Scaling Grafana Assistant: Context, confidence, and AI observability

As Grafana Assistant expanded its capabilities, the challenges changed. The hard part was no longer just adding functionality. It was making the Assistant’s agentic workflows actually useful, improving it with confidence, and understanding how it behaved in the real world. In this session, members of the Grafana Labs engineering team share what they learned scaling Assistant: how better context makes it more useful, how trusted feedback loops make iteration safer, and how observability helps them understand Assistant once it is live.