For three years the loudest man in finance told the world that artificial intelligence could replace his employees — and that firing them was the entire point. Sebastian Siemiatkowski slashed Klarna’s workforce from roughly 5,500 people to 3,400 and bragged that one OpenAI-built assistant was doing the work of 700 human agents. Now the same man is quietly hiring humans back and conceding that the all-AI operation delivered worse service. The most expensive part of the AI revolution was never the machine. It was the layoff that was supposed to pay for it.
What Siemiatkowski did is not a quirk of one fintech. It is the shape of the entire AI job purge, and that purge is now running in reverse. Companies spent two years announcing that AI had made their workers redundant, watched their share price pop on the news, and are now discovering that the savings never showed up and the people had to come back. More than half of the employers who cut jobs for AI now say they regret it, and the retreat is costing more than the cuts ever saved. The genie did not climb back into the bottle. It sent the company an invoice for the rehire.
Forget the executive’s ego and feel the arithmetic of what these reversals cost. One workforce analytics firm put it bluntly: for every dollar a company claws back by cutting people, it ends up spending about $1.27 putting the work back together — severance, lost productivity, recruiting, onboarding. Nearly a third of the organizations surveyed spent more on rehiring than they saved by automating in the first place, and another four in ten roughly broke even. That is not a savings program. It is a round trip you paid for twice.
Sebastian Siemiatkowski — He Called AI Worth 700 Workers, Then Admitted It Wasn’t
The Klarna chief was the movement’s poster child, the man who made “we replaced staff with AI” sound like a boast instead of a warning. He cut the payroll from about 5,500 to 3,400 and told the world his assistant handled the work of 700 agents. Then customer satisfaction slipped, and the arithmetic caught up with the story. Siemiatkowski conceded that the all-AI approach had produced lower-quality service, and insisted that customers must always be able to reach a human being when they want one. Klarna started hiring people again. The man who sold a future without workers is now staffing it. The most efficient company in Europe just learned the cheapest worker is the one it already trained, then fired.
The Gartner Data — Firing People to Fund AI Buys You No Extra Return
The reversals are not sentiment. They are data. Gartner surveyed 350 executives at companies doing more than a billion dollars in revenue and found the pattern those anecdotes kept hinting at: eight in ten had cut headcount in the name of AI, and it made no difference to their returns. The companies that fired the most showed nearly identical financial results to the ones that fired the least — and the ones that cut less sometimes did better. The firms actually extracting value from AI were doing something else entirely: using it to make their existing people more productive, not to replace them. Cutting people frees up budget room. It does not create return, and the study’s author warned that chasing value only through layoffs is the shortest-sighted move an executive can make. A confession is not a strategy. It is just a confession with a severance spreadsheet.
Ford’s Charles Poon — The Engineers AI Couldn’t Replace Had to Be Hired Back at a Premium
The cost stops being abstract the moment a company has to rebuild what it broke. Ford spent years trimming experienced engineers, then found itself rehiring hundreds of veterans to fix quality problems the automated systems never caught. The company’s head of vehicle hardware engineering admitted Ford had wrongly assumed it could feed the AI its design requirements and get a good product without seasoned humans watching the result. The engineers it had paid to leave had to be lured back — at market rates that had moved in their favor. Rehired workers are now commanding premiums of 20 to 35 percent over what they earned before the axe. The layoff did not end the relationship. It repriced it.
Commonwealth Bank of Australia — It Fired 45 “Redundant” Workers, Then Undid the Firing
Sometimes the whole collapse fits inside a single month at a single company. Australia’s biggest bank cut 45 customer service roles and credited an AI voice bot with shrinking its call volumes. The calls did not shrink; they climbed, and team leaders were pulled back onto the phones to cope. The bank reversed the cuts, conceding its assessment of which roles were redundant had been wrong — all in the same year it booked a record cash profit. The technology was not the failure. The conviction that a spreadsheet could replace judgment was.
Here is the human cost the boardroom math keeps refusing to see. Somewhere right now there is a person who was told their job was redundant, watched an AI take the title, and is now being called back — at a higher rate, with an apology, and with the quiet understanding that they were never redundant at all. They were the cheapest thing the company owned, until the company sold them off and had to buy them back at a markup. The people who ran the purge did not lose their jobs. The people who trusted the purge did. The layoff did not make the survivors more productive; it made them afraid, and frightened people do not teach a machine to do the work better.
What changes next is not that companies abandon AI. It is that they stop pretending the technology and the firing were the same decision. The companies that cut first and asked questions later are now paying twice — once to remove the capability, once to rebuild it. The ones that proved the technology before reshaping the team are getting the outcome everyone wanted in the first place: the machine doing the volume, and the humans doing the judgment. The AI did not fail. The layoff did. And the invoice for that mistake is only now arriving.
Sources: CNBC, IBTimes UK.