Executive Summary

Key Figures

Four numbers follow. The first two are shares of workers who say they were asked back, the third is the share that picked a pay increase as the condition for returning, and the fourth is the rehiring share counted on the organization side.

Sources: Fortune, "AI rebounding" (November 9, 2026) and CNBC Make It, "AI-driven layoffs are creating a new class of boomerang employees" (October 30, 2026).

SHL, a firm working in talent assessment and talent management, surveyed workers in the United States and the United Kingdom across June and July 2026. The fieldwork went to the market research company Propeller Insights, and 2,036 adults currently in work answered. That is 1,029 in the U.S. and 1,007 in the U.K., with ages running from 18 to 65 and over. The margin of error is ±5 percentage points. The results came out as a report titled "Talent Mobility at Work in 2026: The Hidden Workforce."

The item that stood out was the callback. Among U.S. respondents, 19.1% said that after being laid off or displaced because of AI, the same employer offered to take them back. In the U.K. the same answer drops to 13%, and across both the U.S. and the U.K., the figure was 16%. Splitting by age widens the gap. Among respondents aged 18 to 24, 24% said it had happened to them personally, and 36% said they are close to someone it has happened to. CNBC put the same item at 19.6%, so the decimal moves a little from outlet to outlet.

One item counts more than personal experience. Adding those who went through it, those who know someone who did, and those who have heard of it happening, 52% of all respondents said they had encountered the pattern in some form. Counted as a personal loss of a seat it is one in five; counted as talk moving around a workplace it is more than half. The two measure different things, and it is better not to read them together.

SHL called the pattern AI rebounding. A rebound is something that bounces away and ends up back where it started, and the word went to people who were let go and then walked back in. CNBC on October 30 and Fortune on November 9 put the name straight into their headlines, and within about ten days it had settled as the name of the thing.

Not many have shut the door for good. The share who say they would not go back under any conditions stays in the single digits. Even here the sources part. CNBC and People Matters put it at 6.6% of U.S. workers, while SHL's own press release took being replaced by AI as the denominator and wrote one in 13, or 8%. With different denominators the two cannot stand on the same line, and what they say together stops at this: the refusers are a minority. That leaves the question of what the willing ones ask for, and the terms are not light. Section 4 takes them up.

Lucy Beaumont, Global SVP for Product and Design at SHL, defined the pattern in a single sentence. "AI rebounding is what happens when a company automates a role before it has genuinely assessed the human capability inside it." The order of the decision has been inverted. In SHL's press release the next sentence gives the result. "The AI doesn't deliver, the gap becomes obvious, and the business ends up rehiring the same experts they just let go of, at a far greater cost."

Beaumont also named what was missing. Organizations, she said, "are making fast, sometimes myopic, decisions about their people based on an incomplete picture of what their employees can do with or without AI." The incomplete picture here means assessment material. A company had confirmed that AI could handle a given task, but it had not measured what the person in that seat was holding beyond the task. Her phrasing to CNBC is shorter. Some employers "pulled the plug too quickly."

One more item inside the same survey supports the diagnosis. Close to half of the respondents answered that their employer is bringing in AI tools, and the number who answered that roles and work processes are being rebuilt around what those tools can do was 24%. That 24% is a different item from the age figure in the previous section, and only the number matches. Tools come in; rebuilding the work around them happens at one site in four.

The same report measured the people side as well. Fewer than half of workers, 42%, believe their manager fully understands their skills beyond their job title, and in the U.K. that falls to just over a third, 36%. At the table where someone decides what to automate, the file on the people involved is less than half full. That is where the report's subtitle, "The Hidden Workforce," comes from. Capability is not absent from the organization, the report wrote; it is not visible enough to reach the decision.

Kathy Ross, a VP analyst at Gartner, called the pattern a "cut fast, rehire later" trap. The basis for reducing headcount, by her account, was "expected AI value rather than demonstrated results." That one line between the expected and the measured is where rebounding begins.

SHL asked individual workers. Two surveys this year asked organizations instead. In a report published in March 2026 by Orgvue, an organizational design analytics company, 32% of the organizations that had cut staff because of AI went on to hire people again. Robert Half, the staffing firm, put the question to U.S. hiring managers in July and came to 32% as well. The answer covers roles removed because of AI that were brought back as the same job or a similar one.

Robert Half also asked why they were brought back. The most common reason was the organizational knowledge and context AI did not take with it (40%), followed by needing a person to manage relationships (39%). That the amount of work itself had grown came in at the same 39%, and that human oversight and quality checking were needed more than expected followed at 38%. Separating the four, only the growth in work points to something outside AI; the other three point at nobody having counted what the person in that seat was holding before sending them out.

Put the three figures in one place and the direction looks the same. The denominators, though, have to be read alongside them. SHL's 19.1% is the portion of individual workers who received a callback, and the 32% in the other two is the portion of organizations that cut for AI and then hired again. All three count the same event from a different place, and they are not values to add up or set against each other.

