The Algorithm That Broke
The Algorithm That Broke
When a chatbot meant to replace 700 humans starts failing at complex disputes, a fintech darling’s valuation plunges 65 percent.
[Speaker 1]: If you bought Klarna stock at the IPO in September, you have currently lost nearly sixty-five percent of your money. [Speaker 2]: And if you’re looking at the chart this morning, it’s ugly. The stock is trading between thirteen and fourteen dollars. [Speaker 1]: Right. Down from a listing price of forty. The collapse happened fast. It started last Thursday, when the Q4 earnings report dropped. Revenue was actually up, but two other numbers absolutely terrified Wall Street. First, a massive, unexpected provision for bad debt. [Speaker 2]: And second, the admission that the company is quietly returning to human agents to do the job that AI was supposed to have already solved. [Speaker 1]: Today, we’re not just looking at a bad earnings report. We are investigating whether this is the market’s first true "AI Crash"-where an algorithm didn’t just fail, it actually broke the business model. [Speaker 2]: We’re going to look at the lawsuit alleging that Klarna used "AI washing" to inflate its IPO price. [Speaker 1]: And we’ll look at the number "700." That is how many human agents the CEO claimed he replaced with AI two years ago. [Speaker 2]: A claim that looks very different today, now that internal engineering teams are being asked to answer the phones. [Speaker 1]: It’s Monday, February 23, 2026, and you’re listening to The Angle. [Speaker 2]: To understand why this stock chart looks like a cliff edge, we have to go back to early 2024. This is when the narrative was built. [Speaker 1]: The "AI Efficiency" narrative. [Speaker 2]: Exactly. In early 2024, Klarna CEO Sebastian Siemiatkowski made a very loud, very specific claim. He said their new AI customer service chatbot was doing the work of seven hundred full-time agents. [Speaker 1]: And that wasn't just a tech flex. That was a financial pitch. [Speaker 2]: It was. The company froze hiring. They cut staff. The pitch to investors was essentially: "We are not a bank. Banks have heavy overhead and branches and call centers. We are a tech company. We have tech margins." [Speaker 1]: And the market bought it. Hook, line, and sinker. They IPO on the NYSE on September 10, 2025. They list at forty dollars a share. It opens at fifty-two. The valuation hits fifteen billion. [Speaker 2]: For a few months, it looked like the future. But behind the scenes, late last year, the metrics started to slide. We don't have the internal emails yet-those will likely come out in the class-action discovery-but reports suggest the AI started struggling with "high-stakes" interactions. [Speaker 1]: Meaning complex disputes. [Speaker 2]: Right. If you want to know my balance? The bot is fine. If you want to dispute a charge because a merchant sent you a counterfeit watch? The bot gets confused. It loops. It creates frustration. Customer churn started ticking up. [Speaker 1]: So fast forward to last Thursday, February 19. The Q4 earnings call. They have to admit two things. One, they "focused too much on efficiency"-which is corporate speak for "we fired too many people"-and they are rehiring humans. [Speaker 2]: And then, the bomb drops. In the middle of the report, there’s a line item for "Provision for Credit Losses." [Speaker 1]: Okay, pause on that. Because this is the mechanic that actually sank the stock. Explain what that provision actually represents. [Speaker 2]: Think of a "Provision for Credit Losses" as a rainy day fund, but the rain is already falling. It’s money the company moves from the "profit" column to a "reserve" column because…