Klarna's AI assistant took two-thirds of support chats in its first month

Klarna put an OpenAI-powered assistant in front of customer service chats. It handled 2.3 million conversations in month one. In 2025 the CEO said quality had suffered and Klarna started hiring human agents again.

Organization: Klarna
Industry: Fintech
When: Feb 2024
Evidence: Company reported
Agents and tools: OpenAI GPT-4

Reported outcome

2.3M conversations in the first month, about two-thirds of all service chats, the work of roughly 700 full-time agents (company claim).

2.3M
Conversations, month one
~2/3
Share of chats
11 min → <2 min
Resolution time
23 / 35+
Markets / languages

Caveats

Stated by the company that uses the agent (blog, earnings call, interview). Not independently audited.

All launch numbers come from Klarna and OpenAI. No independent audit.
In May 2025 the CEO said a cost-first AI push lowered service quality, and Klarna began recruiting human agents again so customers can always reach a person.

What the agent does

The assistant answers customer service chats inside the Klarna app. It handles common requests such as refunds, returns, payment questions, and disputes. It hands the conversation to a human when it cannot help.

What happened

Klarna said the assistant matched human agents on customer satisfaction. It also said repeat inquiries fell by 25%. The company estimated a $40M profit improvement for 2024.

About 15 months later, the CEO told Bloomberg that cost had been "too dominant" a factor. Klarna started hiring remote human agents again. It still uses AI heavily.

What to copy

Measure quality, not only volume. Deflection rate looks good on a dashboard. Repeat contacts, complaints, and churn tell you if it works.
Keep a human path. Make it easy and fast to reach a person.
Start with high-volume, low-risk intents. Refund status and "where is my payment" are good first targets.

Starter prompt

Written by us, not by Klarna. Adapt it to your own policies.

You are the support assistant for {company}. Answer only from the policy
documents and account data provided. If the answer is not in them, say so
and offer a human agent.

Always:
- Confirm the customer's goal in one sentence before acting.
- Never promise refunds, credits, or dates that the data does not confirm.
- Offer a human agent if the customer asks, is upset, or the issue involves
  fraud, safety, or a legal complaint.

End each answer with: "Did this solve your issue?"