Klarna says AI does the work of 700 customer service agents

Klarna said the AI customer-service assistant it had built on OpenAI’s models handled 2.3 million conversations in its first month live — two-thirds of all customer-service chats — and that the volume of work was equivalent to that of 700 full-time agents. The company reported that customers resolved issues in under two minutes on average, against 11 minutes previously, with a 25% drop in repeat inquiries and customer satisfaction scores matching human agents. The assistant operated across 23 markets and more than 35 languages, and Klarna projected it would contribute $40 million to the company’s profit over 2024.

The “700 agents” figure was a productivity-equivalence claim rather than an announced headcount reduction: Klarna did not state at the time that 700 people had been let go, only that the chatbot’s throughput matched that scale of staffing. It nonetheless became one of the most widely cited numbers in the early debate over generative AI and jobs, repeated across business press and executive commentary as evidence that large language models could substitute directly for service labour, not merely assist it.

The claim later became a reference point for the opposite argument as well. Klarna chief executive Sebastian Siemiatkowski said in 2025 that the company had begun rehiring human agents after concluding that AI-only service had degraded quality for complex queries, a reversal that critics of the original announcement cited as evidence the 2024 figures had oversold what the assistant could sustain. The February 2024 announcement remains one of the earliest and most concrete corporate claims that generative AI had already displaced measurable amounts of white-collar labour, ahead of the broader wave of AI-linked layoff announcements later in the decade.