Metabase

Building a dashboard is trivial now. Verifying what it tells you is the hard part.

AI is changing data roles fast. Visualizing data and running reports? Anyone can ask a question and get an answer in seconds. Judgement, context, and verification - that’s where the value is now: knowing whether that answer is right, what it's actually based on, and being able to prove it when someone asks.

**In this AI Analytics Week session, we'll get into:**

*   Why AI gives confidently wrong answers on real schemas, and why that's a context problem, not a model problem
*   What a semantic layer is, and why curating one is now the highest-leverage thing a data person does
*   How to set it up in Metabase, so every answer, whether reached by a human or an agent, is grounded in the same context and can be verified the same way
*   Why doing it in Metabase helps you maintain an end-to-end view over your data, from transform to question.
*   A live build: raw tables → transforms → business context in the semantic layer → a canonical dataset → an answer you can open up and check yourself
*   What happens when something upstream breaks: change a transform, and lineage flags what it takes down, in the same place you built it
*   Where this already works today: in-product, Slack, SQL assist, and MCP, in every plan including open source

A session for anyone who'd get asked to explain a number: analysts, data leads, and the people who now have AI-generated answers landing in their inbox.
