They fired workers for AI. Two-thirds are hiring them back.

The 2026 AI layoff has a sequel nobody put in the press release: two-thirds of the companies that cut staff for AI are quietly hiring some of them back.

They fired workers for AI. Two-thirds are hiring them back.

The AI layoff had a very clean story in 2025. The technology could do the work, so the people doing the work were surplus. Cut them, book the savings, tell the analysts the word “efficiency” four times, collect the multiple.

The story had a second half. Nobody wrote it into the press release, because the second half is: some non-trivial share of those people are now being asked to come back.

Writing in Forbes on July 17, John Werner called it a boomerang, and the metaphor is annoyingly accurate. You throw the thing away with force and conviction, and it returns to your face.

The numbers, such as they are

The figure making the rounds comes from outplacement firm Careerminds, reported by CNBC and Fast Company: roughly two-thirds of companies that laid people off citing AI have since rehired at least some of them.

Sitting next to it is a survey stat that keeps getting quoted — that of the leaders who made AI-driven redundancies, about 55% now say the decision was wrong. Take the exact percentages with the usual grain of salt; survey wording moves these things around. But the direction is not subtle, and it’s coming from multiple outlets at once, which is not what you’d expect if it were noise.

Then the forecasts. Forrester expects roughly half of AI-attributed layoffs to be reversed “in some form” by the end of 2026. Gartner has floated a similar line — half of the companies that blamed AI for headcount cuts rehiring for comparable roles by 2027.

Read those two sentences again. The analysts who spent 2025 modeling how many jobs AI would erase are now modeling how many of those same jobs will quietly reappear. Same spreadsheet, minus sign flipped.

What actually broke

The mechanism is not mysterious, and it’s the least surprising thing in the whole saga. AI took over most of a job and choked on the rest.

The example everyone reaches for is IBM, which routed a big chunk of HR to an AI system. By its own account the system handled around 94% of routine requests fine. The other 6% were the ones with a human problem inside them — the escalations, the edge cases, the “this policy doesn’t cover my situation” conversations. IBM has since been hiring again, and notably tripled entry-level hiring in some functions. Ford, per the same reporting, brought back around 350 engineers.

Notice the shape. 94% is a genuinely impressive number and a completely useless one for capacity planning, because the residual 6% doesn’t sit quietly in a corner. It’s the part that generates the complaints, the compliance exposure, the customer who churns. You cannot staff for 94% of a job. Jobs don’t come in 94%.

So the trade that looked like “replace a person with a model” turned out to be “replace a person with a model, then hire a person to handle everything the model can’t, then discover that person is roughly the person you fired.”

The part that isn’t funny

It would be easy to file this under corporate slapstick, and some of it is. But the boomerang doesn’t come back to the same hand.

The worker who was cut ate the gap. Months without income, a resume with a hole in it, a job search in a market where 54% of layoff events this year name AI as the reason. Some get rehired — often as contractors, often without the seniority or the equity vesting clock they had before. The company records a rounding error. The person records a year.

And “rehired” is doing heavy lifting. Two-thirds of firms rehiring some people is not two-thirds of people getting their jobs back. A company that cut 200 and rehired 15 is inside that two-thirds. The boomerang stat describes employer behavior, not worker outcomes, and the two are very different things.

If your job just got automated

The useful read here is not “you’re safe, they’ll hire you back.” Most won’t. The useful read is about where the value sits.

The 6% is the whole game. If your role is being pitched as automatable, the question is not whether the model can do the easy 94% — assume it can, or soon will. The question is whether you’re the person who owns the residual: the judgment calls, the exceptions, the moments where being a human in the loop is the actual product. That’s the part that keeps boomeranging back into the building, because it never left in the first place. It was just, briefly, declared redundant by someone who hadn’t met the other 6%.


Sources: Forbes, CNBC, Fast Company.

Keep reading