Case Study: Ramp improves AI agent performance with Langfuse

Ramp cuts AI agent session times by 30% with Langfuse

Ramp, a smart financial infrastructure company, uses Langfuse to build and manage a portfolio of AI agents that power nearly all of its operations. These customer-facing and internal agents handle critical tasks like expense approvals and data analysis, and the challenge was to effectively observe, trace, and improve their performance at scale.

The solution was implemented using the self-hosted, open-source Langfuse platform. Ramp built an observability layer with Langfuse that allowed them to trace agent activity. They then created an agent called Reflect, which uses Langfuse to score, cluster, and analyze these traces to automatically diagnose issues and propose fixes. This resulted in a 10-20% decrease in token usage, 15% fewer tool calls, and 30% faster session times for their agents.

Ramp