Klarna Rehires Staff It Replaced With AI as Service Quality Drops

Klarna is rebuilding its human customer-service team less than two years after CEO Sebastian Siemiatkowski claimed an AI assistant was doing the work of 700 full-time agents. The Swedish buy-now-pay-later company confirmed the reversal to Fortune in early November 2026. It is not a full retreat from AI. It is a recalibration — and part of a wider pattern.

What happened to the 700-agent claim?

In February 2025, Siemiatkowski told the Financial Times that Klarna's AI assistant had taken on the equivalent of 700 customer-service roles. Klarna had said in early 2024 that the assistant was built with OpenAI. The company had already frozen hiring in August 2024 and cut headcount by roughly 22%, citing AI productivity gains.

By mid-2025 the picture changed. Siemiatkowski acknowledged the push had produced "lower quality" support, and Klarna began recruiting humans again through a remote, Uber-style pilot program.

A Klarna spokesperson told Fortune that human agents were needed "for complex, sensitive cases" and that the company had begun directly hiring a "small number" of high-skilled people to identify where the "human touch" adds most value.

How common is 'AI rebounding'?

Klarna's U-turn fits a measurable trend. Research from talent-assessment firm SHL, published November 9, 2026, found that 19.1% of U.S. workers replaced by AI said they were later asked to return — a phenomenon SHL calls "AI rebounding."

Those who came back had demands:

• 51% wanted a substantial pay increase
• 42% required written job-security guarantees
• 40% wanted transparency about how AI would be used in their roles

A separate Robert Half survey of hiring managers found that 32% of organizations that cut positions because of AI later reinstated the same or similar roles. The reasons were concrete:

• 40% said the work required institutional knowledge AI could not replicate
• 39% cited relationship management
• 39% pointed to increased business demand
• 38% said more human oversight and quality control were needed than expected

Was the original claim ever sound?

The evidence suggests the 700-agent figure measured ticket volume, not resolution quality. An industry tracker assessed that Klarna's AI "struggled with edge cases, emotionally complex interactions, and multi-step problem resolution." Customer satisfaction scores fell, and rehiring costs exceeded the original savings, according to the same tracker.

Klarna is not alone in overpromising. Ford hired and promoted 350 experienced engineers in June after AI-powered quality-control tools fell short of expectations; the automaker said AI contributed gains but stressed the need to combine it with deep technical expertise.

Gartner analyzed more than 1 million layoffs in 2025 and found that cuts genuinely driven by AI productivity — where one worker became so productive that fewer were needed — accounted for under 1% of the total. The firm estimates that by 2029, 30% of employees terminated and replaced by AI will be rehired, often at higher cost, unless organizations overhaul workforce planning.

What do the layoff numbers show?

Challenger, Gray & Christmas data shows U.S. employers cited AI in 121,401 announced job cuts through October 2026 — more than double the 54,836 for all of 2025. Yet the same data shows AI dropping from the top-cited reason for cuts in March–July to fifth place in September. Companies including Zillow, Etsy, LinkedIn, Microsoft, Uber and HubSpot explicitly denied AI was driving their latest rounds.

What remains unclear?

Klarna has not disclosed how many roles it has rehired, the net headcount change since the 2024 freeze, or whether the "small number" of new hires represents a fraction or the bulk of the 700-agent equivalent it once claimed to have replaced. The company says it continues to use AI for routine queries while keeping humans for complex cases — a hybrid model.

The mechanism behind these reversals is consistent across the research: companies automate the measurable parts of a role — volume, speed, script adherence — while underestimating the tacit knowledge, judgment and relationship work that only becomes visible when it disappears. Gartner's forecast implies that the bill for that blind spot will land by 2029, and it will be higher than the savings.