How would it help your business if your data agent answered 84.5% of questions right on the first attempt - when the best general-purpose coding agents manage around 52%? This is the claim Databricks is making with their Genie Ontology supercharging their AI-coworker Genie One, which can be connected to Teams and Slack for easy coworking. Databricks is bringing the knowledge graph out of deep theory and using it to generate value for customers.
Knowledge graphs and ontologies have been around for over twenty years, but efforts to apply them in the enterprise context have been lacking. Databricks is looking to change this by bringing ontologies into the BI world with its new offering of Genie Ontology. Genie Ontology uses the structures of Unity Catalog to help build a knowledge graph that represents your organization, and links data outside of traditional BI from Slack, Jira, Google Drive, SharePoint and other apps to your data platform context.
Visibility into the underlying knowledge graph is limited today - the ontology itself is in a gated preview - so if you're evaluating this as a replacement for an existing ontology or catalog, it's unclear yet whether it fits the bill. The additional enterprise data in Genie can provide an improvement on the accuracy of the agent by providing solid context for its operation, so let’s dig into what this is, what it isn’t.
Building enterprise ontology with Databricks Unity Catalog
Genie Ontology is part of a pipeline of releases which position Databricks as a provider of agentic data platform solutions:
Unity Catalog's semantic layer is arriving in pieces. Domains and Pages are available now in Beta - a Page is a governed, authoritative definition of a business term or KPI that Genie One prioritizes over inferred context and cites in its answers. Genie Code can even bulk-import Pages from existing documentation. The full Business Glossary is still coming, but the mechanism for feeding human knowledge into the Ontology is already here - and it's exactly the groundwork worth starting today. (One note for regulated customers: Page content is stored unencrypted, so definitions only - no sensitive data.)
Rethinking where your reports live
Databricks has invested heavily in AI development, and one new feature is the ability to take existing reports from Power BI and Tableau and generate Databricks Dashboards served by Databricks Metrics inside the Databricks application. For organizations committed to Databricks, this can change the math on keeping additional applications mainly to serve BI reports: Dashboards now run natively on the same governed, secured platform, and Databricks has shipped a potential migration path from existing BI tools. Fewer boundaries mean the data engineering and analytics team can move faster and worry less about cross-system integration and think more about how to add value to the organization. To help teams try out these new features, Databricks has paused invoicing on Genie One and Genie Agents through January 31, 2027.
Agents for building and running the platform
Databricks has also released several products to help development and maintenance on the platform. Genie ZeroOps is your own application maintenance service in the data platform. It can troubleshoot pipeline failures and propose solutions through a sandboxed dev environment before a human looks at anything. Genie Code offers similar capabilities to best-in-class agent harnesses such as Codex and Claude Code, enabling your teams to make progress on their tasks quickly.
Unity AI Gateway answers the governance question every organisation running agents eventually hits: who is spending what, on which model, and are they allowed to. Hard spend caps, budgets per user and team, content filtering, and approval policies for sensitive agent actions.
Caveats: What is available today?
Most of this is brand new. Genie One, Genie Agents, and Genie Code are GA - but Genie Ontology itself runs behind the scenes. You can't open it, inspect it, or explore it. The ways to feed the Ontology are only just arriving - Pages in Beta, Business Glossary still to come - so there's a real gap between what Databricks demoed on stage and what you can put into production today.
That gap is the window. The Ontology learns from what exists: certified tables, governed metrics, a curated glossary. Customers who build that foundation now will get value on day one of access. The ones who wait will spend the preview period cleaning up their catalog instead.
That said, this direction of development represents an exciting opportunity to reshape the data and analytics platform landscape. Databricks is offering a coherent view for enterprises on how to adopt and remain AI-ready.
