From RAG to Ontology: Databricks Bets on Business Context for AI Trust

At its Data + AI Summit, Databricks unveiled Genie Ontology, a new context layer designed to give enterprise AI agents a shared understanding of business operations. This offering represents a shift from retrieval-augmented generation (RAG) approaches toward more structured knowledge foundations.

genie Ontology automatically extracts business context from various data sources—dashboards, queries, pipelines, documents, and applications—organizing it into a living graph that AI agents can use for decision-making. The system uses a ranking inspired by Google’s PageRank to prioritize authoritative definitions within an organization, ensuring agents rely on the most trusted information.

“One definition feeding every agent means you stop getting three different answers to the same question,” explained Michael Leone of Moor Insights and Strategy. “An ontology gives the agent meaning that simple search can’t—understanding what terms mean and which sources to trust.” This consistency directly addresses a major barrier to AI adoption, where decision-makers often distrust outputs they cannot verify.

While promising, analysts note that ontologies require data readiness. Stephanie Walter of HyperFRAME Research cautioned, “Ontologies improve context but don’t guarantee correctness—agents can still make errors with incomplete or misinterpreted data.” Keeping ontologies accurate as businesses evolve will also be an ongoing challenge requiring clear ownership and governance.

Databricks’ approach arrives amid a growing trend in the enterprise AI space. Competitors like Snowflake and Microsoft have introduced similar context-building capabilities, creating both opportunity and potential confusion for CIOs navigating this evolving landscape.