Klarna automated two-thirds of its service chats, then brought people back

Klarna handled enormous volumes of routine service requests (refunds, returns, payment status) across dozens of markets and languages, and it took a customer an average of 11 minutes to resolve an issue.

It built an assistant on top of its own account data and pointed it at the high-volume, well-defined requests. It also pushed generative AI inside the company: OpenAI’s customer story reports 90% of Klarna employees using its tools daily, with communications, marketing and legal at 93%, 88% and 86%.

In the first month the assistant had 2.3 million conversations, two-thirds of all customer service chats, doing the equivalent work of 700 full-time agents. Customers resolved issues in under 2 minutes instead of 11, satisfaction matched human agents, repeat inquiries fell 25%, and Klarna estimated a $40 million profit improvement for 2024. By the first quarter of 2025 the cost per transaction for customer service had fallen 40% over two years, and by the third quarter of 2025 the company said the agent did the work of more than 853 agents and had saved $60 million.

In May 2025, CEO Sebastian Siemiatkowski told Bloomberg that “cost unfortunately seems to have been a too predominant evaluation factor when organizing this, what you end up having is lower quality,” and that “it’s so critical that you are clear to your customer that there will be always a human if you want.” Klarna began hiring for a flexible customer service workforce alongside the AI. The AI didn’t go away: two-thirds of inquiries were still automated in late 2025.

The automation delivered what was promised: speed, cost and scale. The reversal shows what happens when cost is the only number on the dashboard. The independent research points the same way. In a study of 5,179 customer support agents (Brynjolfsson, Li and Raymond, published by the National Bureau of Economic Research), giving agents a generative AI assistant raised productivity 14% on average and 34% for newer, less-skilled workers, while improving customer sentiment and employee retention. The most reliable gains came from AI helping people, not replacing them.