Bankrupt by AI
All case files
On the recordWalked back · May 2025

Reversal

Klarna and the 700-agent headline

Klarna · Fintech · Sweden

In February 2024 Klarna said its AI assistant did the work of 700 agents. Fifteen months later its CEO said cost had crowded out quality — and the company started hiring humans back.

Feb 2024

700

agent-equivalent · two-thirds of chats

May 2025

Humans

rehiring · “lower quality,” said the CEO

Fifteen months from headline to correction.

The record

  • Klarna said its OpenAI-powered assistant handled 2.3 million conversations — two-thirds of its customer-service chats — in its first month. (Klarna press release, Feb 2024)
  • The company put the assistant’s output at “the equivalent work of 700 full-time agents” and projected a roughly $40 million profit improvement for 2024. (Klarna / OpenAI, Feb 2024)
  • In May 2025, CEO Sebastian Siemiatkowski told Bloomberg that cost had been “a too predominant evaluation factor,” and “what you end up having is lower quality.” Klarna began recruiting human agents again. (Bloomberg, May 2025)

The headline

The number traveled further than any product ever could. In late February 2024, one month after its OpenAI-powered assistant went live, Klarna announced that the bot had handled 2.3 million conversations — two-thirds of its customer-service chats — and was doing “the equivalent work of 700 full-time agents.” OpenAI co-published the figure; Klarna projected a $40 million profit improvement for the year. For two quarters it was, in effect, the most-cited customer-service statistic in the industry.

It was also a company’s own launch metric, unaudited and framed for maximum reach. That is not a criticism of the technology, which plainly did handle real volume. It is a note about what kind of number it was: a projection and a work-equivalence estimate, presented as a milestone.

The walk-back

In May 2025, the same CEO who had championed the automation told Bloomberg the company had leaned too hard on it. “As cost unfortunately seems to have been a too predominant evaluation factor when organizing this, what you end up having is lower quality,” Siemiatkowski said. Klarna began a recruitment drive so that, in his words, “there will always be a human if you want.”

This was a correction, not a collapse. Klarna still calls itself AI-first; the assistant still carries most of the chat volume; the rehiring began as a limited, flexible pilot rather than a full restoration of the roles it had stopped hiring for. But the direction was unmistakable — quality, not cost, was now the stated priority, and a human backstop was now a brand principle.

What the 700 actually was

The load-bearing lesson is in the number itself. “700 full-time agents” was never a headcount that got cut; it was Klarna’s estimate of the work the assistant did, benchmarked against agents the company would otherwise have needed to hire while it grew. Read as a capability metric, it was defensible. Read as a redundancy count — which is how much of the coverage read it — it overstated what had happened to people and understated what would happen to quality.

The failure, to the extent there was one, was letting a single variable — cost — carry a decision that also priced in trust, edge cases and the option of a human. When that option was cut, the loss did not show up in the deflection dashboard. It showed up fifteen months later, in the CEO’s own words.

The lesson

A workload-equivalence headline is a marketing figure, not a headcount plan. When cost becomes the metric that crowds out the others, quality is the variable that pays — quietly, and on a delay.

How we’re reading this

“700 agents” was Klarna’s estimate of work-equivalence, not 700 people dismissed, and the $40M was a projection the company never publicly audited. Klarna still describes itself as AI-first: this was a recalibration toward a human-backstopped hybrid, not an abandonment of AI. We report what the company stated in 2024 and what its CEO said about quality in 2025.

Sources

  1. 01
  2. 02
  3. 03
  4. 04
  5. 05
The pattern, anonymizedThe 90-day headcount experiment

Compiled from public filings, court records, company statements and reputable press. Figures are attributed to their sources; allegations are labeled as such. Not legal or investment advice.