Share
Databricks wants its AI agent to stop asking for permission. On Wednesday, the company unveiled an updated version of Genie that can take on multi-step jobs requiring extended work, and, according to Databricks, finish them without supervision.
What Databricks actually shipped
The core of the update is something Databricks calls an ontology. Think of it as a detailed map of a company’s data: what each table means, how pieces connect, and which numbers matter for which questions.
Databricks built this ontology layer to pull context from the information customers already store on its platform. That context is what lets Genie tackle a wide range of assignments without someone guiding each step.
The product, called Genie Ontology, is scheduled to become generally available on November 9, 2026.
Alongside it comes Agent mode, which adds deeper reasoning and support for unstructured data. That means files like PDFs and images stored in Unity Catalog volumes, Databricks’ governed storage area, can now feed into the agent’s work.
The rollout also includes new ways to reach the agent. A Chrome extension enters beta on November 10, 2026, followed by a Genie One desktop app for Windows in beta on November 11, 2026.
How the Genie lineup got here
This is not Genie’s debut. Databricks introduced Genie One and Genie Agents at its Data + AI Summit in June 2026, setting the stage for the current expansion.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
We respect your privacy. Unsubscribe anytime.
Genie Agents is the piece that ties most directly into the new ontology. It lets users complete tasks unsupervised by drawing on inferred knowledge about their data along with semantics that users define themselves.
Adoption numbers suggest the earlier releases found an audience. By mid-2026, more than 1.5 million Genie Spaces had been created, giving Databricks a sizable installed base to upsell on autonomy.
The broader strategy is to turn Genie from a single assistant into a family of autonomous, domain-specific agents. Databricks points to business functions such as marketing and finance as targets, where an agent tuned to a specific department could handle that team’s recurring workloads.
Pricing stays usage-based, with no per-seat fees. Organizations can receive up to $10 in free credits per user each month, which lowers the barrier for teams that want to experiment before committing budget.
Why the ontology bet matters
The biggest obstacle for enterprise AI agents has never been raw intelligence. It has been context.
By grounding Genie in a structured understanding of each customer’s data, Databricks argues the agent will make fewer things up. The enhancements are expected to significantly reduce hallucinations, the industry term for AI confidently stating falsehoods.
The research behind the launch suggests that Databricks’ emphasis on contextual efficiency is likely to reduce reliance on expensive token usage, the per-word processing charges that make large language models pricey at scale.
