Case Studies LangGraph LangSmith How Klarna's AI assistant redefined customer support at scale for 85 million active users

Klarna, with 85 million active users and 2.5 million daily transactions, built an AI assistant on LangGraph and LangSmith to handle payments, refunds, and escalations. The assistant routes requests through a LangGraph framework, dynamically tailoring prompts to reduce token costs and latency. With LangSmith’s tracing, the team performed test‑driven development, pinpointing issues step‑by‑step. Prompt optimization, including meta‑prompting, was iterated based on LangSmith insights. The result was an agent that performs the work of 700 full‑time staff, reducing average query resolution time by 80%, automating 70% of repetitive support tasks, and improving root‑cause identification for escalations.

Klarna’s partnership with LangChain enabled a scalable, multi‑agent system that cut operational costs and improved customer experience across global markets. The case study demonstrates how LangGraph’s routing and LangSmith’s observability can be combined to deliver high‑volume, high‑quality AI support.

The assistant’s success is measured by faster resolution, higher automation, and improved accuracy, illustrating the power of a well‑engineered agent architecture.