Is the observability UI as we know it dead? And what happens next?

AI is changing how engineers investigate production systems. Agents are taking a more active role in exploring telemetry, testing hypotheses, and finding evidence. Where should that work happen? Should investigations live inside a vendor's proprietary agent, or should teams be able to choose their own agents and connect the tools and data they need?

We'll explore what it takes to support that choice. How does MCP give agents access to investigative tools and data beyond navigating existing dashboards? What role does SQL play in letting them ask new questions and produce queries that engineers can inspect? And why is ClickHouse suited to supporting these investigations over detailed telemetry?

That leads to a broader question: if agents can help engineers investigate from the environment they choose, what do they still need from an observability UI?

Teams still have recurring questions, reporting needs, and operational workflows. How could those interfaces better reflect their own systems, rather than a vendor's interpretation of them?

Join Kelsey Hightower and Mike Shi from ClickHouse for a discussion about how the investigation surface is changing, where agents fall short, and what engineers still need from their interfaces. Through demonstrations of investigation with an agent and custom apps powered by ClickHouse SQL, we'll explore where each approach fits and what it takes to trust and maintain them.