Fortune pointed to Ford and Klarna as companies that called people back. Ford hired and promoted 350 experienced engineers in June 2026, after AI-powered quality control tools failed to perform as expected. Of the design engineers it took back, the company said, "These engineers carry the hard-earned wisdom of decades of design." Klarna, which had moved customer service over to AI, put human agents back in place "for complex, sensitive cases."

How much there is to undo can be gauged from the layoff side. Of the announced job cuts counted by Challenger, Gray & Christmas from January through October 2026, employers cited AI in 121,401. All of 2025 came to 54,836, so ten months more than doubled it. Taking only the cuts announced in May 2026 out of the same count, about 40% gave AI as the main reason. This count, too, gathers the reasons companies wrote into their announcements, so it is not a number that confirms AI was the actual cause.

There is a line to stop at here. All three surveys are self-reported. SHL counted experiences that workers wrote down as having happened to them, and Orgvue and Robert Half counted decisions that people on the organization side reported about themselves. None of it is a tally matched against employment records. So these numbers cannot say how many people AI actually replaced, or how many of those who got a callback actually went back. Whether companies met the terms sits outside the survey too. Cost is the same. SHL's press release wrote that the side doing the rehiring pays "a far greater cost," but no item in this survey measured how much greater. Three independent surveys pointing the same way supports the existence of the current; it does not measure its size.

Asked what would have to be in place before they went back, the answers gathered into four branches. At the top is money, but the three underneath do not convert into money.

Two places in this table need care. One is rounding. This article takes as its basis the 51%, 42% and 40% that Fortune wrote, while CNBC and People Matters carried the same items as about 50%, 43% and 40%.

The other is the denominator. Fortune wrote these shares as values "Among the returnees," and CNBC wrote them as values for "surveyed U.S. workers." The gap between the two is not small. On the first reading the share is counted inside the group who got a callback; on the second it is a hypothetical put to all workers. SHL's own press release carries no word pointing at a denominator in front of these shares. The same numbers are circulating on different denominators, and which one holds is settled only by opening the report itself.

The second and third rows aim at the same place. Job security in writing asks for a promise that the decision will not be reversed a second time, and transparency about AI use asks to be shown in advance how far that decision reaches. Beaumont said returning workers are "going to be a lot more demanding of what they expect from an employer," and she put the reason briefly. "You've really broken the trust with those people." She also carried over the question an employee holds while deciding whether to go back. "How do I know that this isn't going to happen again?"

From the company's side the third row is the hardest. Explaining to employees where AI is used means having that written down first, and as the previous section showed, the share saying roles are being redesigned was 24%. More companies have nothing to explain yet.

Turn Beaumont's definition around and one requirement falls out of it. Assess the capability of the people inside a role before automating it. Assessing needs something to measure, and measuring needs a record. What the companies where rebounding happened lacked was not AI performance. It was that record.

This shape matches what the Pebblous blog has written about data more than once. We have covered how a company's own verdict that its data is ready gets reached without measurement, and separately what a company looks at when it writes down that a layoff was caused by AI. This time the subject is work rather than data, and the place where no record exists to check the self-assessment is unchanged.

Automation comes back to being a data problem in the end. Judging which work can be handed to AI requires that what the work is made of survives as data. If the record does not hold who does it and how often, how many exceptions arrive in a month, and what judgment enters when those exceptions are handled, the only basis left is the expected value. That is the spot Kathy Ross meant by "expected AI value rather than demonstrated results." Joseph Fuller, who teaches management practice at Harvard Business School, took the same spot from the other side. "The more you know about important processes, the more value you can extract from the AI." You get out of automation only as much as you know about the work, and knowing comes down to how much was written down.

Three questions can be put to your own organization. First, is there a record of the volume, the handling time and the exception rate for the work you mean to hand to AI? Second, is what the person doing that work held beyond the task written down anywhere? Third, once AI is in, can you measure again with the same yardstick and line up before against after? With any one of the three missing, whether the automation decision was right or wrong gets known only after the person has gone.

Primary Sources

• 1.SHL. (2026). "Workers Replaced by AI Would Demand Higher Pay and Greater Job Security to Return, SHL Research Finds." Press release announcing Talent Mobility at Work in 2026: The Hidden Workforce, October 29, 2026.
• 2.Orgvue. (2026). "New research exposes the complexity of deploying AI systems in the workforce." March 12, 2026.
• 3.Robert Half. (2026). "Navigating the AI Era: Adoption and Workforce Change."

News Coverage

• 4.Hong, J. (2026). "'AI rebounding': Nearly 20% of U.S. workers replaced by AI are being asked back—now they're demanding higher pay, job security and AI transparency." Fortune, November 9, 2026.
• 5.Caldwell, S. (2026). "AI-driven layoffs are creating a new class of boomerang employees—companies are realizing they 'pulled the plug too quickly'." CNBC Make It, October 30, 2026.
• 6.Srivastava, S. (2026). "Nearly 1 in 5 workers laid off due to AI are asked to return, survey finds." People Matters, November 2, 2026.