The Pulse
Databricks plans always-on Genie agent for enterprise work
Databricks is preparing an always-on version of Genie that can manage ongoing assignments and ask employees for decisions. The company says the new agent is coming to preview and beta programs, while its publications give conflicting availability information for Genie Search.
Work that continues in the background
Databricks is preparing an always-on version of Genie that can take responsibility for ongoing enterprise assignments, delegate work and ask employees for decisions. The second-generation agent and a tool for managing its business context are coming to preview and beta programs, the company said in its November 11 announcement.
The company says Genie will run on its own computing resources, track progress and follow up when information is missing. Employees will set a goal and boundaries for the work; a priority inbox will surface items needing attention, while email alerts will link back to the app and group subsequent updates in the same thread.
Its proposed uses range from monitoring email and preparing for meetings to tracking a quarterly forecast or customer usage patterns. The new agent is designed to handle multiple jobs, use sub-agents for parallel tasks and run on a schedule or in response to triggers.
Business context and search
Databricks says the agent will work from Genie Ontology, a layer that brings together a company’s definitions, metrics, people and processes. The existing Genie product and the ontology are generally available to customers, separately from the forthcoming always-on agent.
The company says the ontology learns from work such as dashboards, queries and notebooks, and is being extended to business logic in external analytics tools and organizational relationships. The forthcoming Ontology Manager is intended to let teams review that context, identify gaps and refine definitions.
The announcement also describes Genie Search, which brings information from company data and business applications into a search experience for people and agents. It can index content where it resides, ingest it through Lakeflow Connect or query applications live through MCP connections; the company lists Google Drive, Jira and Confluence among the applications.
Its availability is described differently in Databricks’ own publications. The press release lists Search as coming soon to preview and beta programs, whereas a separate company blog post and the platform release notes describe it as generally available.
Permissions and spending controls
Databricks says the agent will operate within a customer’s existing environment and use Unity Catalog permissions, including source-system permissions for indexed content. Employees can decide which actions the agent performs independently and which require their approval, within the limits set by their organization.
Unity Gateway lets administrators set tool access, approval rules and spending limits, and trace actions, according to the company. Customers pay for Genie based on usage rather than by seat, and administrators can set limits by team or user.
The company also describes data-driven slides, dashboards and interactive reports that update as their underlying data changes. Genie Agents can work across documents in Unity Catalog volumes alongside tables, while OpenSharing allows an agent to be shared with partners without exposing the underlying data.