Metabase Data Studio builds semantic layer

Metabase is trying to own the trust layer for analytics, not just the dashboard layer. That matters because AI answers are only as reliable as the metric definitions underneath them, and this release focuses on the unglamorous plumbing that keeps those definitions consistent.

• –SQL and Python transforms let teams clean, join, and pre-aggregate raw tables before they reach analysts or AI
• –Lineage and dependency diagnostics make it safer to change shared metrics without breaking downstream dashboards
• –Publishing curated datasets to a library turns Metabase into more of a semantic-layer workflow than a simple BI frontend
• –The AI angle is credible only if definitions are centralized first; this is really about reducing ambiguity before an LLM ever answers a question
• –For existing Metabase users, folding modeling and analytics into one product is a stronger story than stitching together separate BI and semantic-layer tools