The AI Boomerang: Regretting AI Layoffs, Rushing to Rehire

How Many Workers Have Actually Lost Their Jobs to AI?

Workers have lost their jobs to AI in six-figure numbers so far in 2026 alone. Challenger, Gray & Christmas recorded 101,743 AI-attributed layoffs between January and June 2026 — nearly double the 54,836 recorded across all of 2025. AI has now led every layoff reason category for four consecutive months, from March through June 2026.

Behind that acceleration is a sector that moved faster than almost any other in its history. Technology alone accounted for 139,156 job cuts through June 2026 — and even Challenger’s own analysts think the real AI-driven number is higher than what companies are willing to admit. Andy Challenger, the firm’s chief revenue officer, has said employers increasingly reach for softer language like “technological update” specifically to keep the AI label off the announcement.

AI-cited layoffs nearly doubled in the first half of 2026 compared to all of 2025.

That distinction matters more than it sounds. A company that says “AI” is making a bet about the future. A company that hides behind “technological update” already suspects the bet didn’t pay off — and would rather nobody ask what happened to the people who were cut for it.

None of this happened by accident. Behind every one of those cuts was a leadership team that believed, with real conviction, that artificial intelligence could do the job a person used to do — and that belief is where this story actually starts.

Why Did Executives Believe AI Could Replace Their Employees?

Executives believed AI could replace their employees because the belief was contagious before it was tested. A 2026 Mercer survey of company leaders, reported via Tom’s Hardware, found that nearly all of them expected AI to drive at least some headcount reduction within two years. Fewer than a third believed their own organization could actually combine human and AI labor effectively — meaning most leaders were cutting for a capability they privately doubted they could manage.

That gap between confidence and readiness wasn’t hidden. It was published, discussed, and then acted on anyway. Oxford Economics went further and found that many companies announcing AI-driven layoffs were, more accurately, correcting for pandemic-era overhiring — and used AI as the more palatable story to tell investors and boards. The term now circulating for this is financially motivated cuts wearing a technology narrative: “AI-washing.”

The pressure to look decisive mattered as much as the technology itself. A company trimming headcount for “cost-cutting” reads as retreat. A company trimming headcount for “AI transformation” reads as strategy. Forrester’s research found many organizations announcing AI-related layoffs didn’t yet have a mature, tested AI system ready to fill the roles they were eliminating — the technology was the story, not the plan.

None of this means the technology was irrelevant. It means the decision to cut came before the proof that it would work — and the next question is whether anyone checked.

Did Companies Have Data to Justify Laying Off Employees for AI?

Companies mostly did not have data to justify laying off employees for AI. Thomas Davenport’s survey for Harvard Business Review, covering more than 1,000 executives, found that AI-related workforce cuts were made almost completely in anticipation of AI’s potential — while the companies making those cuts were still waiting for generative AI to deliver the productivity gains they’d promised investors.

The gap gets more specific the closer you look at it. Orgvue’s own research found that 23% of companies making AI-linked layoffs admitted the decision was based on general assumptions about what AI could do, not a role-by-role analysis of what the eliminated employees actually did all day. Careerminds found something similar from the other side of the table: 55.1% of HR leaders said reskilling or redeployment for the affected employees was never formally discussed before the cuts were made.

Most AI layoff decisions leaned on assumptions rather than role-specific analysis.

That check would have shown up somewhere. It didn’t — not until the layoffs were already done and the people were already gone.

Why Couldn’t AI Actually Replace These Employees?

AI couldn’t actually replace these employees because it excelled at the routine share of the work and failed at exactly the part that required judgment.

Ford: The Graybeard Return

Ford found this out in its own factories. The company had leaned on automated systems for quality inspection and design review, cutting experienced engineers along the way — until quality problems started surfacing that the automated systems weren’t catching. Charles Poon, Ford’s vice president of vehicle hardware engineering, put it plainly:

Ford rehired, promoted, or newly hired 350 veteran engineers — a group the company internally calls its “graybeard” engineers — to mentor younger staff and catch what automation missed. Ford topped J.D. Power’s 2026 Initial Quality Study for the first time since 2010 after bringing them back.

IBM: The Last 6%

IBM’s version of the same problem was narrower, and in some ways more revealing. Its AI system resolved 94% of routine HR requests without any issue.

