Klarna AI Assistant Case Study: 2/3 Chats Handled by AI

Klarna is a global digital payments and shopping company operating across numerous markets and languages. Its large customer base generates substantial volumes of payment, refund, return, and account-related enquiries.

Where Klarna’s Customer Service Was Losing Time 

Klarna needed to provide fast, round-the-clock support while managing growing customer-service demand across different countries. 

Major enquiries involved repetitive processes, but customers still expected accurate and contextual answers. Expanding the traditional support model in proportion to customer growth would have increased operational costs and management complexity.

The AI Product Klarna Introduced

Klarna launched an OpenAI-powered customer-service assistant within its application.

The assistant was designed to:

• Answer payment and account questions.

• Support refunds and returns.

• Provide information based on customer context.

• Operate continuously in 23 markets.

• Communicate in more than 35 languages.

• Transfer customers to human support when required.

OpenAI reports that Klarna connected its customer-service processes with generative AI while also making ChatGPT Enterprise available to employees across functions such as technology, marketing, communications, and legal. 

Sources: OpenAI’s Klarna case study

Results Klarna Reported

During its first month, Klarna reported that the AI assistant:

• Completed 2.3 million customer conversations.

• Handled approximately two-thirds of its customer-service chats.

• Processed a workload equivalent to 700 full-time agents.

• Reduced average resolution time from 11 minutes to under two minutes.

• Reduced repeat enquiries by 25%.

• Achieved customer-satisfaction scores comparable to human agents.

Sources: Klarna’s official AI assistant announcement

By May 2024, Klarna said the assistant had interacted with more than four million customers and was expected to deliver approximately USD 40 million in annualised savings. 

Sources: Klarna Q1 2024 results

The USD 40 million figure should be understood as a company-projected annualised saving rather than independently audited realised profit.

The Later Shift Towards Human Support

Klarna’s experience also demonstrates the limits of treating customer-service automation purely as a cost-reduction programme.

In 2025, Klarna began hiring additional customer-service employees and shifted its AI focus from cost reduction toward business growth. The AI assistant remained part of its operations, but the company placed greater emphasis on maintaining access to human support. 

Sources: Reuters’ Klarna update

This does not necessarily mean the AI product failed. It indicates that routine, high-volume enquiries and complex or sensitive customer situations require different service models.

What This Means for Your Business 

Businesses implementing AI assistants should measure more than the percentage of conversations automated.

A balanced performance framework should include:

• Resolution time.

• Repeat-contact rate.

• Customer satisfaction.

• Escalation accuracy.

• Incorrect-answer rate.

• Customer retention.

• Performance on sensitive and complex cases.

• Cost per successfully resolved inquiry.

AI can absorb repetitive demand, but customers should be able to reach a qualified human when judgement, empathy, or financial consequences are involved.

Deploy AI With Proven Expertise 

Rubixe develops AI chatbot solutions that can connect with CRM, ERP, and customer databases to provide contextual responses and automate routine customer actions.

Rubixe can also help businesses establish:

• Human escalation rules.

• Response-quality monitoring.

• Multilingual support.

• Data-security controls.

• Conversation analytics.

• Continuous model improvement.

• Integration with existing support platforms.

Curious About What AI Could Do for Your Business? 

Book a consultation with Rubixe AI consultants to evaluate where conversational AI can improve customer support without compromising service quality.

Note: Rubixe was not involved in Klarna’s implementation. This is Rubixe’s independent analysis of publicly documented results.