Companies fired thousands and blamed AI. Now 55% regret it and they’re quietly rehiring.
Mark Zuckerberg just admitted Meta “made mistakes.” Klarna’s CEO says it “went too far.”
Mark Zuckerberg just admitted Meta “made mistakes.” Klarna’s CEO says it “went too far.” A growing pile of evidence suggests the great AI layoff wave was less about robots being ready and more about executives using a buzzword to do what they wanted anyway.
It’s possible we’re watching the “AI replaced them” story quietly fall apart, and the people who pushed it hardest are the ones now backpedaling.
This week, Mark Zuckerberg issued a rare apology. In a leaked internal memo, the Meta CEO admitted the company “made mistakes and will almost certainly make more” during a restructuring that hit roughly 20% of its workforce, cutting 8,000 jobs and shoving another 7,000 people into AI projects. He promised no more company-wide layoffs this year.
It’s a notable admission. It’s also far from the only one, and once you line them up, a pattern emerges that should make everyone who got laid off “because of AI” very, very suspicious.
Meta’s apology comes with a $145 billion asterisk
Before anyone mistakes Zuckerberg’s mea culpa for a change of heart, look at what Meta is doing with the other hand.
The same company apologizing for layoff chaos just raised its 2026 capital spending forecast to as much as $145 billion, nearly double what it spent in 2025. The “mistake” Zuckerberg is owning isn’t the AI bet itself. It’s the messy, morale-destroying way the cuts were handled.
So the apology is really about smoothing ruffled feathers, hackathons, bigger offsite budgets, while the underlying strategy of betting the company on AI accelerates. Read that way, the memo isn’t a retreat. It’s damage control on the way to spending more.
Klarna already ran this experiment, and it failed
If you want to know how the AI-replacement bet actually plays out, Klarna already ran the whole experiment in public, and it’s the cautionary tale the rest of tech is pretending not to see.
The Swedish fintech proudly replaced the workload of around 700 customer service agents with an OpenAI-powered chatbot, bragging that it cut resolution times and would add tens of millions in profit. CEO Sebastian Siemiatkowski wanted to be “OpenAI’s favorite guinea pig.”
Then the bill came due. Customer satisfaction dropped, service quality cratered, and trust eroded. By his own admission, Siemiatkowski conceded the company “focused too much on efficiency and cost,” and flatly: “We went too far.” Klarna is now rehiring humans, and in the most on-the-nose twist imaginable, it’s repackaging live human support as a premium “VIP” feature. The thing they tried to delete is now the upsell.
Klarna isn’t alone, and that’s the real story
One company overreaching is an anecdote. The list is now long enough to be a trend.
Salesforce openly bragged about cutting roughly 4,000 employees because it “needed less heads with AI,” then spent the back half of the year quietly reloading headcount as the AI proved less capable than advertised.
IBM automated big chunks of its HR with a system called AskHR, hit delays and morale problems, and reversed course to rehire human talent in other divisions. The pattern repeats: bold AI-replacement announcement, quiet walk-back once reality lands.
And the data backs up the vibe. According to Forrester’s 2026 workforce report, 55% of employers now regret laying people off for AI. Gartner projects that about half of all AI-attributed layoffs will reverse by 2027. This isn’t a few unlucky companies. It’s most of them.
The quiet part: AI was cover for cost-cutting
Here’s the uncomfortable thesis under all of this, and the evidence increasingly supports it.
For a lot of companies, “we’re replacing them with AI” was never really about the technology being ready. It was a socially acceptable, even investor-pleasing, way to do what companies have always wanted to do: cut labor costs. “We’re an AI-first company now” plays a lot better on an earnings call than “we over-hired during the pandemic and want to juice margins.”
The smoking gun is in how the rehiring is happening. Forrester predicts most companies that reverse course won’t rehire openly at the old salaries. They’ll do it “quietly, but offshore or at significantly lower salaries.” Read that again. The endgame for a chunk of these AI layoffs isn’t a robot doing the job. It’s a cheaper human doing the same job in a cheaper country, with AI as the story that made the swap palatable.
Even the scary original numbers were shakier than they sounded. IBM’s famous “7,800 jobs AI could replace” was a five-year projection of automatable roles, not a layoff plan. Klarna’s “700” was a workload-equivalence calculation, not 700 specific people fired. The headline figures that powered the panic were inflated from the start.
The fair case, because it’s not all a con
Now the honest other side, because “it’s all a scam” would be just as lazy as “AI is replacing everyone.”
AI genuinely is automating real work. It handles tier-one, high-volume, repetitive tasks, basic customer queries, routine code, first-draft copy, well, and teams using it really can do more with fewer people. That’s not fiction, and pretending otherwise helps no one. Some of the 2024-2025 cuts were also a legitimate correction to the absurd over-hiring of the pandemic boom, which had little to do with AI at all.
And the technology is improving fast. The capabilities that aren’t quite there in 2026 may well arrive. The companies betting big aren’t crazy to bet, they’re arguably crazy not to. The error wasn’t believing in AI. It was believing it was ready now, and firing people based on a promise instead of a proven result.
The consistent finding across every case study is the same: hybrid beats full replacement. AI plus humans outperforms AI alone on both cost and quality, every time. The companies that quietly won are the ones that used AI to make their people more productive instead of trying to delete them.
What this actually means
Strip it all down and here’s where it lands. The “AI is taking the jobs” narrative was oversold, by executives who had every incentive to oversell it, and the receipts are now coming in: regret surveys, reversals, rehirings, and CEOs admitting they went too far.
That doesn’t mean AI is fake or that the disruption isn’t real. It means a lot of companies pulled the trigger on a future that hadn’t arrived yet, used a buzzword to justify it, and are now quietly cleaning up the mess, often by rehiring the same work at a lower price somewhere else.
The people who paid for that miscalculation aren’t the executives who made it. They’re the workers who got walked out the door so a chatbot could fail at their job for a year. Zuckerberg gets to call it a “mistake” and move on with $145 billion to spend.
The 8,000 people Meta cut don’t get a memo. That’s the part of this story nobody in a corner office is putting in a leaked apology.
This op-ed reflects the views of the author.
ClownfishTV.com strives to be an apolitical, balanced and based pop culture news outlet. However, our contributors are entitled to their individual opinions. Author opinions expressed in this article do not necessarily reflect the views of our video hosts, other site contributors, site editors, affiliates, sponsors or advertisers.
Hat Tips:
Entrepreneur and Reuters (June 12-15, 2026), verified for the Zuckerberg “made mistakes” memo, the 8,000 cuts and 7,000 reassignments, the 20% workforce figure, and the $145 billion capex forecast
Forrester Predictions 2026 via Metaintro (February 2026), verified for the 55%-of-employers-regret figure, the “future promise of AI” framing, and the “quietly rehire offshore or at lower salaries” prediction
Gartner via industry reporting (2026), verified for the projection that roughly half of AI-attributed layoffs reverse by 2027
MLQ and DigitalApplied (2025-2026), verified for the Klarna reversal, the “we went too far” quote, the ~700-agent figure, and the human-support-as-premium-feature pivot
Cloud Engineer Academy and Emerald Book (2026), verified for the Salesforce ~4,000 cuts and reload, and the IBM AskHR reversal
ToolDirectory (June 2026), verified for the context that IBM’s 7,800 and Klarna’s 700 were projections and workload-equivalence figures, not direct headcounts