. Better data and AI technology increase our productivity and impact on the company’s growth strategy, perhaps more than any other team in HR.

My co-founders and I have designed Compa as a compensation intelligence company aimed squarely at powering smarter compensation decisions.

We are committed to bringing responsible, high-impact agentic AI into the domain of enterprise compensation.

…And with that commitment, we made a small change on our website:

Peer Group is a newsletter for comp leaders navigating competitive talent markets! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

No, probably not.

Developing a smart and compelling offer involves building out four-year cash flow scenarios. It’s a fun model — shoutout Matt Olson for teaching me how to do this thirteen years ago.

But unless the candidate volunteers a detailed and accurate picture of that $1m walkaway value (you can’t ask!), you are stuck with using a lot of weak assumptions and spotty information… right?

It turns out market data in Compa can give you excellent information to:

• Validate what the candidate shared,

• estimate walkaway details, and

• develop a winning offer without overpaying.

Here’s how to do it.

Check market unvested equity values

First, the $1m unvested value looks steep:

This is the market total unvested value for M5 Software Engineering in Compa, calculated using current value from all grants for current employees based on grant-level market data from stock admin systems.

• Median is $715k

• 75th is $901k

• The candidate’s reported $1m is at the ~90th percentile

So we have validated that $1m walkaway looks high.

One possibility is the candidate is embellishing a tad, and based on my experience this is common. (“Well, if our stock keeps going up it will be $1m by the end of the year.”) Another way to think about it is it’s very unlikely a competitor would pay the full amount.

Share

Break down the walkaway value timing

How much of that $1m vests this year? In future years?

When you build your four-year cash flow model, you can “pay” for some of the unvested value through both your new hire grant and future refresh grants. But the candidate’s objection will be that she has to wait for the new grants to vest, leaving a gap this year.

So let’s go back to the market data in Compa and toggle from Total Unvested to Vesting next 12mo:

A much different picture emerges:

• Median is $353k

• 75th is $445k

• The candidate’s 90th percentile reported value is ~$500k

With this market information, we can infer that about 50% of walkaway value vests in the next 12 months.

Interestingly, we can substantiate this ratio as a good rule of thumb looking across other levels:

Also, this makes sense based on typical NH:Refresh grant ratios (2.5 for more senior M levels) and how grants stack up.

Take an employee with three years of tenure and annual refresh grants:

The higher new hire grant final vest props up the unvested value vesting over the next 12 months.

Subscribe now

How much does the market cover?

Now my favorite part — how much does the market cover walkway value in the first gap year?

Most companies use a 1-year cliff in their new hire grant schedules (although we’re seeing some companies ditch it in their refresh grants) to make sure the employee sticks around for the first year.

This creates a gap:

• Our candidate will vest ~$500k over the next 12 months

• Our stock comp pays nothing over the next 12 months

So this is where we use a sign-on bonus. Do we have to pay the whole $500k to get the offer accepted?

Again, we can turn to market data to estimate how much gap the market covers. And this is my favorite because it combines two types of market data in Compa, offers-based and grants-based, to compare sign-on bonuses and unvested equity.

Here’s what we find:

The market only covers 20% of the value vesting over the next 12 months.

That is, your candidate can expect to take an 80% hit on first year cash flow to take the new job; in our example, that’s $400k.

From the employer perspective, a $100,000 sign-on bonus covers $1,000,000 in total walkaway value, about half of which likely vests over the next 12 months.

Why does a candidate accept this? And how do you sell it?

Tell your story:

• Better base and bonus target

• Our stock price is growing and yours is declining

• Bigger role, better manager, exciting vision

• While the employer pays a sign-on bonus to unseat the candidate, the candidate pays a larger opportunity cost to join the new company

One more thing

Compalluminati last week was incredible!

300+ comp leaders from the world’s best companies gathered at DoorDash HQ in San Francisco to discuss the future of comp. Products were announced, Compallumilattes were served, and a sweet Compa truck took to the streets of SoMa:

Thank you to my team for organizing, DoorDash for hosting, and to everyone who joined us!

We love bringing this special community together.

Peer Group is a newsletter for comp leaders navigating competitive talent markets! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

I propose looking at a simple ratio to understand current market conditions:

Offer/Employee ratio = Median offer / median employee

You can predict market behavior with this ratio:

• When offers exceed employee pay (ratio > 1.0), the market heats up, and companies focus on competition. Offer acceptance rates fall, attrition rises, exceptions grow, and the employee-friendly market drives more employee concessions like remote work and meal perks.

• When offers equal employee pay (ratio ≈ 1.0), the market is stable, and companies focus on fairness. In this balanced market, compensation is more predictable, and pay equity is well-managed.

