Databricks Agent Technology Switches To Always On

The next generation of the company’s Genie agentic assistant goes beyond performing specific tasks to autonomously managing long-running work processes and proactively making decisions for people.

Databricks has launched the second generation of its Genie agentic assistant that operates in an “always-on” autonomous mode, running in the background using its own compute resources to perform ongoing tasks.

The improved Genie agent takes ownership of work when an employee gives it a goal, such tracking customer sales trends and product requests and notify employees of significant shifts, according to the data and AI platform giant, or monitoring and optimizing online content for generative engine optimization.

The new Genie can delegate tasks to subagents when needed, follow up on missing inputs, and reach out when a person needs to make a decision.

[Related: Databricks CEO Predicts 10-Year AI Services Boom For Channel]

“Genie is no longer just an AI that you talk to. It’s a long-running agent that can work on tasks autonomously and actually bring things to your attention,” said Ken Wong, Databricks senior director of product management, in an interview with CRN.

The new offering builds on the Genie One product that debuted at the Databricks Data + AI Summit in June.

Wong said an employee can give Genie a broad task like “keep my sales pipeline healthy,” and the agent will ask questions about accomplishing that task, determine appropriate metrics, and so on. “And then [the agent] will go and manage that, bringing to your attention things that it needs your help with and seeking your approval to do work along the way.”

The new Genie is designed to address one of the biggest challenges facing agentic technology: Complex projects often require working across multiple platforms and even accessing employee knowledge. That requires understanding a business, being able to reach systems that teams use, and following correct permissions and access controls.

Key to the improved agent’s functionality is Genie Ontology, a context technology layer that captures everything within a business or organization, from definitions and metrics to people and processes, and utilizes that context information as an agent carries out its assignments.

Ontology, also introduced at the Data + AI Summit event in June and now generally available, can tap into business logic from external business intelligence tools to extend its understanding of analytical data, according to the company. And a new “organizational structure” connector provides both context for understanding people and organizational relationships.

“Genie Ontology is meant to be a horizontal layer that can capture the context of your entire business,” Wong said.

Genie’s Extended Functionality

Genie’s activities are governed by Databrick’s Unity Catalog and Unity Gateway products, so the agent’s actions stay within the permissions, controls and guidelines that companies already have in place.

The new Genie release utilizes sub-agents that Genie can delegate tasks to and work in parallel, offers a priority inbox that helps users understand what items require a decision, and provides email alerts with agent updates. Genie can now also generate dynamic slides, dashboards and interactive reports and keep them current as data changes, according to the company.

Also new is the Genie Search tool that can reach more than 100 data sources including the Databricks Lakehouse and business applications such as Google Drive and Atlassian’s Confluence and Jira software. Genie Search indexes content within those applications, including structured and unstructured data, without moving or copying and ingests it through the company’s Lakeflow Connect or query applications live through MCP.

And a new Ontology Manager helps teams manage and govern business context and flag frequently questioned concepts that require refinement, the company said.

“What all this means for customers is that Genie and any agent that’s connected to the Ontology like Genie, are now smarter and faster and more accurate,” Wong said.

“They’re more accurate because now they have the right context to make the right translation of the user’s question into structured queries and things like that,” he said. “And they also have the ability to do it in a much more [AI] token-efficient manner because instead of searching all over the systems, they can exactly pinpoint where to go. And that means faster response time and fewer tokens being used.”

The expanded Genie can address a wide range of use cases and Wong said that creates opportunities for Databricks partners, including solution providers and systems integrators, to build out the Ontology context layer and develop solutions that take advantage of the agent’s expanded autonomy.

The new Genie agent, Genie Search and Ontology Manager will be available soon in preview and beta programs. In the meantime, Databricks is offering the original Genie and Genies agents free through January to encourage trial adoption and Wong said that also creates partner opportunities.

Leave a Reply

Your email address will not be published. Required fields are marked *