Klarna Said AI Did the Work of 700 People. Then It Hired Humans Back.

Transcript

700 full-time agents. That’s the workload said one AI assistant was carrying. Its CEO goes further, saying AI can do every job humans do. So, the company will simply stop hiring. In 24 months, the payroll falls from 5,527 people to 3,422 without a single mass layoff announcement.

Then 15 months after the boast, he sits down on Bloomberg TV and admits it. Cost had become too dominant, he says. And what you end up with is lower quality. So Klarna starts hiring humans back. All this from a company that had already lost roughly 85% of its value in one brutal year.

If the most aobsessed company on earth had to learn this the hard way, how many boardrooms are about to pay the same tuition?

A fast food kitchen in Sweden, friars, headsets, the smell of grilled beef. That is where Sebastian Siemiatkowski met one of the men he would later start a company with. He was young, born in Obsala, Sweden in October 1981. The son of parents who had moved there from Poland the year before, he would go on to the Stockholm School of Economics and take a master’s degree. But the job behind the counter mattered more than the diploma because of who was standing next to him.

Years later, that Burger King shift would be worth billions of dollars. Simatowski has said he credits a Swedish government policy for putting him on this path at all. Sweden subsidized home computer ownership and that is how his family got their first machine. His heroes were not bankers. They were Richard Branson and Ingvar Camprad.

The man who built IKEA out of a Swedish province and a flatpack box. Keep that in mind because the company he was about to build would borrow more from Ikea than from any bank. He started building it at 23 while taking a sbatical from the Stockholm School of Economics. He was working in sales offering accounts receivable services to small businesses. Boring work, invoices, chasing money, plumbing.

And in that plumbing, he noticed something that would not go away. Online payments broke again and again in the same places. Shoppers did not trust the internet with their card details and merchants did not trust shoppers. Somebody had to stand in the middle and carry the risk. In 2005 in Stockholm, Sweden, Siemiatkowski founded the company with Nicholas Adalbirth and Victor Jacobson.

They did not call it Klarna, they called it creditor and it grew out of the Stockholm School of Economics owned startup incubator SSC Business Lab. The idea was almost embarrassingly simple. let the shopper receive the goods first and pay afterwards without ever typing in a card number. Creditor would process the payment and take on part of the merchants’s credit and fraud risk. In other words, three students in their 20s offered to stand behind strangers they had never met. That is a beautiful pitch and a terrifying balance sheet because someone has to fund the gap between delivery and payment and someone has to eat the loss when the money never arrives.

They needed capital and they needed people who could actually build the thing. They got both from one person. Jane Walerud, an angel investor, saw the idea presented and offered them 600,000 Swedish crona for 10% of the company. That values the whole business at 6 million Swedish crona. Three founders, one product, a rounding error next to what came later.

The money was not the important part. Walerud connected them with developers who helped build the platform itself. Those developers brought a programming language called Erlang into Klarna’s early technical architecture. A choice tied to Walerud’s own background at Erlang systems. Erlang was built for systems that must not fall over.

Whail later put it bluntly in the Swedish press. Without Erlang, no Klarna, the code they wrote in a Stockholm office would one day carry millions of transactions a day. The model worked because it solved both sides of the same fear at once. The shopper paid after the box arrived. The merchant got paid either way.

Sweden first, then Norway, Finland, and Denmark, then Germany, and the Netherlands. In 2007, the Swedish investment company Investment Obison put money in. And in 2009, creditor changed its name to the one you know, Klarna. But a business built on handing strangers goods before they pay has one obvious question buried inside it. What happens when you scale that to millions of people and billions of dollars?

In

December 2011, a Swedish payments company almost nobody outside the Nordics had heard of announced a number that made the tech press sit up. General Atlantic led an investment round of 155 million US dollars in Klarna joined by DST Global with the money earmarked to push the company beyond its home region. That same year, the Daily Telegraph put Klarna on its list of Europe’s promising young technology companies. For a firm that had started in a Stockholm business school incubator, that was a strange kind of arrival, and it was not the first outside bet. The year before, Sequoia Capital had invested in Klarna.

One of the most famous venture firms in the world had flown to Sweden for an invoice business. The money did what money does. Klarna had already spread from Sweden into Norway, Finland, and Denmark, then into Germany, and the Netherlands. In 2013, it bought the German online payments company sofag, the operator behind software tuberwa, and the combined business became Klarna group. A year earlier, it had launched Klarna checkout, an embedded checkout that merchants could drop into their online stores.