• When employee pay exceeds offers (ratio < 1.0), the market cools, and companies focus on cost. Offer acceptance rates rise, attrition falls, layoffs are more likely, and the employer-friendly market drives more employer concessions, like RTO.

And to put it in a table:

Why I think an index ratio metric like this matters:

• Share with my Comp Committee to set context for market conditions when we make big decisions

• Understand how my own offers and employees compare — am I carrying retention risk or layoff risk?

• Comp strategy is a three-legged stool between competitiveness, cost, and fairness — the ratio suggests a framework for when priorities might shift

Share

How to measure this

It is simple to back-of-envelope calculate this.

Let’s look at a couple examples, all P4 US roles, with median offer base salary over the last 6 months compared to median employee base salary today, using Compa:

AI/ML Engineering: $241,000 / $221,000 = 1.09
Product Management: $205,000 / $200,000 = 1.03
Legal Counsel: $213,000 / $220,000 = 0.97

Now to put real weight into this, we’ll want to sharpen pencils on methodology (and remove some noise).

• Compare across the same group of companies, and consider weighting: observations or companies

• Account for metropolitan areas

• Analyze across multiple levels, or weight job-level-metro offer/employee pairs by market volume

• Consider other pay elements beyond base salary, at least target total cash

• Analyze over time and consider seasonality

Another question: where is equilibrium? Is the equilibrium ratio 1.0, as I suggested, or something slightly higher than 1?

To the degree we recruit passive candidates well-employed for like-level roles, we expect to pay some non-zero switching cost, often (but not always) expressed in the form of a higher base salary.

What market are we in now?

Based on the trend data in Compa and my daily conversations with compensation leaders, I think we are in a mixed market.

The AI race is heating up offers for technical roles, but many of those same companies are hiring back from layoffs across the rest of the business at much lower costs.

But we’d have to put our Insights team to work on this to sharpen methodology and understand the market holistically. We certainly have an amazing dataset for it.

Do you think we should work on this?

Let me know your thoughts.

And excited to see you all at Compalluminati next week!

Peer Group is a newsletter for comp leaders navigating competitive talent markets! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

I’ve already been asked a dozen times since yesterday whether any comp teams are reacting to the tariff-induced stock market volatility.

This could be a defining moment for the talent market in 2025 and beyond, much like COVID in 2020 or interest rate hikes in 2022… or it could blow over.

To help you monitor market change, I asked my team to put together a flash poll. Participate by clicking the button below:

Participate in Flash Poll

We’ll keep the poll open through Friday and return results early next week. Results are available to poll participants only.

Flash poll questions cover:

• Equity targets

• Retention grants

• Benchmarking

• Employee communication

• Hiring plans

It should take about five minutes to complete.

What are the comp risks?

It’s too early to closely analyze this yet, and the poll results will be revealing whether teams are making moves.

With that said, a couple initial thoughts:

• If the stock market stays in correction territory, then the seasonal timing is acutely bad for employees at the many companies that made refresh grants in March (I actually wrote about this timing risk a few months ago)

• If we see pull backs on IPOs and M&A, this could create a liquidity crisis in tech, including employee access to secondary/public sales, acutely for late-stage privates, as well covered here, here, and here

• We could see big private tech companies gain a short-term edge over their public rivals in attracting AI talent. Well-funded and recently funded companies like OpenAI, Databricks, and Stripe can tell a longer-term, less market-sensitive value story, while their public counterparts face declining unvested equity values

• Companies with flexible comp programs may see employees flee to cash safety in their next election, driving up cash spend and reducing share burn, and some frank conversations with finance

• More generally, sliding towards recession could bring back difficult and all-too-recent memories from 2022/2023 — hiring freezes, layoffs, pay range reductions, and more. It’s way too early and fast-changing to know, let’s hope not

Share

Peer Group is a newsletter for comp leaders navigating competitive talent markets! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

I’m excited to announce Compalluminati 2025 is next month!

Compalluminati is our annual summit for compensation professionals to exchange ideas, build community, and learn from the very best minds in comp.

We started this event last year and the experience was incredible: 120+ people joined us (right in the middle of comp planning!) at two of our customers’ offices, Roblox and Stripe, to talk about the future of compensation.

This year Compalluminati is on April 23-24 (much better!) and co-hosted with DoorDash in their beautiful San Francisco offices.

Register Now

What you can expect

Compalluminati is exclusively in-person and features two days of immersive discussions, industry talks, and panels led by comp leaders from Cloudflare, Databricks, DoorDash, Instacart, Micron, Shopify, T-Mobile, Walmart, Workday, and more top companies.

