NitronEdge TechnologiesHelping Enterprises Harness Innovation
Data · Databricks

Databricks Genie goes always-on with Ontology and Search

Databricks has launched a second-generation Genie that works in the background. What changed, the controls involved and what data teams should do next.

Databricks has released the second generation of Genie, its AI agent for business users, and moved it from a question-and-answer tool towards something that keeps working in the background. The announcement came on 11 November 2026. Genie and a new layer called Genie Ontology are generally available to all Databricks customers now. Genie Search, an Ontology Manager and the always-on mode itself are listed as coming soon in preview and beta.

What changed

The headline idea is simple. Instead of waiting for a person to type a question, a user gives Genie a goal and some limits, and Genie carries on with the task using its own compute. It can split a large piece of work across several sub-agents running in parallel, keep track of progress, ask people for missing inputs, and stop to request a decision when one is needed. Databricks gives examples such as sending a weekly revenue brief to executives, watching product usage for account teams, or keeping a quarterly forecast on track.

Results arrive in a priority inbox. Each new alert comes as its own email, and later updates stay in the same thread with a link back to the Genie app. Genie can also produce slides, dashboards and reports built on live data.

Genie Ontology is the part to watch for data teams. It stores business definitions, metrics, people and processes, and Databricks says it updates as the business changes. Genie uses it to ground what it does, so that "active customer" or "net revenue" means the same thing every time. Definitions from external BI tools can feed into it, and a new connector maps the organisation chart and reporting lines. The upcoming Ontology Manager will point out concepts that people ask about often but which lack a clear definition.

Genie Search extends the agent beyond the lakehouse to business apps such as Google Drive, Jira and Confluence. Content can be indexed where it sits, brought in through Lakeflow Connect, or queried live through MCP, and Databricks cites more than 100 connectors. Genie Agents can also read PDFs, contracts and slides stored in Unity Catalog volumes without a separate vector index.

People can reach Genie through the web, mobile, email and tools such as Slack, ChatGPT, Claude and Microsoft Copilot. Pricing is based on usage, not seats.

Why it matters

Most enterprise AI pilots stall at the same point. A chatbot that answers questions is easy to demo, but an agent that acts on its own needs clear definitions, permissions, cost limits and an audit trail before anyone will trust it with real work. Databricks has tried to answer each of those points inside the platform.

Genie runs within the customer's existing Databricks environment. Its actions follow Unity Catalog permissions, so an agent cannot see or change data the user could not. Approval rules and budgets are set in Unity Gateway. Admins choose which tools Genie may use, set spending caps by team or by user, and can audit every action it takes. Genie Search respects the access rules already in Unity Catalog, which avoids building a second search index with its own security model.

The ontology matters as much as the agent. An autonomous agent working from vague or conflicting metric definitions will produce confident but wrong answers, and it will do so repeatedly. Teams that have already invested in a semantic layer or a clean set of certified metrics will get far more out of this release than teams that have not.

There are open questions too. Always-on agents consume compute even when nobody is watching, so usage-based pricing needs firm caps from day one. Connecting Genie to email, Slack and document stores also widens the set of data an agent can touch, which raises the stakes on access reviews.

What this means for organisations in the Middle East, India, Africa and North America

For many organisations in the UAE, Saudi Arabia, India and South Africa, the deciding question is where the agent runs and where the data goes. Because Genie works inside the customer's own Databricks workspace and follows Unity Catalog controls, residency follows the workspace region rather than a separate service. Teams should still confirm which cloud regions the preview features reach, and whether any connector or external AI tool sends data outside the country, before switching them on.

Regulators are paying closer attention to automated decisions. Under the UAE and Saudi personal data protection laws and India's DPDP Act, organisations need to show what personal data an automated process used and why. Genie's action audit helps, but it only works if the audit logs are kept, reviewed and linked to a named owner. In the USA and Canada, teams aligning with the NIST AI RMF or OSFI guideline E-23 on model risk will want each always-on agent recorded in their AI inventory, with its goal, its data sources and its limits written down.

For mid-sized firms with small data teams, the ontology may be the bigger win. Writing down what key metrics mean, once and in one place, pays off whether or not an agent is involved.

What to do next

  • Pick two or three high-value, low-risk goals, such as a weekly sales summary or a usage alert, and pilot Genie on those before anything that changes records.
  • Review your metric definitions and certified datasets first. Feed them into Genie Ontology and fix gaps before agents start relying on them.
  • Set spending caps per team in Unity Gateway and agree who approves raising them.
  • Check Unity Catalog permissions for the users and groups who will start agents, since Genie will act with exactly those rights.
  • Add each always-on agent to your AI risk register with an owner, a purpose and a review date.

NitronEdge helps teams plan and run Databricks work of this kind. See our Databricks partner page or our agentic AI services for more.

Sources

  1. Databricks Launches New Always-on Genie: Autonomous AI for Complex Enterprise Work (2026-11-11)
  2. Databricks' new AI agent tells CEO Ali Ghodsi the unvarnished truth (2026-11-11)
Related: Databricks
Enquire

Talk to us about Databricks

Get a free consultation from our AI, data and security specialists. Honest advice, clear options and a reply within 6 hours.

Reply timeOur team replies within 6 hours

Request a callback

Free consultation. Our team replies within 6 hours.

Your details are safe. Our team replies within 6 hours. By sending, you agree to our privacy notice.

WhatsApp