The AI layoffs are real. They just won't happen where you think.

Your firm's layoff risk comes from whether you keep winning work, not from the AI tools you use. Through June 2026, U.S. companies attributed 101,743 job cuts to AI, about 23% of all cuts this year and 85% more than all of 2025. AI has led the list of layoff reasons for four months straight, according to Challenger, Gray & Christmas.

If you run a service firm, those numbers look like a countdown. The assumed chain is simple: adopt AI, decide you need fewer people, and start cutting. Most owners I talk to assume that math.

Firm-by-firm data tells a different story. Companies that invest in AI grow revenue and their employee count at the same time, while firms that skip AI lose more employees than firms transformed by it. Cuts cluster at firms that replace people with AI across the board, and at firms losing work to faster competitors.

What layoff trackers leave out

That assumption shows up before any cut happens: a budgeted hire gets deferred one more quarter. A senior person leaves and the replacement never gets posted. The team notices the freeze and draws its own conclusion, and the two people you most wanted to keep start taking recruiter calls. None of that reaches a layoff tracker.

Owners who read the trackers as a countdown are responding to the way the data is published. There is nothing cynical in that assumption: the only public data connecting AI to jobs is a layoff count, so a layoff is what it predicts.

The tracker numbers are real, but they mix together three different stories. AI may have replaced the work, the firm may have been losing already and used AI as a respectable explanation, or the cuts may have been coming anyway, with AI as the perfect excuse. The trackers count all three the same.

Some losses are exactly what they look like: Stanford's Digital Economy Lab found that workers aged 22 to 25 in jobs where AI can do the most work saw a 16% employment decline relative to older colleagues at the same firms. The drop appears where AI automates work, but stays muted where AI helps workers. Entry-level roles built mostly from tasks a model can do are getting squeezed. That part of the fear is earned.

For the owner of a firm between $1M and $20M, the total number of AI-linked cuts answers the wrong question. The useful question is which firms are cutting. When you separate them, the pattern flips.

Most firms that adopt AI hire more people, not fewer

A study in the Journal of Financial Economics tracked companies investing in AI from 2010 to 2018. They grew faster in sales, employment, and market value than comparable firms. Most of that growth came from making more new products. They found more work to do, so they needed more hands.

PwC's 2026 AI Jobs Barometer found the same pattern in current data: employee counts at companies where AI can do the most work are growing faster than at companies where it can do the least, and productivity growth at those companies is 40% higher. PwC's conclusion is that AI may create more jobs when firms use it to open new markets. Some of that hiring comes from money pouring into companies built around AI, I give you that. Keep the conclusion in view a little longer.

The most useful detail for a smaller service firm sits in a 2026 Atlanta Fed survey of nearly 750 CFOs: large companies expect to shed workers because of AI. Smaller firms anticipate modest employment growth. The same paper finds that productivity gains come with innovation and demand, not cost reduction alone. A U.S. Chamber of Commerce survey points the same way: 82% of small businesses using AI increased their workforce over the past year.

The obvious question is this: If AI automates work, how does a firm that adopts it end up needing more people?

Cheaper delivery grows the work

Banks answered that question when ATMs spread through the 1990s: the machines automated the core task of a bank teller, and everyone predicted the job would disappear. Instead, teller jobs did not decrease. Between 1988 and 2004, the number of tellers needed to run an average urban branch fell from 20 to 13. Cheaper branches let banks open 43% more of them in urban areas, and the tellers who remained shifted to relationship work, the part machines couldn't do.

The same pattern is running now. Torsten Slok, chief economist at Apollo, calls it the Jevons employment effect: when professional work gets cheaper, more customers can afford it, so the market expands and the field grows to include more firms and workers. New business formation in the U.S. is running at the highest weekly levels in its history. Slok reads that as likely driven by AI.

For a service firm, "more branches" means more clients served, offers you couldn't price before, or turnaround times that win deals. Cheaper delivery creates growth only if you use the hours you free up to win more work. I wrote about this in the extra time AI buys you is already gone: saved hours evaporate into slack unless you decide in advance where they go.