That product would go on to hold roughly 40% of the market in Sweden and over 20% across the Nordic region. in its home market. Then Klarna wasn’t a payment option. It was close to being the checkout itself. Then came the market that would decide everything. Klarna launched in the United States in 2015 and it became one of the company’s main growth markets.

That same year, Sweden’s then Minister for Enterprise and Innovation, Mikael Damberg, listed Klarna as one of the country’s five unicorn technology companies alongside Spotify, Majang, Skype, and King. Think about that company Sebastian Siemiatkowski was suddenly keeping. Skype had changed how the world made phone calls. Klarna had changed the order in which you paid for a pair of shoes. But a payments company is not a bank and seekowski wanted the real thing. In June 2017, Sweden’s financial regulator finance inspection granted Klarna a full banking license.

That license meant Klarna could operate across the European economic area under EU banking rules and later reach outside it into the United States and the United Kingdom through local subsidiaries. The next year it bought Clo Brothers Retail Finance in the United Kingdom, strengthening its position in British retail finance. And in 2019, Klarna raised $460 million US at a valuation of $5.5 billion, which made it one of Europe’s highest valued fintech companies at the time. 5.5 billion. Remember that figure because within 2 years, it would look almost quaint.

In 2020, Ant Financial, the financial affiliate of China’s Alibaba Group, bought a minority stake in Klarna as part of a partnership. Then the pandemic boom hit online shopping and by now pay later went from a Nordic habit to a global one. In June 2021, a funding round led by Soft Bank’s vision fund two valued Klarna at about 46 billion US that made it one of the most valuable private fintech companies in the world. For Seatcowski personally, the arithmetic was just as dramatic. He had been minted a billionaire that year as the valuation climbed and by 2022 his net worth was estimated at $3.2 billion.

Not bad for a man who met one of his co-founders while working at Burger King. Flushed with cash, Klarna started buying. In July 2021 alone, it acquired APRL, a Stockholm influencer marketing platform and Stockcard, a mobile wallet app for loyalty cards. It launched a physical Klarna card in the UK in January 2022, then a card in the United States supported by Marquetta. It completed the purchase of PriceRunner, the Swedish price comparison service to bolt shopping discovery and reviews onto the payments network.

That last deal looked like a sideshow at the time. Keep it in mind because years later, Pricerunner would drag one of the largest companies on Earth into a Swedish courtroom. So here was Klarna at its peak. A licensed bank, a shopping app, a card, a price comparison engine, and a valuation of $46 billion built on the promise that the growth would keep coming. All of it rested on cheap money and rising consumer spending.

Both were about to go. What happens to a $46 billion company when interest rates rise and the money stops?

It is May 2022 and Sebastian Siemiatkowski has a number to deliver to his own staff. Klarna will cut about 10% of its workforce. At the time the company employs roughly 7,000 people, so that is around 700 jobs. The reason he gives is blunt. Business sentiment has turned.

Inflation is climbing and there is a war in Ukraine. For 17 years, the story had only gone one way. Now it went the other way. Then came the decision that would follow him for years. Siemiatkowski published a list of affected employees on LinkedIn.

The idea, as Klarna described it, was voluntary. Departing staff were asked whether they wanted to share their availability on platforms like LinkedIn, and those who said yes had their details posted by the company and by the chief executive himself. Some people called it generous. Plenty of others called it insensitive. One report counted 560 names on the list.

A layoff had become a public document. In Stockholm, Sweden, the reaction from organized labor was colder still. The Swedish Financial Sector Union said it had learned about the job cuts from the media rather than being consulted first, and it sued the company over the layoffs. That fight did not end quickly. The following year, the union’s union in Invaria’s in Genura announced strike action at Klarna’s Stockholm headquarters.

The strike was called off only after Klarna agreed to join the financial sector employers organization and accept a collective bargaining agreement from the start of 2024. The company that liked to call itself a disruptor had just been pulled back into the oldest Swedish institution of all. But the layoffs were only the human half of the story. That July, with technology valuations falling and interest rates rising, Klarna raised $800 million. The valuation attached to that money was $6.7 billion.

