Nearly one in five American workers say they have lived through an unusual workplace reversal: their employer replaced them with artificial intelligence, only to ask them to come back afterward. That is the central finding of new research from talent assessment firm SHL, which has labeled the phenomenon “AI rebounding” — an early signal that some companies may be discovering the limitations of swapping human employees for AI.

“Organizations are making fast, sometimes myopic, decisions about their people based on an incomplete picture of what their employees can do with or without AI,” Lucy Beaumont, global senior vice president of product and design at SHL, said in a press release. “AI rebounding is what happens when a company automates a role before it has genuinely assessed the human capability inside it. 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.”

The survey, conducted in June and July, found that 19.1% of American respondents said they had personally been replaced by AI and were subsequently asked to return to their jobs. Across both the U.S. and the U.K., the figure stood at 16%.

The conditions attached to a return were notable. Among those asked back, 51% wanted a substantial pay increase before agreeing to return, SHL found. Another 42% said they would require written guarantees about future job security, 40% wanted transparency about how AI would be used in their positions.

Those responses also show what workers say would matter when an employer reverses an AI-related replacement decision: compensation, security, and a clear explanation of how the technology would fit into the job. The other surveys and company examples point to a related distinction for employers: automating individual tasks is not necessarily the same as eliminating the role responsible for institutional knowledge, relationships, or quality control.

Companies are finding out AI can’t do everything

The pattern is echoed elsewhere. A July survey from staffing firm Robert Half found that 32% of U.S. hiring managers whose organizations eliminated positions because of AI later had to reinstate the same or similar roles. Among that group, 40% said the positions required institutional knowledge or context that AI couldn’t replace. Another 39% cited relationship management, 39% pointed to increased business demand, and 38% said the roles required more human oversight and quality control than expected.

Several prominent companies have already run into those limitations. Klarna, the Swedish buy-now, pay-later firm, began recruiting human customer service representatives after leaning heavily on AI chatbots. CEO Sebastian Siemiatkowski acknowledged the company had gone too far in its reliance on automation, according to Bloomberg.

A Klarna spokesperson told Fortune that human agents were needed “for complex, sensitive cases.” The company began directly hiring a “small number” of high-skilled humans into the customer service process to identify where the “human touch” can bring the most value to customers.

Ford offers another example. The automaker hired and promoted 350 experienced engineers in June after AI-powered quality control tools failed to perform as expected. The company said AI had contributed to quality improvements but stressed the importance of combining the technology with experienced engineering teams.

“These engineers carry the hard-earned wisdom of decades of design,” a Ford representative previously told Fortune. The company said AI is very important to quality gains, “and that, in tandem with deep technical expertise, is what’s needed.”

AI layoffs have been looming, but jobs are growing

Despite the rise of AI rebounding, the number of employers citing AI as a reason for workforce reductions has grown substantially. According to outplacement firm Challenger, Gray \u0026 Christmas, employers cited AI in 121,401 announced job cuts through October — more than double the 54,836 announced layoff plans for all of 2025.

Younger workers are especially vulnerable to AI-led job displacement. A study from the Stanford Digital Economy Lab found that employment among workers aged 22 to 25 in occupations highly exposed to AI was 19% below where it would have been had it grown at the same pace as employment among their less-exposed peers.

SHL’s research does suggest some promise for younger workers, however. The firm found they were more likely to report experiencing “AI rebounding”: among respondents aged 18 to 24, 24% said they had personally been replaced by AI and later asked to return, compared with 19.1% of U.S. respondents overall.

This story was originally featured on Fortune.com.