Databricks' new Genie AI coworker hits 91% on Turing benchmark

Databricks positions the new Genie as a step beyond AI assistants that run analyses or summarize meetings but leave employees to manage the surrounding work. The company says most enterprise work is continuous, complex, and data-intensive, which is why Genie is designed to stay on a job, follow up on missing inputs, and act as new information arrives. That design depends on business context: Genie Ontology supplies definitions, metrics, relationships, and organizational knowledge, while Genie Search and Unity Gateway and Unity Catalog supply governed information retrieval and access controls.

The benchmark with Turing is the body's main evidence for this approach. Databricks compared Genie, powered by Genie Ontology, with two leading frontier agents on the same tasks in the same workspace. Genie answered correctly in 91% of task runs, 1.6 times the rate of the stronger frontier agent, while using less than half as many tokens. The tasks required connecting knowledge spread across tables, dashboards, and notebooks, not simply accessing data, which suggests the gain is tied to assembling context rather than raw model capability.

Databricks also expands how teams can work with Genie: through Genie One, email, Slack, and Microsoft Teams; through artifacts such as slide decks and dashboards built in conversation; and through MCP connectors to tools like Slack, Zoom, Microsoft 365, Jira, and GitHub. Genie Agents extend specialized expertise to entire teams, and Genie plugins bring the same business context into AI tools such as ChatGPT and Claude. Whether these additions translate into broad adoption beyond the thousands of customers already using Genie in production is not established in the body.