12 months earlier, Soft Bank’s vision fund 2 had led a round valuing Klarna at about $46 billion. That is a drop of roughly 85% in a single year. Think about what that means in raw arithmetic. Close to $40 billion of paper value gone in the time it takes to run one financial year. The founders’s own fortune moved with it.

In 2021, when Klarna’s valuation hit $31 billion, Siemiatkowski was minted as a billionaire. In 2022, his net worth was estimated at $3.2 billion. By 2025, he was no longer a billionaire at all. He had built one of Sweden’s five unicorns, named alongside Spotify and Skype by the country’s own enterprise minister, Mikael Damberg. And now the market had repriced the whole thing.

So what does a chief executive do when the cheap money is gone and the headcount is already cut? Klarna had to keep growing because growth was the entire promise. It still had to serve tens of millions of consumers in dozens of markets across more than 30 languages with fewer people answering the phone. Siemiatkowski’s answer came from a conversation he had in 2023 when he told OpenAI’s Sam Altman that he wanted Klarna to be ChatGPT’s favorite guinea pig. He was not joking.

Within months, Klarna stopped hiring altogether. If you can never add another employee, where on earth does the next wave of growth come from? In 2023, Sebastian

Siemiatkowski made an offer to Sam Altman, the head of OpenAI. He said he wanted Klarna to be ChatGPT’s favorite guinea pig. Think about that phrase for a second. Not a customer, not a partner, a test subject. The chief executive of a Swedish bank regulated by the Swedish Financial Supervisory Authority volunteering his own company as the laboratory animal for a technology nobody had fully mapped yet, and he meant it.

On the 27th of February 2024, Klarna put out a press release from New York in the United States. The AI assistant powered by OpenAI had been live around the world for exactly 1 month. In those 30 days, it had handled 2.3 million conversations. That is 2/3 of every customer service chat the company received. Do the arithmetic and it comes to more than 75,000 conversations a day, every day, including weekends.

It worked in 23 markets and spoke more than 35 languages, and it never slept. Then came the line that made the whole thing famous. Klarna said the assistant was doing the equivalent work of 700 full-time agents, 700 people, one piece of software. The company said customer satisfaction scores were on par with human agents, that errand resolution was more accurate, and that repeat inquiries had dropped by 25%. The average customer used to wait 11 minutes to get a problem solved.

Now it was under two. Klarna estimated the whole thing would add about $40 million US to its profit in 2024. Brad Lcap, the chief operating officer of OpenAI, said Klarna was at the very forefront among its partners in AI adoption. Siemiatkowski framed it as a win for everyone. Better experiences for customers, better prices, more interesting work for staff, better returns for investors.

But in the same statement, he added something stranger and heavier. He said the launch underscored the profound impact AI would have on society. He urged politicians to think carefully and called for considerate, informed, and steady stewardship through the transformation. The man selling the machine was also warning you about it. That press release traveled further than any product Klarna had ever shipped.

Here was a company serving 150 million consumers and 2 million transactions a day, saying out loud what every board in the world was whispering. The number 700 became a headline in every language. And the timing was perfect for Klarna because behind the scenes, the company was preparing for something enormous. It had confidentially started work on a US stock market listing in 2024. A fintech that could grow revenue while shrinking its payroll was exactly the story Wall Street wanted to hear.

So Klarna leaned in harder. It ripped out Salesforce and Workday and stopped using around 1,200 smaller software services because Siemiatkowski wanted all the company’s data consolidated in one place where AI could actually read it. By his own account, the workforce that had stood at about 5,5002 years earlier was heading toward roughly 3,000. Later, he would put the fall even more bluntly from 7,400 people down to 3,000. He was not hiding it, he was advertising it.

He even appeared on an earnings call as an AI generated video replica of himself, speaking in his own voice and likeness. For a while, the story had no second half. The chatbot was fast, cheap, multilingual, and tireless, and everybody quoted the 700. But speed is not the only thing a customer wants when a payment goes wrong and a debt letter arrives. Somewhere in those millions of chats, people were trying to reach a human being.

And what Siemiatkowski would admit about that roughly a year later would undo the most famous claim his company ever made. What did those customers find on the other end of the line? And why would the man who built it soon call the result lower quality? December 2024.

Sebastian Siemiatkowski is in front of a camera again and he is not hedging. Klarna would not be adding headcount, he said, because AI can do all of the jobs that we humans do and the only question left was how you apply it. No caveats, no soft landing. It was the sort of line a chief executive can never take back. And he said it out loud on the record as the boss of a Swedish bank with a banking license.

