An AI coworker that doesn’t wait to be asked
Most workplace chatbots sit idle until someone types a question. On November 11, 2026, Databricks announced a different idea: the second generation of Genie, which the company describes in its press release as an always-on, autonomous agent for enterprise work. Instead of answering one prompt at a time, Genie is meant to take a goal, own the work, and come back only when a person has to decide something.
That shift matters because it moves AI from a search box to a teammate that runs in the background. It also raises the obvious question every company asks first: how do you let software act on your behalf without losing control? Databricks’ answer is to build governance into the design rather than bolt it on.
What Genie now does on its own
According to Databricks, Genie can run in an autonomous mode on its own compute. A user shares a goal and clear boundaries, and Genie manages the task, delegates parts of it to sub-agents working in parallel, tracks progress, and follows up when inputs are missing.
Two interface changes make that practical. A priority inbox now greets users with the items that need attention, so they can see pending decisions instead of digging through old conversations. And each new alert arrives as its own unread email, with later updates added to the same thread so the inbox doesn’t fill up. Each email links back into the Genie app.
Databricks lists examples of the work Genie can take on: staying on top of email or Slack, preparing for meetings, keeping a quarterly forecast on track, monitoring customer usage patterns for a success team, or delivering a weekly brief on revenue trends and top feature requests. Marketers could ask it to monitor performance and suggest content.
Genie Ontology: giving the agent business context
The company argues that an assistant can only take on complex work if it knows how the business actually runs. Much of that knowledge lives across platforms and in people’s heads: which accounts need attention, what triggers an escalation, who approves a change.
Genie Ontology is Databricks’ answer. It captures what the company calls horizontal context, including definitions, metrics, people and processes, and keeps it current as the business changes. It draws on dashboards, queries, notebooks and external BI tools, and a new organizational-structure connector adds an understanding of people and reporting relationships. A new Ontology Manager lets teams govern that context in one place, flagging frequently asked concepts that still need a clear definition.
In the release, CEO Ali Ghodsi said the agent knows “what we mean by revenue, which numbers are official, and who can see what.”
Genie Search and the open ecosystem
Genie Search is built to give people and agents one governed way to find information across the lakehouse and business apps such as Google Drive, Jira and Confluence. Databricks says it offers more than 100 connectors, and teams can ingest content through Lakeflow Connect, index it in place, or query applications live through MCP. The company says a single search, instead of checking each app separately, cuts unnecessary tool calls and token use, which should make agents faster and cheaper to run.
Genie can also be reached from a browser, a mobile app, email, and AI tools including ChatGPT, Claude, Microsoft Copilot Cowork and Grok Bot, through Genie MCP. New live, governed artifacts let it generate slides, dashboards and interactive reports that stay current as the underlying data changes.
Governance and availability
Because Genie runs inside a customer’s existing Databricks environment, it follows Unity Catalog permissions and Unity Gateway approval rules and budgets, and every action is auditable. Databricks says admins can set spending limits by team or user.
On availability, Databricks states that Genie and Genie Ontology are generally available to all customers, while Genie Search, Ontology Manager and the new always-on Genie are coming soon to preview and beta programs. Customer comments in the release, from Headspace and HelloFresh, are company-supplied; HelloFresh cites a 75% reduction in the time to identify ad winners on Meta.
What it means
For data and IT leaders, the story is less about one product than about a pattern: agents are being tied to a shared business-context layer and to existing permission systems. Teams evaluating this kind of tool should look at who can see what, how approvals and budgets are enforced, and how auditable each action is. Since the always-on mode is not yet generally available, the practical advice is to watch the preview program and test on a narrow, low-risk workflow first.
Conclusion
Genie’s second generation shows where enterprise AI is heading: proactive, context-aware and bounded by governance. Whether it delivers on that promise will depend on preview results over the coming months.