Inside Klarna's Systematic Withdrawal from Human Customer Service
In February 2025, Klarna CEO Sebastian Siemiatkowski logged onto X and announced that the company's AI assistant had handled 2.3 million customer service conversations in its first month of operation — roughly two-thirds of all customer contacts. He framed it as a productivity story. The workforce data told a different one.
Between January 2024 and March 2025, Klarna reduced its total headcount from approximately 5,000 employees to around 3,800, a reduction of more than 24 percent, according to figures the company disclosed to reporters and subsequently cited in its IPO prospectus filed with the SEC in July 2025. Customer service roles bore the largest share of that contraction. Siemiatkowski told the Wall Street Journal in an interview published in August 2024 that the company had stopped hiring in functions it expected AI to absorb, and had allowed attrition to run without replacement. He used the phrase "natural reduction."
What happened at Klarna over those fourteen months is now one of the most thoroughly documented cases of AI-driven workforce contraction in a consumer-facing company — not because it was hidden, but because Klarna's leadership narrated it in real time, apparently calculating that transparency served the company's valuation better than discretion.
The Mechanism Behind the Numbers
The AI assistant Klarna deployed in January 2024 was built on OpenAI technology under a partnership announced publicly by both companies. Klarna said at the time that the assistant was performing work equivalent to 700 full-time customer service agents. That figure — 700 — is the one that circulated most widely in press coverage, including in reports by Reuters, the Financial Times, and Bloomberg. What received less attention was the underlying methodology: Klarna calculated the equivalent by dividing the volume of resolved conversations by its historical per-agent resolution rate. It was not claiming 700 people had been let go in a single action. It was describing what 700 people's annual output had been redirected toward a system that cost, by the company's own account, a fraction of equivalent human labor.
The distinction matters because it shapes how the displacement actually unfolded. Workers were not assembled and told their roles had been automated. Instead, teams shrank through a managed attrition strategy that Siemiatkowski described to the Wall Street Journal as deliberate: the company set a hiring freeze in AI-exposed roles and let the headcount drift downward as people left voluntarily. By the time Klarna filed its IPO prospectus, the company was presenting the workforce reduction not as a cost it had incurred but as a margin improvement it had engineered.
The prospectus disclosed that Klarna's operating expenses had fallen materially even as revenue grew. The company reported revenue of approximately $2.8 billion for the twelve months ending March 2025, alongside a return to operating profitability after years of losses. Goldman Sachs, one of the banks underwriting the IPO, highlighted the cost structure improvement in analyst notes previewing the offering. The word "AI" appeared more than 100 times in the prospectus, according to a count by the Financial Times.
What the prospectus did not detail was what happened to the workers whose attrition powered the margin improvement. Klarna employed a large share of its customer service workforce through third-party contractors and outsourcing arrangements, particularly in markets including Germany, the Netherlands, and the United States. Those workers, not on Klarna's direct payroll, do not appear in its reported headcount figures. Researchers at Oxford's Internet Institute who have studied platform labor displacement have noted this structural feature as a consistent pattern in AI-driven service automation: the workers most exposed tend to be the least visible in corporate reporting.
What the Company Said, and What It Left Out
Siemiatkowski gave a notable interview to the Swedish newspaper Dagens Industri in September 2024, in which he was asked directly whether Klarna was using AI to eliminate jobs. His answer was careful. He said he believed AI would create new categories of work and that Klarna intended to hire in engineering and product roles. He acknowledged the workforce was smaller. He did not dispute that customer service headcount had fallen because of automation.
In a separate conversation at a fintech conference in London in October 2024, he went further, saying that companies which did not pursue this kind of efficiency would be outcompeted by those that did — a framing that positioned the workforce reduction as a competitive necessity rather than a choice. That framing was picked up by Bloomberg's coverage of the event.
The company's public communications consistently emphasized what the AI system could do rather than what the workforce reduction meant for the people affected. There was no disclosure of severance arrangements, no accounting of contractor displacement, and no statement from Klarna about retraining support for affected workers. A company spokesperson, responding to a question from Reuters in early 2025 about whether displaced workers had received transition assistance, said Klarna did not comment on individual employment matters.
This gap — between the precision with which Klarna quantified its AI productivity gains and the near-total absence of information about affected workers — is what researchers studying the displacement have highlighted as the defining feature of the case. The McKinsey Global Institute, in its June 2025 update to its automation displacement series, cited the Klarna case as an example of what it called "asymmetric disclosure": detailed reporting on operational efficiency paired with minimal disclosure on workforce impact.
The Stanford AI Index 2025, published in April of that year, included Klarna in its analysis of enterprise AI deployment at scale, noting that the company had become a reference point in corporate presentations about return on investment from large language model deployment. The index did not independently verify Klarna's 700-agent-equivalent claim but noted its prominence in industry discourse.
Aria Chen, who ran the software engineering team at a mid-sized logistics firm before her own displacement in 2024 and now publishes AI displacement research at clankers.today, reviewed the Klarna IPO prospectus and the available press record for a briefing published in August 2025. Her assessment was characteristically direct: the Klarna case demonstrated that managed attrition, because it produces no single layoff event to report, generates almost no sustained press coverage of its human cost, even when the aggregate numbers are substantial.
Klarna's IPO ultimately priced in July 2025 at a valuation of approximately $15 billion, below its 2021 peak of $45.6 billion but representing a significant recovery from its 2022 down round. The workforce reduction, narrated as an AI efficiency story, was a material part of the financial case presented to investors. The 1,200 or more workers whose roles disappeared through attrition over those fourteen months did not have a comparable venue in which to present their own account.
PROTECT YOUR INCOME NOW
If AI automation threatens your job, preparation is the only honest answer: know your risk, build a runway, and reskill before you have to. The Clankers 90-day programme structures exactly that — free, which is exactly why today is the right time to start.
Start the 90-Day Programme — Free →GET THE CLANKERS BRIEF — FREE
One email every Monday: the week's AI displacement moves that matter, and what to do about them. No spam. Unsubscribe anytime.