Swedish fintech giant Klarna has become the most prominent — and most instructive — case study of what happens when a financial services company bets aggressively on replacing human workers with artificial intelligence. The buy-now-pay-later pioneer has cut its workforce from approximately 5,500 employees in 2022 to around 3,000 today, with CEO Sebastian Siemiatkowski projecting further reductions to roughly 2,000 by 2030. But the path from headline-grabbing AI triumph to quiet human rehiring tells a more complicated story than most coverage has captured.
For anyone building or operating financial infrastructure — open banking platforms, BaaS providers, payment processors, accounting integrations — the Klarna experiment is not just a news story. It is the closest thing the industry has to a controlled test of what AI-first workforce strategies actually produce when they meet the reality of regulated financial services at scale.
The Timeline: From Hiring Freeze to AI Replacement to Rehiring
Klarna’s AI journey began in earnest in 2023, when the company became one of the first major financial services firms to adopt the enterprise version of ChatGPT across its operations. At the time, the company had approximately 5,000 employees. Rather than laying off staff directly, Klarna implemented a hiring freeze and allowed natural attrition to reduce headcount while AI tools absorbed the departing employees’ workload.
By February 2024, the company announced that its AI assistant — built in partnership with OpenAI — was handling approximately two-thirds of all customer service interactions, equivalent to the work of 700 full-time agents. Resolution times reportedly dropped from an average of 11 minutes to under two minutes. Internally, Klarna developed Kiki, a generative AI assistant designed to help remaining employees with document analysis, contract drafting, and workflow optimisation. The company reported that 96% of staff were using AI tools in their daily work.
The financial results appeared to validate the strategy. Klarna’s third-quarter 2025 revenue reached $903 million, a 28% year-on-year increase, while the company reported a 152% increase in revenue per employee since early 2023. Operating costs remained flat even as revenue grew, and the savings from reduced headcount allowed Klarna to raise average salaries for remaining employees by approximately 60%, from around $126,000 to $203,000.
Siemiatkowski was explicit about the direction of travel. In public appearances and investor communications, he stated that AI could already perform all human jobs at the company and predicted that Klarna’s headcount would eventually fall to around 2,000. The company was held up across the fintech industry — and the technology sector more broadly — as proof that AI-driven workforce replacement was not only possible but already delivering results.
What Went Wrong: The Quality Problem That Metrics Missed
By mid-2025, a different picture had emerged. Customer complaints about the quality of AI-driven support had been accumulating. The issues were not with the volume of interactions handled — the throughput metrics remained strong — but with the quality of resolution on complex, emotionally charged, or multi-step cases. Customers reported generic responses, inflexible scripts, and the frustrating experience of being looped through automated flows before eventually reaching a human agent, only to have to explain their issue again from scratch.
The problem was structural, not incidental. AI systems trained on customer service documentation and historical interaction data performed well on routine queries — balance checks, return policies, payment schedules — but struggled with the long tail of interactions that require judgment, empathy, and contextual understanding. A customer disputing a charge they believe is fraudulent needs more than a policy citation. A merchant dealing with a reconciliation error across multiple transactions needs someone who can navigate ambiguity. These interactions represent a minority of total volume but a disproportionate share of customer satisfaction impact and brand perception.
Siemiatkowski acknowledged the overcorrection publicly. In a Bloomberg interview in May 2025, he admitted that cost had become too dominant a factor in workforce decisions and that the result was lower quality. His statement was notable for its directness: the CEO who had declared AI could do every human job was now saying that human presence in customer service is critical for brand credibility and customer trust.
“Cost unfortunately seems to have been a too predominant evaluation factor. What you end up having is lower quality, and that’s not sustainable.”
