Beyond the dashboard: an MCP-connected AI agent that reads your dbt lineage
An AI agent that reads dbt lineage from inside the BI layer answered three data mysteries in about the time it takes to get coffee.
Overview
Traditional BI copilots struggle with questions like "why is this number wrong?" Shows an AI agent integrated with dbt via MCP that pulls report context, analyzes compiled SQL, traces column-level lineage back to source tables, clarifies business logic, and runs validation queries. Includes a live demo investigating a data discrepancy end to end.
What you'll take away
- 01
AI copilots bolted onto dashboards can summarize a chart but can't explain why a number is wrong, because they only see BI-layer gold tables, not the business logic in the dbt models underneath.
- 02
Connecting one agent to the dbt, Databricks, and Looker MCP servers let it trace a misclassified product back to a single CASE statement and file a Jira ticket with QA queries attached in under three minutes.
- 03
Because the agent inherits whatever access each MCP server already enforces, governance came for free: it's read-only end to end, and every query lands in the warehouse's audit trail.
Session speaker
Mars Dauer