Behind the sound bite sat a quieter decision that had already been running for more than a year. Klarna had simply stopped hiring in 2023. At the same time, it pushed AI deeper into its own ranks. That was the trick, and it was almost boring in its mechanics. You do not need to fire people if you refuse to replace the ones who walk out the door.

Siemiatkowski put the natural attrition at a company like his at 15 to 20% a year. We have simply communicated to our employees that what we’re going to do is we’re going to shrink. So, we’re going to stop hiring. He said shrinking as a strategy announced to the staff in advance. The numbers in Klarna’s own IPO perspectus told the story cleanly.

At the end of December 2022, Klarna had 5,527 full-time employees. Two years later, 3,422. That is more than 2,000 people gone from the payroll in 24 months. Roughly 100 a month, month after month, without a single mass layoff announcement. The company credited the drop to its use of AI and to lowering overall headcount and said it expected the number to keep falling, not stabilize, keep falling.

Then came the part most people missed and it was the part that actually mattered. Klarna ripped out Salesforce and Workday, two of the biggest software systems a modern company runs on. Siemiatkowski said the reason was data. To learn anything about one client, Klarna’s people had to dig through Google Suite, Slack, Workday, Salesforce, and more. So, the company cut around 1,200 smaller software services as well to pull its data into one place where AI could actually read it.

1,200 contracts unwound. You cannot feed a machine an organization that is scattered across a thousand login. And he went further than the systems. He went for the job at the top. In May, Simakowski appeared on an earnings call not as himself, but as an AI generated video replica, delivering scripted remarks in his own voice and likeness.

Klarna also ran a 24/7 AI CEO hotline, answering customer questions in his conversational style in English and in Swedish. The message to investors was unmistakable. If the founders’s face can be automated, nobody’s can. He was living it at home, too. Telling Bloommerg he used the technology all the time and that his wife complained because once the kids were asleep, he wanted to go and vibe code.

By then, Klarna had become the world’s favorite exhibit in the argument about AI and work. And Siemiatkowski was not trying to soften it. I feel a lot of my tech bros are being slightly not to the point on this topic, he told Bloomberg, warning of a massive shift coming to knowledge work in banking and in society at large. He pointed at Brussels in Belgium where thousands of people still work as translators, a job he said AI could already largely do. New jobs would come, he admitted.

But that does not help the Brussels translator who is not going to become a YouTube influencer tomorrow. He had made himself the loudest proof on earth that machines replace people. There was just one group he had not asked. What happens to the man who automated his own face when his own customers start refusing to talk to the machine? Bloomberg TV May 2025.

Sebastian Siemiatkowski sits down and says something no tech chief executive likes saying out loud. Cost, he admits, had become too dominant a factor in how Klarna organized its customer support. And what you end up with, he says, is lower quality. Really investing in the quality of human support, he adds, is the way of the future. 15 months earlier, the same man had celebrated a chatbot doing the work of 700 full-time agents.

Now he was hiring people back. The reversal was not a press release. It arrived in pieces through a spokesperson. Klarna’s Claire Nordstrom told the trade outlet CX Dive that customers should always have the option of speaking to a human. AI gives us speed, she said, and talent gives us empathy.

The company framed it as evolution rather than retreat. The chatbot still handled 2/3 of all inquiries. Response times had improved by 82% since launch. Repeat issues were down 25%. All true.

But a year earlier, the story had been that the machine was the answer. And now the machine was the easy stuff. So what exactly was Klarna hiring, not a call center? Siemiatkowski described in Yubertype setup a pilot program, remote work, competitive pay, full flexibility. Nordstrom said Klarna wanted highly educated students, professionals and entrepreneurs in a role that mixed frontline service with realtime product feedback.

And the target was bigger than it looked. One of the goals, Siemiatkowski told Bloomberg, was to replace the thousands of workers Klarna was outsourcing. Read that again. The company famous for letting AI replace humans was now planning to let humans replace an outsourcing bill. In June, he took the argument on stage in London in the United Kingdom at SXSW.

My wife taught me something. He told the crowd, “Two things can be true at the same time. Yes, Klarna had stopped hiring and rolled out AI agents that cut support costs and lifted revenue per employee. Yes, the workforce had fallen from 5,500 2 years earlier to about 3,000. That is roughly 2,500 people gone in 24 months. more than a hundred a month every month.

