In February 2024, the buy-now-pay-later company Klarna made one of the boldest public claims of the early generative-AI era. In a press release dated February 27, 2024, it said its OpenAI-powered customer-service assistant had, in its first month, handled 2.3 million conversations - about two-thirds of all chats - which it called “the equivalent work of 700 full-time agents.” It claimed parity with humans on satisfaction, a drop in resolution time from 11 minutes to under 2, a 25 percent cut in repeat inquiries, and a roughly 40-million-dollar profit improvement for 2024. The number became shorthand for the idea that AI was about to gut customer-service jobs.
The story did not stay that simple. Through 2024 Klarna paired the AI push with a hiring freeze and a shrinking headcount, and CEO Sebastian Siemiatkowski leaned into the message that AI could do the work of people. Then in 2025 he publicly reversed the emphasis. In interviews around the company’s results, he acknowledged that cost had become “a too predominant evaluation factor” in how support was organized, that this had produced “lower quality,” and that Klarna was reinvesting in human agents - including a model that lets people work flexibly, almost like ride-share drivers, while always giving customers the option to reach a person.
Klarna did not abandon AI; its own later results kept reporting falling customer-service costs per transaction. The walk-back was about balance, not retreat. But the episode became a cautionary tale told alongside Duolingo’s “AI-first” memo: a vivid headline number, a quieter correction once quality and customer experience were weighed against pure cost.
Note on sourcing: the 2024 “700 agents” press release was fetched and verified live as the Tier 1 primary. The 2025 reversal statements were made by Klarna’s CEO in interviews tied to company results and are reported consistently across independent outlets; they are summarized here as attributed claims rather than cited to a single fetchable company filing.