Some famous adopters treated the capacity AI freed as a reason to cut people, and it went badly.

What happens when a firm replaces people with AI

In February 2024, Klarna announced its AI assistant was doing the equivalent work of 700 full-time customer service agents. About fourteen months later, CEO Sebastian Siemiatkowski reversed course and started recruiting human agents again. His explanation: "As cost unfortunately seems to have been a too predominant evaluation factor when organizing this, what you end up having is lower quality."

IBM took the same approach with a different ending. Its AskHR system absorbed a couple hundred HR roles by automating 94% of simple, routine HR tasks like vacation requests and pay statements. Then IBM's total headcount went up. CEO Arvind Krishna: "Our total employment has actually gone up, because what [AI] does is it gives you more investment to put into other areas." Those areas were software engineering, marketing, and sales.

The same tool led to two different choices:

	Klarna	IBM
What AI absorbed	The equivalent work of 700 full-time support agents	94% of simple, routine HR tasks
Where the savings went	Headcount and cost	Budget for engineering, marketing, and sales
What happened next	Quality fell; recruiting human agents again 14 months later	Total employment went up

The difference matters more at your size than at theirs because in a service firm, quality is the product. A client pays for your judgment, delivered on time, at a standard they can't buy cheaper elsewhere. AI replacing that judgment weakens the thing the client pays for. Using AI to sell and deliver more gives the firm the growth pattern in the data.

So where do the layoffs come from?

How firms that fall behind lose their people

I've watched hundreds of service firms adopt AI through a franchise network's delivery data. In the first half of 2026, more than one client told me they were gaining market share. When I probed, the trail led to the same place each time: AI tools, redesigned processes, faster delivery, a clearer view of their own data. The math checked out: better outcomes, better Google reviews, more referrals, more business.

Notice what's missing from that chain. Nobody announced a switch. No client of a slower competitor said "we're leaving for the firm built around AI." The shift showed up in review scores and referrals moving toward the faster firm, one client decision at a time.

The other side of that drift is audible if you listen for it. In January 2024, the owner of a content development agency he had run since 2015 posted this: "Slowly, but steadily, we are losing clients to ai. We do produce strategy etc, but all writing jobs are going to the bots. I have had to lay off writers." His layoffs were caused by AI, and none of the work was automated inside his firm.

The pattern predates AI. In a 2016 study of firm data through 2014, the OECD called the most productive companies frontier firms and the least productive companies laggards. It found that frontier firms gained significant market share relative to laggards, with productivity diverging between the two groups and many available technologies "remaining unexploited by a non-trivial share of firms." AI speeds up that split.

This is where the tracker numbers come from: a laggard's layoff arrives late and looks ordinary. Revenue softens, the pipeline thins, and eighteen months after a competitor rebuilt their delivery, an owner lets two people go. AI never appears in that layoff memo. It does not have to.

Let the work you win set your staff size

Your staff size follows how much work you win. "Can AI replace someone on my team" is the wrong question. Ask whether you will win enough work to keep everyone busy next year. Firms that answer it well end up at Stage 3 on the Delivery Model Ladder, where revenue climbs without adding people at the same pace.

You change that outcome in an unglamorous way: apply AI to one part of the work, then save time there or improve the quality of what comes out. Decide ahead of time what those hours buy: the next process, training, or better deliverables. Pick one more part and repeat. AI forces you to look at how you do business, and that loop is the whole method.

Skipping coordination is the expensive version. Roll AI out with a company-wide order and you create shadow AI use you can't see, plus anxiety in the people who assume the tool is their replacement. Either way the firm loses faith in the technology before it pays.

This works only when there is more work to win. In a market that is genuinely contracting, faster delivery moves your share of a smaller pool and does not create demand. The Jevons argument needs real demand underneath it, so check that customers are there before you spend a year on delivery speed.

Read the scoreboard. If your delivery gets faster and better each quarter, hiring will be your problem. If it does not, somewhere in your market a firm is building on its gains, and your reviews and referrals are funding its next hire.

If you want these pieces as they publish, the newsletter is the easiest way to follow along.