And yes, he insisted there was still an opportunity for humans at his company. Then came the line that reframed the whole strategy. Human customer service, Siemiatkowski said, is always going to be a VIP thing. He compared it to clothing stitched by hand instead of by machine. People pay more for that.

So, Klarna would use AI to take away the boring manual work and promise customers a human connection on top, not a climb dome, a luxury tier. Outside analysts were less poetic about it. Julie Jeller of Infoch Research Group said Klarna had embraced AI with an eye on cost savings and perhaps underestimated the tradeoff. As customers voiced frustration about impersonal service and limited access to human help, she said the approach risked undermining the very experience it was meant to improve. That friction, she added, carries real financial and reputational consequences. The numbers backed her up.

A varant survey found more than twothirds of customers had suffered a bad chatbot experience with the top complaint being that it could not answer the question. A 59 survey in March found 86% of customers rated empathy and human connection above a quick response. Siemiatkowski kept living inside the technology. Anyway, he told the London audience he was using chat GPT as a private tutor, dropping Slack threads into it to check whether the engineering made sense. He said Klarna had shut down around 1,200 small software services to consolidate its data.

And when someone asked about the initial public offering, he gave nothing away beyond a smile and a remark that he was happy there was less turbulence in the market. 3 months later, he would be standing on Wall Street. Two things can be true at once, he said. But which one would be true for the investors about to buy his stock? April 2025 had not gone

to plan. Klarna had filed its public registration statement in March. The IPO was ready. And then President Donald Trump’s sweeping tariff announcement tour through the markets in early April. Klarna pulled back.

So did the ticket marketplace StubHub and the stock trading app E Toro. The company that had been preparing this moment for a year simply stopped and waited and waited. Then on the 10th of September, Klarna’s ordinary shares started trading on the New York Stock Exchange in the United States under the ticker KL. The offering was $ 34,311,274 shares at $40 each. That price was above the expected range and it raised 1.37 billion US for the company and its existing shareholders.

Look closely at who was selling though because that is the real story of the day. Klarna itself sold 5 million shares. Everything else, 28.8 8 million shares came from existing shareholders cashing out. Klarna, the company, walked away with $222 million. The people who had backed it walked away with nearly 1.2 billion.

The stock opened at 52. Then it slid through the day and closed at 45.82, up about 15%. And that put the company at roughly $17.3 billion. It was the biggest IPO of the year on the New York Stock Exchange, but hold that valuation next to the one from 4 years earlier. In June 2021, a round led by Soft Bank’s Vision Fund 2 had valued Klarna at about $46 billion US.

So, Klarna went public worth a little over a third of what a Japanese investment fund had paid for a slice of it in a private round. Soft Bank had watched the value of its stake plunge. For one investor, the arithmetic ran the other way. Sequoia Capital first backed Klarna in 2010 and had put in $500 million US in total. It sold just 2 million of its 79 million shares in the offering.

At the offer price, that meant an overall return of about $2.65 billion US. Andrew Reed, a partner at the firm, told CNBC he had still been in college when Seoia first invested in what he called an alternative payments company in Stockholm. 15 years later, he was standing in New York, describing a business with over 100 million consumers, over 100 billion US of volume and close to a million merchants. Staggering, he called it. Sebastian Siemiatkowski was on CNBC that day and he did not sound like a man at a party.

To me, it really just is a milestone, he said. Then he reached for the comparison that stuck. It’s a little like a wedding, he said. You prepare, you plan, you throw a big party, but in the end, the marriage goes on. And the marriage he meant was with public shareholders, people who would now check his numbers every 3 months.

So what did he actually sold them? Not just buy now, pay later. Klarna had spent recent months talking up its move into banking with a debit card and personal deposit accounts in the United States. Simacowski said 700,000 people in the US had signed up for the card and 5 million more were sitting on a waiting list. He drew a line between his card and the one from rival a firm which had attracted 2 million users since 2021 saying Klarna was pulling a slightly different audience and there was a shadow over all of it.

