← News

Databricks Genie Ontology screen showing a map of business domains such as Products, Finance, Sales and Marketing, with top experts and referenced pages in a side panel

11 Nov 2026

Databricks

Databricks turns Genie into an always-on AI coworker that runs jobs on its own

Databricks said on Nov. 11, 2026, that its Genie assistant is now an autonomous AI coworker that handles long-running enterprise work within guardrails admins set, backed by a new Genie Ontology for business context, Genie Search across company systems, and Unity Gateway controls for access, tracing and spending caps.

Every enterprise software company now sells an "AI coworker"; the difference is whether it actually knows what your company means by "active customer." Databricks' edge is that the data already lives there, and the Ontology is its bet on context over raw model smarts. The CEO-truth-serum demo is charming. Spending caps being a launch feature tells you what buyers are actually scared of.

Databricks said its Genie assistant is now an autonomous coworker for long-running enterprise work. Enterprise, here, means a large company. Genie One launched earlier this year. Databricks says thousands of customers already use Genie in production, including Zillow, Unilever, adidas, Atlassian, and AstraZeneca. The announcement is dated Nov. 11, 2026. The product change and the customer list are Databricks's.

The new Genie takes on jobs that stay open. It tracks progress, follows up when an input is missing, and takes action as new information arrives. People set the guardrails, supervise the work, and approve the steps that need a sign-off. A guardrail, here, is a limit the company sets on what the assistant is allowed to do. A job can run on a schedule, or on a trigger, such as a metric dipping or more than 10 customers asking for the same feature. It can run many jobs at once, with a control panel for the ones that are active. Those lines are Databricks's.

People can reach it through Genie One, email, Slack, and Microsoft Teams, and on a browser, a desktop app, iOS, and Android. Plugins carry its business context into ChatGPT and Claude. Those paths are Databricks's.

The examples Databricks gives: watching insurance claims for fraud patterns and routing the case files to underwriters; running marketing campaigns and proposing weekly budget changes for someone to approve; reading your email and Slack to prepare your day. Those examples are the company's.

Genie Ontology is a map of the business: the definitions, the metrics, and the relationships between them. Databricks's line for how it is made is "built by agents and governed by people." A personal memory keeps a user's preferences. Genie Search pulls from Google Drive, SharePoint, Jira, Confluence, and Workday, plus Databricks dashboards and tables. In the chat it can build artifacts. Here that means slide decks, dashboards, and interactive HTML pages. Connectors built on MCP reach Slack, Zoom, Microsoft 365, Jira, GitHub, and a calendar. MCP is a plug that lets the assistant call an outside tool. Shareable skills are stored in Unity Catalog, Databricks's system for governed data. Those lines are Databricks's.

Unity Gateway is the admin side. Admins control which tools Genie can use, trace and audit every action, and set budgets for a workspace and for each user, with spending caps and alerts. Those controls are Databricks's.

TechCrunch watched chief executive Ali Ghodsi use his own Genie. It picked out weak spots in an employee satisfaction survey, summarized a sales pitch that went poorly, and prepared him for a podcast. He cited Dunbar's number, about 150, the rough count of relationships one person can keep. With 12,000 employees, he said, he cannot hear everything directly, and Genie surfaces what people will not tell the boss. On cost, he said Unity Gateway lets a company set usage limits and swap in cheaper open-weight models. An open-weight model is one a company can run itself, weights included, often for less money than a closed model. Those lines are his, as TechCrunch reported them.

The picture is the Genie Ontology screen, a map of business domains such as Products, Finance, Sales, and Marketing, with top experts and referenced pages in a side panel. The credit is Databricks.

In plain terms, Databricks turned Genie from a question box into a coworker that keeps a job running, inside limits an admin sets, and that is supposed to know what the company means by its own words. The customer list and the examples are Databricks's. The spending caps shipped with the product.

RELATED

ONLINE…

Comments

guidelines

Loading…

Loading…

Sources