AI Layoffs Can Start With the Clients You Lose

If you run a service firm, I wouldn't read every AI layoff headline as a preview of what happens when you give your own team the tools.

AI can replace work and people can lose jobs because of it, that part is real. But your firm can also lose jobs because somebody else gets better at serving the market and you stop winning enough work to keep the team busy.

That's the risk I want you to look at alongside the tool itself. From my perspective, the safer question is not just how many people AI could replace, it's whether the way you use it helps your company keep winning and delivering work.

I've watched hundreds of service firms adopt AI through a franchise network's delivery data, and in the first half of 2026, more than one client told me they were gaining market share.

When I asked about what had changed, I found a connection to AI tools, redesigned processes, faster delivery and a clearer view of their own data. Better outcomes, better Google reviews, more referrals, more business, that was the chain I could see.

I'm describing those clients' experience and what I found when I probed it. I don't have a study showing how much of their competitors' lost work was caused by AI, or how many jobs those competitors then cut.

But you don't need to pretend that part is measured to take the operating risk seriously. If your clients get a better result somewhere else, your own decision not to automate doesn't protect the work they're paying you for.

Now let me give the layoff numbers their proper place, because I am not dismissing them. In Challenger, Gray & Christmas's report covering the first half of 2026, employers cited AI in 101,743 announced job cuts.

That was about 23% of all cuts announced in that period, and roughly 85% more than the 54,836 AI-cited cuts reported for all of 2025. In June, AI led the reasons listed for the fourth consecutive month, those are the June report's figures, not a running total through today.

The important word is "cited." Challenger records the reasons employers give, it doesn't audit every job to establish whether AI replaced it, whether the business was already cutting costs, or whether several things happened together.

The job loss is real for the person affected. The explanation still needs more care than treating every mention of AI as the same event.

And a layoff count won't show you every change in the team. A firm can delay a budgeted hire, leave a vacancy unfilled, or decide not to bring in juniors before it announces a single cut, so I wouldn't use the tracker as the only way to understand employment either.

There is serious evidence of pressure on younger workers, and it belongs in this argument. Stanford's Digital Economy Lab revised its Canaries study in August 2026, using ADP payroll data covering millions of US workers through June 2026.

It reported that employment among workers aged 22 to 25 in AI-exposed occupations stood 19% below where it would have been had it kept pace with their less-exposed peers. Experienced workers did not show a comparable gap.

That is a relative employment gap, not a finding that 19% of all young workers lost their jobs. The study also found declines concentrated in occupations where AI use mainly substitutes for tasks, rather than complements workers, but the authors call these descriptive indicators, not causal estimates.

So yes, entry-level work deserves attention. I wouldn't reassure a junior by saying that firms can grow with AI and therefore their particular role is safe, because a growing firm can still change which people and skills it needs.

There is another side to the evidence, though. A Journal of Financial Economics study of AI investment from 2010 to 2018 found that AI-investing firms grew faster in sales, employment and market value, with product innovation explaining much of that growth.

That period predates the current generative-AI wave, and the growth was concentrated among larger firms, so I wouldn't treat it as a forecast for a small consultancy. It does show why automating some tasks and growing the workforce are not opposites.

PwC's 2026 AI Jobs Barometer reports faster headcount growth at companies most exposed to AI than at those least exposed, along with productivity growth that is 40% higher. Exposure means the work is more open to AI's effects, it is not the same thing as a measured rollout at every firm.

PwC's read is that AI can expand jobs when firms use it to open markets and create more value. That's a reason to examine the growth route, not a guarantee that buying tools causes a company to hire.

The size distinction is useful too. A 2026 Atlanta Fed working paper drawing on nearly 750 CFOs found large companies expected to shed workers because of AI while smaller firms anticipated modest employment growth.

Those are expectations, not hiring outcomes the researchers later observed. Separately, the paper links productivity gains with innovation and demand, rather than cost reduction alone, and I would keep those two findings separate.

A 2025 US Chamber of Commerce survey reported that 82% of small businesses using AI had increased their workforce over the preceding year. Again, that is an association in a survey, not proof that AI created each of those jobs.

The reasonable conclusion is that adopting AI does not require your headcount to fall. What happens to demand, what you sell and how you use the capacity all matter.

You can see the mechanism in an older example. James Bessen's account of ATMs for the IMF describes banks needing fewer tellers per urban branch, from 20 to 13 between 1988 and 2004.