IBM’s chief human resources officer, Nickle LaMoreaux, later explained the company’s decision to triple entry-level hiring in the US: “If we don’t continue to invest in entry-level hires, what happens in three to five years? There’s no pipeline; the well simply dries up.”

The same gap shows up in how AI itself gets trained. A technique called Reinforcement Learning with Verifiable Rewards (RLVR) has become one of the primary ways companies train AI models without a human checking every output — the model generates an answer, a rule-based verifier checks it against a known correct result, and the model reruns until it matches. It works well for math and code, where “correct” can be checked automatically. It has no answer for the other kind of task. Researchers building RLVR systems have said as much themselves: the method captures correctness, not helpfulness, safety, or tone — the exact qualities IBM’s remaining 6% turned out to require. The industry’s own fix for removing humans from the loop only works on the narrow slice of work that doesn’t need one.

Commonwealth Bank: When the Voice Bot Couldn’t Keep Up

Commonwealth Bank of Australia’s failure was more public. The bank replaced more than 40 customer service roles with an AI voice bot, then watched call volumes surge as the system couldn’t keep up with real customer demand. CBA reversed the redundancies within months and admitted, in a public statement, that its “initial assessment that the 45 roles in our Customer Service Direct business were not required did not adequately consider all relevant business considerations.” A union official called the reversal “a massive win for workers.”

Klarna: 700 Agents, Minus the Judgment

Klarna’s collapse ran the longest before anyone said so out loud. The Swedish fintech cut hundreds of customer service roles after claiming its AI chatbot could handle the equivalent of 700 human agents. Customer satisfaction declined. CEO Sebastian Siemiatkowski eventually acknowledged that cost had been “a too predominant evaluation factor,” and the company began rehiring humans into the roles it had just eliminated.

Four companies, four industries, one repeating pattern: judgment was the part AI couldn’t absorb.

Four companies, four industries, one pattern repeating itself: the work AI couldn’t do wasn’t a technical edge case. It was judgment, context, and the willingness to be accountable for a decision — the parts of the job nobody thought to put in the AI’s job description because no one thought a person would ever need to be hired back to do it.

How Widespread Is AI Layoffs Regret, and What Are Companies Doing About It?

So far, this is a story about a bet that didn’t pay off. What’s left is what companies did once they realized it — and whether it actually worked.

AI layoffs regret is widespread, and companies are responding by rehiring. Orgvue’s 2025 survey of business leaders found that among the 39% who made employees redundant because of AI, 55% later admitted the decision was a mistake. Forrester reached a similar number through separate research: 55% of employers now say they regret laying off workers because of AI. Two different firms, two different survey methods, the same conclusion arrived at twice.

The rehiring isn’t evenly spread. Robert Half’s survey of nearly 2,000 US hiring managers found finance leading the reversal at 44%, followed by HR at 35% and technology at 32%.

Finance leads all sectors in rehiring roles that were cut for AI.

Finance topping that list is not a coincidence. It’s one of the most judgment-heavy functions in any company — the same category of work that keeps showing up as the part AI couldn’t absorb in the previous section.

The speed of the reversal is what makes this hard to write off as caution rather than regret. Careerminds surveyed 600 HR professionals who had made AI-driven layoffs in the prior year and found that 52.1% had already rehired for the eliminated roles within six months, and another 17.8% had started within three. Only 2.1% waited more than a year. Companies weren’t quietly reconsidering over time. They were reversing course almost as fast as they’d made the cut.

That speed says something the regret percentages don’t. A company that takes years to admit a mistake might just be slow to notice. A company that rehires within three months already knew.

Is the Rehiring Actually Succeeding — Or Are Workers Moving On Without It?

The rehiring is not clearly succeeding financially, and a portion of laid-off workers are not waiting for it. Careerminds found that nearly 31% of companies that rehired after AI-driven layoffs ended up spending more than they’d saved from the original cuts, and another 42.4% broke even — meaning fewer than three in ten came out ahead. Those figures don’t count the costs that don’t show up on a balance sheet: lost institutional knowledge, declining morale, customers who’d already moved on.

Fewer than three in ten companies came out financially ahead after rehiring.

There’s a harder truth underneath the rehiring headlines. Forrester’s research doesn’t just predict that AI-attributed layoffs will reverse — it predicts they’ll reverse cheaper. The firm expects much of this work to return “offshore, or at significantly lower salaries,” which means the person answering the phone at the reopened role may not be the person who was let go from it.

