Build a churn dashboard from Databricks Genie data with Claude

Suranga Nanayakkara connected Claude to a Databricks Genie space over MCP and had it query a credit card churn dataset and build a full dashboard, then rebuilt the same thing in Databricks AI/BI to compare.

Done withClaudeinDatabricks· Oct 30, 2026

What they did

Nanayakkara, who works with Databricks, wanted to try the Claude dashboards that kept showing up on LinkedIn. The author connected Claude to a Genie space in their own workspace through the Genie connector, served from Unity AI Gateway, using an OAuth app connection they created in the Databricks account console. The data was the public credit card churn sample set. Nanayakkara then prompted Claude to build a customer churn dashboard, and Claude ran its own queries against Genie and produced the charts, KPIs and a written analysis as an artifact. Afterwards the author asked Claude for the prompt and queries it had used and pasted them into Genie Code to rebuild the dashboard in Databricks AI/BI.

How it went

Claude produced a complete dashboard on the first attempt, and each chart or KPI can be clicked to show the query behind it and the four data sources it drew on. Genie Code rebuilt a matching AI/BI dashboard from Claude's queries. The verdict: Claude's version is more flexible and nicer looking because it is just a responsive HTML page, so it can draw things like a multi-step funnel, but the author would still use AI/BI for production because it is live, versioned and governed in Unity Catalog. The post gives no timings or costs.

Worth knowing

The older Genie MCP URL under /api/2.0/mcp/genie is being switched off on 31 October 2026; use the Unity AI Gateway URL. On Azure you need global admin to create the OAuth app connection, Claude only wants the client ID, and the author expected to re-authenticate about weekly. This was a sample dataset, not production data.

Try it

With Claude· 4 parts to fill in

Connect to my Databricks Genie space [space name] through the connector I have set up. Explore the tables it exposes, then build me a dashboard on [customer churn or another business question] with [number] KPIs and charts that fit the data. Show the query behind every chart and list which tables each number came from. Do not invent fields or fill gaps with guesses; if the data cannot answer something, say so on the dashboard. When it is done, give me the full list of queries you ran so I can rebuild it in [Databricks AI/BI or my BI tool], and stop there.

Square brackets are the parts to fill in. The prompt is ours; the story is theirs.

Where they said it

Related