The branches became cheaper to run, and banks opened 43% more urban branches over that period. Teller employment did not fall in the way the task-automation story alone would have led you to expect, while the job shifted toward relationship work.

That is a historical example, not a claim that banking employment stayed protected forever. The useful part is the extra demand that became practical when the cost of serving it fell.

For a service firm, the equivalent might be work you couldn't price before, more clients you can serve without sacrificing quality, or a faster turnaround that helps you win an engagement. The savings have to help you reach that work, not just leave the same team doing the same amount in the same week.

Apollo chief economist Torsten Slok calls this the Jevons employment effect from AI, cheaper professional work can expand the market for it. In his April 2026 note, he also pointed to unusually high US business formation and read AI as a likely contributor, his interpretation rather than proof of the cause.

That's why the destination of the saved time matters so much. If nobody decides what the capacity is for, you haven't made the growth side of the argument happen.

And cost reduction on its own can take you somewhere you didn't intend. In February 2024, Klarna said its AI assistant was doing the equivalent work of 700 full-time customer service agents.

That is Klarna's workload-equivalent claim, not a verified count of 700 people dismissed because of that assistant. By May 2025, about fourteen months later, the company was recruiting human agents again, including a freelance-style pilot.

CEO Sebastian Siemiatkowski said, "As cost unfortunately seems to have been a too predominant evaluation factor when organizing this, what you end up having is lower quality."

Klarna still described itself as AI-first, so this was not a complete rejection of the technology. It was a public admission that the balance between cost and service quality needed changing.

IBM's account shows a different use for savings. Its AskHR system automated 94% of simple, routine HR tasks, such as vacation requests and pay statements, according to reporting quoting IBM's HR technology leadership, while a couple hundred HR roles were absorbed.

CEO Arvind Krishna said, "Our total employment has actually gone up, because what [AI] does is it gives you more investment to put into other areas." The areas he named included software engineering, marketing and sales.

These are two company accounts, not a controlled comparison of the same tool. They show different decisions about what to preserve and where to put the resources, and neither guarantees what happens to a service firm your size.

But the quality question is even closer to home when the service is what you sell. A lower cost doesn't help much if the client stops trusting the answer, and automating the preparation doesn't remove your responsibility for the judgment in the deliverable.

Now come back to the firm losing work, because its AI-related problem may not appear in an AI layoff announcement at all.

In my client observations, nobody described a formal announcement that customers were switching to an AI-native provider. The changes showed up in outcomes, reviews and referrals, the ordinary ways one firm becomes easier to recommend than another.

There is an older account from the other side. In January 2024, a person posting as the owner of a content development agency wrote, "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."

That is one person's account, not audited business data, and it predates the 2026 figures above. It also describes clients choosing AI instead of buying writing, which is different from choosing a competing agency that uses AI better.

Both routes matter to the owner, but we should not merge them. A customer doing the work themselves and a customer moving to another service firm can both reduce your demand, even though they tell you different things about what to change.

The broader pattern of productive firms pulling ahead is older than generative AI. An OECD study from 2016, using firm data through 2014, found the most productive firms gained market share relative to less productive firms.

That doesn't measure today's AI-driven share transfer. It gives the forecast a historical basis, and my view is that AI can speed up that competition when firms use it to improve the work customers actually buy.

I cannot tell you what share of future layoffs will come from firms losing to better AI users. I can tell you that doing nothing with your delivery operation is not a way to keep competition away from your team.

So I would make the next decision around work you want to win and work you need to deliver better. Choose one part, use AI where it helps, and measure the whole result, including checking and rework, before counting the time as available.

Decide beforehand what any saved time will fund, better deliverables, training, another process improvement, or more client work that is actually there to take on. Then choose the next part and repeat.

That's how you move through the Production Gap, rather than treating an announcement as an operating change. It can support the move toward Stage 3 on the Delivery Model Ladder, where revenue can grow without adding people at the same pace.

Keep the work coordinated too. A company-wide order to "use AI" without clear boundaries can create shadow AI, and silence about the staffing plan can add to people's anxiety about the tool.

Your team deserves to know what you are trying to change, and you need their help to find where the work improves or breaks. That is more useful than asking them to celebrate a tool while wondering whether their job is the saving you have in mind.

There is a real limit to the growth route. If the market is contracting and there is no more work to win, faster delivery does not create demand by itself, and even a firm that gains share may still have to make difficult staffing choices.

Look at demand, quality, win rates, reviews and referrals alongside the cost of delivery. Then make the staffing decision from the work the firm can sustain, not from a headline or a promise on a software box.