Beyond Dashboards: Fixing Text-to-SQL with Semantic RAG

We all want to replace static dashboards with conversational AI. However, studies show that “out-of-the-box” LLMs struggle with Text-to-SQL on complex enterprise schemas. The missing link isn’t a better model - it’s the Semantic Layer.

In this fast-paced session, we will bridge the gap between 90s BI wisdom and the GenAI era.

We will cover: - The Context Gap: Why zero-shot prompting fails and why locking semantics in BI is a dead end for AI agents. - The Architecture: How modern solutions (e.g. UC Metric Views) bring semantics back to the Data Lakehouse, where they belong. - Back to the Future: What we can learn from “ancient” tools like OBIEE/Cognos about governing complex data relationships. - The “Semantic RAG” Pattern: How injecting your semantic model into the LLM prompt can allow the AI to understand table relationships and business logic (like a semantic layer) while retaining the flexibility to write raw SQL for ad-hoc queries.

Join this talk to learn what backend can finally make “Chat with Data” a reality.