Databricks Says Genie Ontology Answers Business Questions on Day One
Summary
Databricks says its Genie Ontology returns business answers on day one, but that accuracy depends on the data foundation beneath it. In a September 1 guide, Srujan Alase and Richard Tomlinson lay out six layers, from entity resolution and Unity Catalog metadata to Metric Views and governance. The company recommends starting with one domain.
Key Points
- Databricks says external agents reach Genie's intelligence through MCP.
- Databricks says dbxmetagen uses large language models to draft descriptions that require human approval before Unity Catalog writes.
- Databricks says the /importBI skill in Genie Code builds a Metric View from Tableau or Power BI semantic models.