Klarna replaced 700 employees with AI. The result? Quality dropped. Customers noticed. The CEO eventually said cost had become "a too predominant evaluation factor." Then they started hiring humans again.
Read that again.
They fired people for AI. The AI didn't perform as promised. They had to rehire people. And the CEO's public explanation was essentially...we optimized too hard for the wrong thing.
That's the whole story. Everything else is footnotes.
A Resume.org survey of 1,000 hiring managers found that 59% of companies admitted AI was a convenient framing for layoffs they were going to make anyway. The Federal Reserve studied data from more than a million firms and found no statistical link between AI adoption and reduced job postings. None. Sam Altman called it "AI washing," which means the people building the technology were publicly embarrassed by how it was being used as an excuse.
And yet the narrative ran because it played better in the boardroom than the truth.
The truth being: they over-hired during COVID, the market tightened, and they needed a story that didn't make them look like they panicked.
"AI is going to handle this." "The technology is ready." "We've done the analysis."
If any of those sentences were said in your organization before a headcount decision, someone should have made one demand..."Show me exactly where in the workflow AI takes over, and exactly what happens when it gets it wrong".
Most people didn't because asking meant slowing down the announcement. Or admitting they hadn't actually done the analysis they said they had.
Forrester Research named this directly in their Predictions 2026 report. They expect half of AI-attributed layoffs to be quietly reversed. A February 2026 survey of 600 HR professionals found that more than a third of companies had already rehired more than half the roles they eliminated, with over half doing so within six months. Only about one in five said AI fully replaced those roles without operational problems.
The part nobody's announcing: "rehiring" often means offshore. Companies aren't standing at a podium saying they got it wrong. They're quietly refilling seats with cheaper labor and internally calling it a strategic evolution. The original decision stays on the books as a win. The people who paid for it don't get that framing. (They rarely do.)
Here's what AI is actually good at:
- Repetitive tasks
- Pattern matching
- High-volume data processing
- First drafts on structured outputs
It handles the repeatable parts of repeatable jobs reasonably well.
Here's what it cannot do:
- Replace someone who knows why your second-largest customer went quiet for three weeks in 2019 and what it took to get them back.
- Replace judgment calls built on years of knowing a vendor's actual behavior versus their contract language. Replace the person who can read a room and knows when not to push.
That capability gap is not closing as fast as the slide decks suggested. That's a fundamental difference between what the technology is and what people convinced themselves it would become.
Only 16% of workers had what researchers classified as high AI readiness in 2025. Only 23% of AI decision-makers said their organizations offered any formal training on using the tools. People are figuring it out on their own, with no governance, no oversight, and no agreed definition of what good looks like. That's not transformation, its improvisation with a risk profile nobody calculated.
If you're running a business and someone in the last eighteen months told you AI was going to significantly reduce your headcount or replace a core function, ask yourself some questions...Did they show you the specific workflow, the specific failure mode, and the specific metric that would tell you whether it worked? Or did they show you a slide with a percentage and a logo?
Those are two different conversations. One is a technology decision and the other is a bet. The companies that lost the bet are now quietly filling the gap and hoping no one connects the dots. Someone should have asked the uncomfortable questions before the check was signed.
That's not a technology problem, it's a leadership one. And it's exactly the kind of thing that goes unanswered when everyone in the room has already decided what they want to hear.
About the Author
Raphael Savastano is the founder and principal consultant of ROFONIC LLC. With 25+ years in IT, 16 years in leadership, including 8 years as CIO scaling technology for a global manufacturer from M to 0M. He now helps growing companies get executive-level technology and operations leadership without the full-time cost. Want to know where your technology actually stands? Take the Founder’s IT Reality Check →