— Sebastian Siemiatkowski, CEO, Klarna
The Rehiring: Rebuilding What Was Dismantled
Klarna began rehiring human customer service agents in mid-2025, ending a hiring freeze that had been in place for over a year. The company adopted what Siemiatkowski described as an “Uber-style” model — remote agents with flexible schedules, targeting students, parents, and workers in rural areas, with pay starting at 400 Swedish krona (approximately $41) per hour. The company even suggested that passionate Klarna users could apply, framing customer service as something closer to gig work than traditional employment.
The shift to a hybrid model — AI handling routine, high-volume queries while human agents manage escalations, complex cases, and emotionally sensitive interactions — represents a significant retreat from the original AI-first vision. It is not a failure of AI technology per se, but a failure of the replacement thesis: the idea that AI could fully substitute for human judgment in customer-facing financial services without measurable quality degradation.
The reversal also came with costs that were not part of the original business case. Recruiting, onboarding, and training new customer service staff requires investment. Institutional knowledge lost during the attrition period cannot be quickly rebuilt. And the reputational damage from publicly declaring AI supremacy and then quietly rehiring humans creates a narrative risk that follows the company into investor presentations and partnership conversations.
The Financial Picture: Strong Numbers, Hidden Costs
Klarna’s financial performance during this period has been genuinely impressive on headline metrics. Revenue roughly doubled between 2022 and 2025. The revenue-per-employee figure increased dramatically. The company completed a high-profile US IPO, with shares surging 30% on debut and the company reaching a valuation of approximately $19.65 billion — a remarkable recovery from the $6.7 billion valuation it received during the 2022 tech downturn.
But attributing all of this improvement to AI-driven workforce reduction oversimplifies the picture. Klarna’s revenue growth also reflects broader BNPL market expansion, geographic expansion, new product lines including its shopping app and bank account features, and favourable macro conditions for consumer credit. The AI narrative — compelling and headline-friendly — has been layered on top of a multi-factor recovery story.
The hidden costs of the AI-first strategy are harder to quantify but no less real. Customer satisfaction erosion on complex interactions affects retention and lifetime value. The rehiring programme carries direct costs. The shift from stable employment to gig-style arrangements may reduce costs per agent-hour but introduces new challenges around consistency, training depth, and institutional knowledge retention. These are the costs that do not appear in quarterly earnings but shape competitive positioning over time.
What the Klarna Case Means for Financial Services
The Klarna story matters to anyone in the open banking, payments, or financial infrastructure space because it is the most data-rich example of AI workforce transformation in European fintech. The lessons are not abstract — they are backed by real revenue figures, real headcount changes, real customer feedback, and a real strategic reversal. Several conclusions are worth drawing.
First, volume metrics mask quality problems. Klarna’s AI performed well on aggregate measures — tickets handled, resolution time, cost per interaction — while quality deteriorated on the interactions that matter most for customer retention and brand trust. Any financial services company evaluating AI deployment should measure quality by interaction complexity tier, not just overall throughput.
Second, the replacement thesis is weaker than the augmentation thesis. Companies that use AI to make human workers more effective — handling routine tasks, surfacing relevant information, reducing context-switching — consistently report better outcomes than companies that use AI to eliminate human workers entirely. The hybrid model Klarna has now adopted is where most successful implementations end up, regardless of where they start.
Third, public AI-first narratives create reversal costs. Klarna’s aggressive public positioning around AI replacement made the eventual course correction more painful and more visible than it needed to be. Companies that quietly integrate AI into workflows without declaring it a substitute for their workforce retain more strategic flexibility.
Fourth, financial services carry higher stakes than most sectors for AI quality failures. A customer service error at a retail company is an inconvenience. A customer service error at a financial services company — a missed fraud alert, an incorrect payment status, a failed dispute resolution — can have regulatory consequences and material financial impact for the customer. The tolerance for AI error in financial services is structurally lower than in other industries, which means the quality threshold for full automation is structurally higher.
Looking Ahead: Klarna’s Hybrid Future
Klarna remains one of the most aggressive AI adopters in European financial services, and nothing about its reversal suggests the company is abandoning AI. The trajectory is toward a hybrid model: AI handling the 60-70% of interactions that are genuinely routine, human agents handling the rest, and the boundary between the two continuously adjusted based on quality data rather than cost targets.
