The semantic layer is dead. Long live the semantic layer!

Semantic layers used to be boring. A solved problem. The thing you built years ago for your dashboards. Then agents showed up.

Suddenly everyone is trying to design what an AI system knows, how it reasons, and what it can trust. They're calling this context engineering, and it's what you've already been doing for years without a name. You weren't just defining metrics. You were building the ERD of your business.

In this session, we’ll cover why your semantic layer is now the most load-bearing piece of your AI strategy. We'll show where MetricFlow and the dbt Semantic Layer are heading, and how the dbt MCP server turns your governed metrics into context your agents can actually use. We’ll also share what teams are actually unlocking when they centralize context at the transformation layer: fewer AI hallucinations, lower token costs, and BI that doesn't contradict itself.