Databricks Powers Genie One With Genie Ontology

Databricks has detailed how its internal product teams are utilizing Genie Ontology to power Genie One, an AI assistant designed to overcome the context limitations of general-purpose models. While standard AI agents often struggle with company-specific data, Genie Ontology provides a continuously updated framework of institutional knowledge. It integrates business definitions, rules, and relationships directly from Unity Catalog, combining human-curated data with learned context from dashboards, notebooks, and SQL queries.

To ensure accuracy, the system uses a ranking algorithm called OntoRank, which evaluates and prioritizes information based on authority signals like certification, usage, and authorship. Genie One also leverages the Model Context Protocol (MCP) to connect with external tools. This allows the agent to search live Google Drive documents, query Jira tickets, and pull field feedback from Slack, combining search capabilities with live SQL analysis of usage tables.

For product managers and data practitioners, this integration automates complex workflows like weekly user adoption reviews. Instead of manually querying databases and cross-referencing stale dashboards, users can prompt Genie One to generate standardized reports, forecast trends, and flag specific accounts. The platform also allows users to inspect citations, view automatically learned "ontology snippets," and convert successful query conversations into shareable, specialized Genie Agents for their teams.