Intercom
Application performance monitoring
Engineers lacked visibility into a slow AI agent. Rebuilt low-level tracing exposed critical bottlenecks and accelerated LLM responses.
- 60% reduction in median customer wait time for first token
- 2-second reduction in median customer wait time for first token