Our theme is skills-based comp: promise and skepticism. After being a skills skeptic for many years, I led Compa in launching a skills-based market data product last year. Since then we have observed rapid adoption of skills data for previously impossible levels of benchmarking precision, especially for hot tech jobs. Together we will dig into how (and how not) to leverage skills data in navigating volatile markets.

Oh, and we will announce new Compa products live on stage!

Share

Why community matters

Compalluminati is all about building community.

I think community matters in our world for two big reasons:

1. Compa is a compensation intelligence company — fundamentally, we aim to answer the question of “what’s happening in the market right now?” in part through our data products, and the rest through community.

2. The world of compensation is rapidly changing. Over the past few years, many exogenous forces have converged to reshape how we practice comp — AI, pay transparency, remote work, and Wall Street pressure, to name a few. Now more than ever, we must learn and adapt, or risk falling behind.

How to join

It’s easy:

1. Register here to save your spot (it’s free)

2. Join us in-person at DoorDash in downtown San Francisco 🌁 on Wednesday, April 23, and Thursday, April 24 (event details here)

3. Tell your comp friends you’re going to Compalluminati!

I’m excited to see you all next month!

Peer Group is a newsletter for comp leaders navigating competitive talent markets! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

Sales professionals at big tech companies get RSUs with their offers just like engineers — align with shareholder interests and incentivize long-term goals, right?

Especially for more senior roles, and especially in tech hubs like SF.

…right?

In February 2025, prevalence of new hire grants for Sales & Customer roles reached an all-time low of only 56% of offers.

This is filtered to US offers by public companies for senior IC levels (P4-P6) in SF, New York, and Seattle from 2022 to ending last month, comparing Technical (blue), Corporate Functions (pink), and Sales & Customer (green) domains.

GTM new hire grant prevalence has always lagged more technical roles, but beginning in early 2024, it dropped way lower.

Share

What is changing with sales comp?

I have a decent guess that I can back up with some data.

But what do you think?

We can look at the offer trend data a couple different ways to learn more:

Over the same period with the same filters applied, median sales target total cash (i.e., OTE) value is up 35%. By comparison, corporate function roles are up 14% and technical roles are up 12%.

And then if we go back to new hire grant prevalence and focus on sales roles comparing Tier 1 (SF, NYC, Seattle) with the rest of US metros:

We see that new hire grant prevalence between tech hubs and other US cities has converged.

My guess what’s happening:

1. Tech companies are reducing stock comp participation, among several levers they must pull to meet new SBC expense targets

2. Comp teams are trading GTM stock comp for increased cash OTE — this reduces stock comp spend in exchange for short-term cash spend

3. Critically, the increase in OTE is focused on commissions/OTI — that is, companies are growing their cash expenses only if they hit sales targets (which I suspect are rising, too)

Meanwhile, sales OTE rises dramatically

On point #3, fascinatingly, you can actually see it in the trend data, focused here on P4 to minimize noise:

Median base salary up 10%, target total cash / OTE up 36%.

Comp teams, this looks like a smart move to make, especially because offer accept rates haven’t dropped at all over the last year.

Go talk to your sales comp friends to see if you can Indiana Jones some stock comp for OTI. 🤠

And you may find that your best salespeople are happy to get bigger quarterly checks when they close big deals.

Peer Group is a newsletter for comp leaders navigating competitive talent markets! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

What do these roles have in common?

• both are Senior level

• both are based in Sunnyvale or San Francisco

• and both jobs are posted with the same salary range of $198,000-$220,000:

However, one job is a SWE on the Ads team and the other is an MLE on the Rider Experience team.

Why do they have the same range?

Two possible reasons:

1. Disclosure strategy — sometimes companies intentionally post the min and max of multiple job families (and levels and locations) to disclose a wider range, providing more discretion on how they transact individual offers. But given the tight 11% spread, this seems less likely for these roles

2. Job architecture — many large tech companies (about half, according to Compa research), do not distinguish AI roles from SWE roles in their job architecture, either as a philosophical choice or due to lack of available reliable data for emerging jobs. As a result, ranges for distinctly different job postings drive off of the same job code. Meanwhile, how recruiters use these ranges vary when they encounter real market differences

So, should we expect the market to command different rates for these two roles, or the same, as the posted range might suggest?

Let’s analyze this in two steps — first using job titles, then using skills.

Share

Analysis using job titles — SWE vs AIE

I looked in Compa at P3 offers by 75 public tech companies in the SF Bay Area over the last 12 months to analyze current market rates.

