Cross-Industry by Revolutio Research

The AI Customer Service Reversal: What Klarna Actually Proves

In 2024, Klarna said its AI assistant did the work of 700 agents and would improve profits by $40M. By May 2025 the CEO was publicly calling the quality cuts a mistake and rehiring humans. The most-cited AI replacement story of the decade turned out to be a two-year lesson in what AI customer service can and cannot do. The macro numbers now say the same thing: AI shows up in a majority of layoff announcements, but full replacement keeps failing the quality test.

The AI Customer Service Reversal: What Klarna Actually Proves

The short answer

The single most-cited AI job replacement story was Klarna. In February 2024 the company announced its OpenAI-powered assistant was doing the work of 700 full-time customer service agents — resolution time down from 11 minutes to under 2, repeat inquiries down 25%, handling two-thirds of all service chats across 23 markets, with a projected $40M profit improvement. Two years later, the honest reading is: the automation worked exactly as claimed on volume metrics, and failed exactly where critics predicted on quality. Klarna’s CEO publicly called the cost-focused approach a mistake in May 2025 (Bloomberg), and the company has been rehiring human agents since — moving to an Uber-style flexible remote support workforce rather than back to a traditional call center. Meanwhile the 2026 layoff data shows AI cited as a factor in 54% of tracked layoff events (173 of 322 through early August, per SkillSyncer). Both things are true: AI is genuinely restructuring customer service, and the “replace the whole function” version of the story keeps failing.

The evidence

2024: the claim

Klarna’s February 2024 announcement was the most aggressive AI replacement claim from any major consumer company, and the numbers were specific:

  • The AI assistant handled two-thirds of customer service chats in its first month
  • Resolution time: 11 minutes (human baseline) to under 2 minutes
  • Repeat inquiries down 25%
  • Available in 23 markets, 35+ languages, 24/7
  • Klarna framed it as the equivalent of 700 full-time agents, with the company simultaneously shrinking its total workforce from roughly 5,500 to 3,400
  • Projected profit improvement: $40M in 2024

These were the company’s own numbers, not independent measurements — a distinction that matters more in retrospect. Klarna was also unusually aggressive in promoting them; the CEO said AI could already do the work, and hiring for affected functions was paused.

2025: the walk-back

The reversal came in stages, and Klarna was more honest about it than most companies would be:

  • By late 2024, service quality complaints were piling up. Engineers and internal staff were pulled into answering support tickets during peak periods
  • In May 2025, CEO Sebastian Siemiatkowski told Bloomberg the cost-cutting focus in customer service had gone too far, and that the company was now actively recruiting human agents again — explicitly framing higher costs as a price worth paying for quality
  • The rehiring model is telling: Klarna is not rebuilding a classic call center. It is building an Uber-style remote workforce — flexible, on-demand human agents layered on top of the AI that still handles routine volume
  • Forbes’ coverage in the same month was blunt: customers prefer talking to people, and the chatbot disappointed them

2026: where it stands

The rehiring has continued through 2026. The structure that is emerging at Klarna — AI as the first tier handling routine volume, humans as a flexible second tier for everything the AI escalates — is the same structure deploying across the industry. What died at Klarna was not AI customer service. It was the “700 humans replaced” framing.

Where the disruption is still real

The reversal does not mean customer service AI is hype. Three things remain true:

  • Volume automation works. Handling routine, high-frequency, templated interactions (order status, password resets, billing questions) is a solved problem. The minutes saved per interaction are real
  • Headcount composition is still changing. Even at Klarna, total support headcount per customer served is down versus 2023. Companies across the sector run leaner support orgs with AI as the base layer
  • Entry-level support roles are still the most displaced job category in the labor data. Customer service employment has declined year-over-year while overall employment grew — a pattern consistent with restructuring, not replacement

The failure was specific: full replacement of a quality-sensitive function, measured only on speed and cost.

The jobs picture

The 2026 macro data says the Klarna pattern is the rule, not the exception:

  • 322 layoff events tracked across tech through August 25, 2026, affecting roughly 205,800 workers (SkillSyncer tracker)
  • AI was cited as a factor in 173 of 322 events through early August — 54% of all announcements
  • The layoff categories where AI is cited are concentrated: customer support, content production, translation, QA, and junior analysis — the same categories our industry pages track as actively disrupting
  • Counter-signal: companies that announced full-function AI replacements in 2024-2025 have disproportionately issued rehiring or “hybrid” announcements since. The full-replacement bet has a poor track record; the tiered-restructure bet is holding

Confidence levels: high on the Klarna timeline (multiple independent outlets, company’s own statements); high on layoff event counts (tracker data, corroborated by Yahoo and TrueUp trackers within ~5%); medium on the AI-cited percentage (tracker methodology depends on layoff announcement language).

What to watch

  • Whether Klarna’s Uber-style flexible human layer becomes the industry template, or gets automated again as models improve — this is the single most informative natural experiment in the field
  • Whether the 54% AI-cited layoff share keeps rising into 2027, or plateaus as companies hit the quality ceiling Klarna hit
  • Whether any major consumer company repeats the “AI replaced N hundred workers” announcement framing in 2026-2027 — the near-total absence of new claims of that shape since early 2025 is itself a data point

FAQ

Did Klarna’s AI actually replace 700 people? The assistant handled the equivalent chat volume of 700 agents. But Klarna’s total workforce fell from about 5,500 to 3,400 through a hiring freeze and broader cuts — the 700 figure was a workload equivalence, not a documented headcount elimination. By 2025 the company was rehiring for support quality.

Is AI customer service a failure, then? No. Routine-volume automation is deployed industry-wide and measurably reduces cost per contact. What failed was the claim that quality-sensitive support could be fully automated — the industry has converged on AI-first tiering with human escalation.

What does this mean for customer service jobs? Restructuring, not elimination. Total support headcount per customer is falling, entry-level roles are the most affected, and hybrid human-AI operations roles are growing. The pure “AI replaced the department” story keeps not surviving contact with reality.

Data sources: Klarna company announcements (Feb 2024); Bloomberg (May 8, 2025); Forbes (May 18, 2025); Customer Experience Dive; SkillSyncer Layoffs Tracker (Aug 2026); Yahoo Tech layoffs tracker; TrueUp. Every claim above is traceable. Counter-evidence included by design.