Siemiatkowski’s revised prediction of 2,000 employees by 2030 still implies significant further workforce reduction. But the framing has shifted from replacement to optimisation — a more sustainable narrative and, based on the evidence, a more realistic one. The question is no longer whether AI can do every human job at Klarna. The question is which jobs AI can do well enough that customers cannot tell the difference, and which jobs still require the judgment, empathy, and adaptability that human agents provide.
For the broader financial services industry — from BaaS platforms considering AI-driven compliance automation to accounting platforms integrating open banking data — the Klarna case offers a simple but powerful framework: automate the routine, augment the complex, and measure quality as carefully as you measure cost. The companies that get this balance right will outperform those that chase the headline of full AI replacement.
Frequently Asked Questions
How many employees has Klarna cut since 2022?
Klarna’s headcount has dropped from approximately 5,500 employees in 2022 to around 3,000 in 2025. The reduction was achieved primarily through a hiring freeze and natural attrition rather than mass layoffs — as employees left, their roles were absorbed by AI tools rather than filled by new hires. CEO Sebastian Siemiatkowski has projected further reductions to approximately 2,000 employees by 2030.
What AI tools does Klarna use?
Klarna deployed several AI systems across its operations. Its primary customer-facing tool is an AI assistant built in partnership with OpenAI that handles approximately two-thirds of all customer service interactions across more than 35 languages. Internally, the company built Kiki, a generative AI assistant that helps employees with tasks like document analysis, sentiment evaluation, and contract drafting. The company reports that 96% of its staff use AI tools in their daily work.
Why did Klarna start rehiring human workers?
Klarna began rehiring customer service agents in mid-2025 after customer satisfaction declined on complex and emotionally sensitive interactions. CEO Siemiatkowski acknowledged that the company had prioritised cost efficiency over service quality and that AI systems could not adequately handle nuanced customer situations requiring empathy and contextual judgment. The company shifted to a hybrid model where AI handles routine queries and human agents manage escalations and complex cases.
Has Klarna’s AI strategy improved its financial performance?
Klarna’s headline financial metrics improved significantly during the AI transformation period. Revenue per employee increased by 152% since early 2023. Third-quarter 2025 revenue reached $903 million, up 28% year-on-year. Average employee salaries increased by roughly 60% to approximately $203,000, funded partly by savings from reduced headcount. However, these improvements coincided with broader BNPL market growth and new product launches, making it difficult to attribute the gains solely to AI-driven workforce reduction.
What is the “Uber-style” hiring model Klarna adopted?
When Klarna reversed its hiring freeze, it adopted a flexible gig-style model for customer service agents. The programme allows workers to choose their own schedules and work remotely from anywhere in Sweden, with pay starting at 400 Swedish krona (approximately $41) per hour. The company targeted students, parents, rural workers, and even passionate Klarna customers as potential agents. This model reduces fixed employment costs while restoring human presence in customer interactions, though it raises questions about training depth and service consistency compared to traditional employment arrangements.
What lessons does the Klarna case offer for other fintech companies?
The Klarna experience highlights several practical lessons for financial services companies considering AI-driven workforce transformation. Volume-based metrics can mask quality deterioration on complex interactions. Full AI replacement strategies consistently underperform hybrid human-AI models in customer-facing roles. Public narratives about AI replacing workers create reputational risk if course corrections become necessary. And financial services carry higher stakes for AI quality failures than most other sectors, meaning the quality threshold for automation must be set higher than cost models alone would suggest. The most effective approach is to automate routine tasks, augment human workers on complex tasks, and measure quality by interaction complexity rather than aggregate throughput.
MyValue Solutions is an independent publication. We are not affiliated with Klarna or any company mentioned in this article. This analysis represents our editorial assessment based on publicly available information and should not be construed as investment or financial advice.