Methodology note: “Senior” roles could be either P3 or P4, it depends on company philosophy — I can’t tell based on a public job posting. Also, I might want to build a tighter peer group than 75 public tech companies, again depending on company philosophy.

Here’s how Uber’s range compares to the market P25-P75th for software engineers and AI / ML engineers:

Benchmarked against Software Engineering, the lower end of the posted Uber range is approximately the 50th-75th percentile, with the rest of the range comfortably above it.

But benchmarked against AI / ML Engineering, the range approximates the 75th percentile.

So if the recruiter defaults to the midpoint of Uber’s range, $209,000, they will pay the Ads SWE well above the 75th percentile and the ML Eng closer to the 60th percentile.

This might produce an unhappy result, especially if hiring in AI is a priority.

But taking this a step further by analyzing the market with skills data reveals a bigger problem.

Subscribe now

Analysis using skills — Node.js vs Pytorch

The job titles reveal differences in these two roles potentially hidden from the comp team by a shared job code. The skills in the job descriptions offer more clues:

The Ads SWE role includes skills like Golang/Go, Java, and Node.js, while the ML Eng role requires ML framework skills in Tensorflow, Pytorch, or JAX. And some skills are overlapped, like Python.

These skills highlight further differences between the roles, and with Compa we can use skills to get more precise market measures.

I did this by building a “custom job,” where instead of relying on Compa’s job match architecture, I composed a set of skills filters based on each job description:

This approach reveals even wider market differences:

Using our custom jobs, the market range for the Ads SWE is $163k-$193k and for the MLE, Rider Experience is $192k-$230k.

This result shows completely different markets for two roles with the same base salary range:

• The MLE, Rider Experience median is $30,000 higher than the Ads SWE

• The Uber range midpoint of $209,000 is at the 44th percentile for the MLE, Rider Experience and above the 95th percentile for the Ads SWE

And stock comp (not pictured) shows even starker differences: total new hire grant median value is $137k higher for the MLE, Rider Experience.

Share

Skills-based comp: promise and skepticism

This analysis uses a simple public job posting to illustrate the precision now available with offers-based market data, supercharged with skills, as an antidote to more traditional blunt instruments.

I had long been a skills skeptic, which I wrote about in-depth here. What changed?

New data (offers!), technology, and methodology make it possible to accurately see the market through skills, establishing a new level of precision comp teams can use to compete in challenging talent markets.

How do our conversations with the business change with skills comp data? Do we go so far as to change our job architectures?

We’re going to discuss the promise (and skepticism) of skills-based comp at Compalluminati 2025 — early invites are out, and I’ll share public details in my newsletter next week.

Peer Group is a newsletter for comp leaders navigating competitive talent markets! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now

This chart shows base salary trends for a P3 Software Engineer, a high volume role with thousands of offers in Compa, from 2022 to 2024 in:

• SF Bay Area

• All California excluding the SF Bay Area (“Other California”)

• A basket of typical Tier 2 cities for reference: Austin, Chicago, DC Area, Denver, Portland, and Salt Lake City

Check out the slopes: while SF salaries rose 4.3%, Other California rose a staggering 10.7% over the three year period. Meanwhile, Tier 2 cities barely increase 1.7%. Note you can stress test the generality of this finding by looking at different percentiles, levels, roles, and pay elements.

Expressing this shift in geo tier terms shows Other California as a percent of the SF Bay Area exiting 2024 at 100%.

Share

You can see a similar trend in new hire grants:

Other California new hire grants caught up to the SF Bay Area in 2024, departing from Tier 2 cities that continued to languish in 2024.

What’s going on?

I moved to Newport Beach in 2020 and expected the cost of living to be lower — but my hipster-certified, single-origin drip coffee (favorite spot) costs minimum $5, and easily $7 for pour-over, just like San Francisco. ☕

As people moved from SF to places like Los Angeles, San Diego, and Orange County, they brought increased demand for expensive lifestyles, fueled by their SF-level wages.

Tech companies no longer offer less to technical talent in Los Angeles than San Francisco; pay has equalized.

If you still have Los Angeles and rest of California in Tier 2, you’re probably observing higher exception rate and compa-ratios.

Analyze this yourself in Compa

I built this analysis in Compa using the new geo tiers feature we launched last week.

You can group cities into tiers as you’d expect, but the best part is you can mix cities with states and create exclusions, like building a tier that includes all of California except the SF Bay Area.

Check it out — it’s a useful feature to pressure test changes to your geo tiers, and discover when a talent market shifts.

Not a Compa customer? Get started here.

Peer Group is a newsletter for comp leaders navigating competitive talent markets! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.

Want more polls and insights about the latest thinking in comp?

Subscribe by hitting the button below.

Subscribe now