Prepare Metabase data for AI
Start with one question colleagues ask every week and check whether the available data, definitions and permissions let an assistant answer it correctly.
Teams can prepare Metabase data for AI by testing a familiar business question before opening up more of the warehouse. The ai-readiness-checklist skill helps you connect the question to the right table, define its metric and check the resulting answer against a known figure.
For a retail team, “weekly sales” might mean paid orders after refunds. If one dashboard includes cancelled orders and another excludes them, improving the prompt alone will not settle the difference. The process records which definition the team accepts and where further work is needed.
What it does
- Choose a narrow question to test with real project data.
- Describe fields whose names are meaningful only to their original authors.
- Check that a metric has one reusable definition.
- Identify which questions or dashboards the team treats as authoritative.
- Separate results verified through a connection from items reported by a colleague.
Install and use
- Install the skill and use a permitted Metabase account. The optional MCP connection can send query results to your agent and its configured provider.
- Ask “We want to ask about weekly sales. Which tables, definitions and permissions should we check first?”
- Compare one answer with an agreed report, then save the remaining actions for the next session.
FAQ
Source
Made by Metabase. Original skill: ai-readiness-checklist, published on GitHub under the MIT licence.
View on GitHub