Sunday, November 15, 2026 Independent reporting on artificial intelligence Newsletter
AI News Beats
AI for Business

Databricks previews an always-on Genie agent for enterprise work

The company says its new Genie agent will track work in the background and flag decisions for staff, but the always-on features are not generally available.

A laptop screen with data charts and lines of code
Photo: Daniil Komov / Unsplash

Databricks announced a second generation of Genie on November 11 that it says will keep working on assignments in the background and ask employees for decisions as needed. The company describes the system as an enterprise agent for tasks that last longer than a chat session, including tracking metrics and following up when inputs are missing.

The announcement separates the products customers can use now from those still to come. Databricks says Genie and Genie Ontology are generally available to its customers. The new always-on version of Genie, along with Genie Search and Ontology Manager, is due in preview or beta programs. The announcement does not specify a general release date for those features.

How a background assignment would work

Databricks says an employee can give Genie a goal and boundaries, then leave the agent to track progress using its own computing resources. The examples in the announcement include watching changes in customer usage and preparing a recurring summary of revenue trends. The company says Genie can split a job among smaller agents, which handle parts of the work in parallel.

The planned interface includes an inbox for decisions that need attention and email alerts for people who are away from the application. Updates about an alert go into the same email thread, according to Databricks. For a team overseeing a long-running job, the relevant measure would be how often the alerts point to a real decision, rather than simply how many updates the system generates.

Company definitions and permissions

Genie Ontology captures how a company defines metrics and relates its data, processes and people. An ontology is a structured map of those relationships. For example, a term such as active customer may have a formal definition in a finance report that differs from its everyday use in a sales conversation. The system is intended to use that formal context when interpreting a request.

Databricks says Genie works inside a customer’s existing Databricks environment and uses controls from Unity Catalog and Unity Gateway. The proposed Genie Search will span more than 100 enterprise data sources, according to the company, including systems such as Google Drive, Jira and Confluence. Databricks says the search feature follows existing access rules and provides sources that teams can inspect.

What customers still need to evaluate

The release describes how Genie is designed to work, not what results an individual customer should expect. For teams considering a trial, a useful first assignment would have a clear definition of success, a limited set of data and an identifiable person who can answer questions or stop the job.

There is also a difference between a useful summary and a reliable ongoing process. A background agent needs to recognize when a source is stale, when two teams use different definitions or when it lacks enough information to proceed. The preview will give customers a chance to see how Genie handles those cases before entrusting it with work that has broader consequences.

Stay on top of AI

Get the stories that matter in artificial intelligence, straight to your inbox.

Free. One email with the stories that matter. Unsubscribe any time. See our privacy policy.