Generative AI holds promise for analytics, but Large Language Models (LLMs) often fail in complex data ecosystems—hallucinating queries, missing schema details, and prolonging debugging. This lightning talk explores how integrating Databricks with dbt via the Model Context Protocol (MCP) addresses this. By surfacing dbt metadata—models, tests, lineage—through an MCP server on Unity Catalog, LLMs gain structured context for schema-aware, reliable responses. We’ll dive into the technical setup and demo how agents dynamically query dbt artifacts, reducing manual intervention. In production, this dbt + Databricks synergy significantly reduced query errors, accelerated exploration, and boosted efficiency, enabling faster forecasting and cost optimization without vendor lock-in. Looking forward, MCP signals a future of human+AI co-piloted workflows where dbt’s evolving semantic metadata fuels predictive modeling, real-time governance, and conversational analytics.