Some workers never waited to find out. LinkedIn’s own data shows “Founder” as one of the fastest-growing titles people add to their profiles — up 60% year-over-year across Europe, with the UK specifically up 69% and the Netherlands up 85%. Labor-market analytics firm Revelio Labs tracked the same pattern in the US: the share of job switchers becoming entrepreneurs nearly doubled between early 2022 and early 2025, and the average age of someone making that switch has dropped to under 34.

The reasoning behind that shift is straightforward. The same AI capability that made a company think it didn’t need a department is available to a single person for a few thousand dollars a year, not a few hundred thousand in payroll. A former employee doesn’t need to out-earn their old salary immediately — they need to out-run the six-month wait Careerminds found most rehired workers actually got.

So the honest picture isn’t “companies fired people and now they’re making it right.” It’s messier: some rehiring is real and expensive, some of it is quieter and cheaper than it looks, and some of the people who got cut have already stopped waiting to find out which one they’d get.

Will AI Ever Fully Replace Human Workers?

AI will not fully replace human workers based on the evidence available now, though credible experts disagree sharply about the future — and their own positions keep shifting. Anthropic CEO Dario Amodei told Axios in May 2025 that AI could eliminate up to 50% of entry-level white-collar jobs within five years and push unemployment to 10–20%. A year later, at a May 2026 briefing alongside JPMorgan CEO Jamie Dimon, he was invoking the Jevons Paradox instead — the 19th-century observation that efficiency gains expand demand for a resource rather than shrinking it.

OpenAI CEO Sam Altman moved in the opposite direction. At a Sydney banking conference in May 2026, he admitted: “I thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened. I’m delighted to be wrong about this.”

The data collected so far leans toward the skeptics. Yale’s Budget Lab found no statistically significant effect on unemployment through March 2026 for workers in high-AI-exposure occupations — though the Lab’s own director has cautioned against reading this as proof the disruption isn’t coming, only that it hasn’t shown up yet. The European Central Bank found AI-intensive firms were about 4% more likely to be hiring than firms that don’t use AI heavily.

My read: the evidence supports a narrower claim than either CEO originally made. AI is displacing codifiable, high-volume, routine work — which is exactly why entry-level roles keep showing up as the most exposed category throughout this article. What the data doesn’t support is AI displacing judgment, accountability, or the ability to repair a relationship after something goes wrong — the same four things Ford, IBM, Commonwealth Bank, and Klarna all rediscovered the hard way.

PwC’s 2026 Global AI Jobs Barometer, drawn from more than a billion job postings across 27 countries, found something that complicates the doom narrative further: companies in the most AI-exposed sectors grew headcount 52% since 2018, compared to 36% for the least AI-exposed companies. The firms most associated with AI disruption are, on average, the ones hiring the most people — not the fewest.

None of this settles whether AI eventually crosses into judgment-based work. Even the two men building it can’t agree, and both have already changed their minds once. It settles what’s true right now, which is that the boundary between what AI can do and what it can’t has moved slower than the last three years of layoff announcements implied.

The Human Layer: How AI and People Should Actually Work Together

Every company that cuts for AI is making a bet about which parts of a job actually need a person. Ford, IBM, Commonwealth Bank, and Klarna all bet wrong about the same four things — and a broader pattern researchers have found across the industry makes their mistake look almost routine: companies that automate aggressively often end up cutting the very staff who would have caught the AI’s mistakes, because nobody categorized oversight as a job worth keeping. The companies pulling ahead didn’t wait to find out the hard way.

The Human Layer: Oversight, Exception Handling, Elevation, and Design.

The Human Layer was never the part of the business AI was going to replace. It was the part keeping AI usable. IKEA and Ford designed for it before the pressure hit, and it shows in their numbers. The companies you just read about designed for it after — while writing rehire offers to the people they let go a year earlier.

Conclusion

The AI Boomerang isn’t a story about artificial intelligence failing. It’s a story about companies discovering the Human Layer only after they’d already cut it — Ford, IBM, Commonwealth Bank, and Klarna all paid for the same lesson in rehires, not savings. AI layoffs regret turned out predictable, not exceptional: Orgvue put a number on the cost, $1.27 spent for every $1 companies thought they saved. IKEA and Ford show the alternative was available all along. If AI is reshaping how your team hires, The 2026 AI Hiring Doom Loop covers the other half.