In the United Kingdom, the government had proposed new rules to bring by now pay later loans under formal oversight over affordability concerns. There was one more thing in the pitch. The thing every analyst wanted to hear about the headcount story. By October, Siemiatkowski was telling Bloomberg that Klarna’s workforce had gone from 7,400 to about 3,000 and pointing at 38% year-on-year revenue growth in the United States. AI had cut the staff.

The cost line had fallen and the market had bought the story. But Klarna had already quietly started hiring humans back. Klarna had sold Wall Street an AI efficiency story. But was anyone checking whether the maths actually worked? Start in Sydney, Australia.

Inside a call center belonging to the Commonwealth Bank of Australia, 45 customer service roles were declared redundant, replaced by an AI powered voice bot. Then the phone started ringing more, not less. Managers offered overtime, and team leaders were pulled in to answer calls themselves. Weeks later, the bank admitted the roles were not redundant at all. Its own statement said the initial assessment did not adequately consider all relevant business considerations and the bank apologized to the workers.

The finance sector union called it a massive win and accused the bank of dressing up job cuts as innovation. So this was not just a Klarna problem. It was a pattern. At IBM, artificial intelligence took over human resources functions and handled roughly 94% of routine requests. The other 6% is where it broke.

Those were the ethical dilemmas, the cases that needed a person. IBM then announced it would triple its entry-level hiring in the United States across all business units in 2026. Its chief human resources officer, Nicl Lamorox, put the reasoning plainly at a summit in New York. If you stop hiring juniors, in 3 to 5 years, there is no pipeline left and the well simply dries up. Ford went the same way, reportedly bringing back hundreds of experienced engineers to fix quality problems the automated systems could not solve. Not everyone blinked.

At Salesforce, chief executive Mark Binoff said on a podcast that he had taken customer support from 9,000 heads down to about 5,000. His words were blunt. He needed fewer heads. That is 4,000 customer support jobs gone from one company. Salesforce said the number of support cases it handled had declined thanks to its agent force bots.

So, it no longer backfilled support engineer roles. Bey had already said over the summer that AI was doing up to half the work at the company, but the analyst Ed Zitron offered a colder reading that AI was being blamed by companies that overhired during the pandemic and now wanted to look efficient for investors. Growth at all costs, he said, even if it makes the product worse. And then the arithmetic arrived. The research firm Gartner predicted that the cost per resolution for generative AI would pass $3 US by 2030.

That is higher than many offshore human agents cost today. The reasons were mundane and brutal. Rising data center costs, AI vendors switching from subsidized growth to chasing profit and use cases that got more complicated the deeper you went. Gartner’s Patrick Quinland said the return on AI investment was far from guaranteed and that full automation would be prohibitively expensive for most organizations. One of his colleagues, Emily Paskki, framed the trade in one devastating line.

Historically, she said, companies bought cheaper technology to replace expensive talent. Now they were buying more expensive technology to replace a cheaper talent source. and she did not know if that would ever pay off. The regret was already showing up in the data. A survey by Orgview found that 39% of business leaders had made staff redundant because of an AI deployment. Of those, 55% admitted the wrong decisions had been made.

Separate figures from Robert Half found that 32% of US hiring managers had eliminated a role primarily because of AI and then rehired for the same or a similar job. That is roughly one in three. One consultancy report put the failure in a sentence. Companies budgeted for technology to replace humans without investing in training and ended up cutting the very people needed to supervise the machines. Which brings us back to Stockholm, Sweden.

Klarna’s founder, Sebastian Siemiatkowski, had said the quiet part loud a year earlier, accusing his fellow tech chief executives of being slightly not to the point about what was coming for knowledge work. He pointed at the thousands of translators still employed in Brussels, Belgium. Work he said AI could already largely do. His warning was that new jobs would come, but not in time to help that translator who was not going to become a YouTube influencer tomorrow. He had also admitted the other half of the truth.

Cost had been too dominant a factor, and the result was lower quality. Klarna had cut its workforce from 7,400 to about 3,000 people, and by the autumn of 2025, its market value sat just over$15 billion US. If the machines were never actually cheaper, then what exactly had all of these companies bought? On a Thursday in the middle of

May 2026, Klarna pushed out a press release with a plain headline and a very loaded number. revenue of$1 billion US in a single quarter, adjusted operating profit of 68 million US, operating income of $17 million, and net income of exactly 1 million. For a company that had spent four years being described as a cautionary tale, that last figure mattered more than its size. It was Black Ink. Underneath sat the machine that produced it. Klarna moved $ 33.7 billion US of goods through its network in those three months.

