Klarna Experiment: Real-World reflections on Agentic AI deployment

Sep 15, 2025
  • artificial intelligence & RPA

Klarna, a global fintech leader, launched an ambitious AI initiative to automate its customer service operations across 23 markets and 35 languages. Partnering with OpenAI, Klarna deployed an AI assistant that handled over 2.3 million customer conversations in its first month, equivalent to the workload of 700 full-time agents.

The assistant managed tasks such as refunds, payment issues, and account queries, delivering faster resolution times and reducing repeat inquiries. 

Klarna reported a 47% increase in customer satisfaction and over $10 million in annual savings, with $4 million attributed to customer service and $6 million to marketing automation. However, the rollout revealed critical challenges. 

Customers began expressing frustration over the lack of empathy and contextual understanding in complex interactions. Klarna’s leadership acknowledged that aggressive cost-cutting had compromised service quality. In response, the company pivoted to a hybrid model, reintroducing human agents for nuanced support while retaining AI for routine tasks. They also launched a flexible staffing model to bring in skilled support on demand. This evolution highlighted the importance of balancing efficiency with experience and reinforced the need for robust governance and monitoring. 

Klarna’s case illustrates how agentic AI can deliver rapid returns but also underscores the risks of over-automation without human oversight. It serves as a cautionary tale and a source of inspiration, showing that sustainable transformation depends on thoughtful integration, clear KPIs, and continuous recalibration. While Klarna’s scale and context may differ from other organisations, the lessons are broadly applicable to any agentic AI journey. 

Key observations 

This use case confirms a set of parameters and formulas that serve as a practical basis for evaluating value, cost, and returns: 

  • Speed of Deployment: Klarna moved rapidly from pilot to production, demonstrating the potential for fast time-to-value when data and processes are well-aligned. 
  • Impact on Roles: The shift raised questions about workforce displacement versus augmentation, highlighting the importance of transparent change management. 
  • Customer Experience: Initial reports suggested mixed outcomes, while some interactions improved, others suffered from lack of empathy or contextual understanding. 
  • Governance & Trust: The experiment underscored the need for robust monitoring and explainability, especially when agents operate in customer-facing roles. 

Relevance to Agentic AI Strategy 

While Klarna’s case is more aligned with AI-assisted and partially agentic automation, it offers valuable insights related to scaling, governance, and human impact. It reinforces the importance of piloting with clear KPIs, engaging stakeholders early, and maintaining oversight as autonomy increases. 

Cautionary Note 

It’s essential to recognize that real-world examples like Klarna’s can be illustrating, but they must always be interpreted through the lens of context. Factors such as company size, industry specifics, and risk tolerance can greatly influence outcomes. It’s prudent to examine them as informative reference points, drawing out relevant lessons while tailoring any approach to the unique contours of your own organization. 

Want to know more ? 

read our full article