Across the group, it now counted more than9 million active consumers, more than a million merchants, and 3.4 million transactions every single day in 26 countries. The Klarna card, the product Sebastian Seamacowski had been selling to Wall Street on listing day, had reached 5 million active users across 16 markets. Remember the waiting list he bragged about in New York? It had turned into customers. But the number that told the real story of this documentary was buried further down the release.

Revenue per employee nearly 1.4 million US. Think about what that means in a building. Klarna had roughly 3,000 people down from 5,527 at the end of 2022. And each one of them was now carrying more revenue than most fintech startups generate in total. That is what the AI years actually bought.

Not a company without humans, but a company where the humans were worth multiples of what they used to be. The credit side held two, which is the part that kills lenders like this. Klarna said it had underwritten about half a trillion US dollars of transactions over 20 years and that its provisions for credit losses came in at 0.55% of volume in the first quarter of 2026. A year earlier, the same figure was 0.54%. In other words, almost nothing moved.

For a business built on lending small sums to millions of strangers, a flat loss rate is the most boring and most valuable sentence in the report. Then came the summer and the windfall nobody had priced in. In July 2026, a Swedish court ordered Google to pay Klarna 1.97 billion US in damages after Klarna won an antirust case. The claim came from Pricerunner, the Swedish price comparison service Klarna had bought back in 2022, which accused Google of favoring its own comparison shopping service in search results. Look at the scale of that award against the operating business.

Klarna’s entire adjusted operating profit for the 2025 financial year was $65 million. One court ruling was worth about 30 times that. That same month, Klarna filed to build the thing it had been circling since 2005. It applied to establish an FDICsured bank in Utah in the United States, which would give it its own American banking subsidiary. A charter would let it expand traditional banking products and pull lending and merchant operations in house, cutting its dependence on third-party banking partners.

And in the same month, Apple chose Klarna as the financial backer and leasing provider for its new Apple upgrade device program. The little Swedish invoice company was now the balance sheet behind the iPhone. So, what do you take from 21 years of this? The founder who told the world that artificial intelligence could do every job humans do ended up running a company that promised every customer a human whenever they wanted one. He called it a VIP product, like clothing stitched by hand instead of by machine.

The automation was real, the savings were real, and the mistake was real, too. Klarna survived not by deciding what machines could do, but by working out expensively and in public what they could not. If a company this obsessed with AI had to learn that lesson the hard way, how many boardrooms are about to pay the same tuition? So let’s settle

the three questions this story opened with. The 700 agent number was real in the narrow sense that Klarna said it was. In its first month, the OpenAI powered assistant handled 2.3 million conversations, roughly 2/3 of all customer service chats, doing work the company valued at 700 full-time agents. And that share held. A year later, the chatbot was still handling twothirds of inquiries with response times improved by 82% and repeat issues down 25%.

But the man who sold that story is also the man who unsold it. Cost, Sebastian Seamacowski admitted, had been too dominant a factor. And what you end up with is lower quality. By 2025, Klarna was recruiting humans again, pitching human service as a VIP product, like clothing stitched by hand instead of by machine. Was AI actually cheaper, less cleanly than the headlines imply?

Gartner now forecasts that generative AI will cost more than $3 per resolution by 2030, higher than many offshore agents, and warns that full automation will be prohibitively expensive for most organizations. The corporate wreckage agrees. Commonwealth Bank cut 45 customer service roles for a voice bot, then admitted the assessment was an error and reversed it. IBM automated 94% of routine HR requests couldn’t handle the other 6% and is now tripling US entry-level hiring. And 32% of US hiring managers say they eliminated a role because of AI, then rehired for the same job.

Which leaves the hardest question. Was it worth it? Klarna went public in September 2025 at about $17.3 billion, a fraction of the $46 billion Soft Bank paid for in 2021. But by the first quarter of 2026, it was posting $1 billion in quarterly revenue, $68 million in adjusted operating profit, and nearly $1.4 million of revenue per employee. A Swedish court ordered Google to pay it $1.97 billion.

It applied for a US bank charter in Utah and signed Apple as a leasing partner. So, the verdict is uncomfortable for everyone. AI did not do all the jobs as its CEO once claimed it could. And the company that pretended it would steal one. It just paid the tuition in public where everyone could watch.