Databricks updates Genie agent to run multi-step business tasks unsupervised, built on a new ontology

Databricks on Wednesday unveiled an updated version of its AI agent Genie, which can now perform multi-step jobs that take extended work to finish. According to the company, corporate leaders can assign the agent a wide range of tasks and it will complete them unsupervised. TechCrunch reported the news after a walkthrough with CEO Ali Ghodsi.

How Genie gets its context

Genie relies on an ontology that Databricks built to give the agent the context it needs from the information customers store on the Databricks platform. The company has not tied the update to a specific underlying model in this announcement.

Ghodsi's own use

Instead of a scripted product demo, Ghodsi opened his actual agent and went through his recent chats and charts, according to TechCrunch. Examples he showed:

• Employee survey analysis: Genie analyzed the parts of an employee satisfaction survey where Databricks did not score well. It then helped him rate, and discard, fixes some executives had suggested for those low scores.
• Sales call summary: It summarized a large sales pitch to a prospect that went poorly, including the features the prospect said it needed before buying.
• Podcast prep: It reviewed his previous appearances and developed likely interview questions.
• Interview advice: During the conversation, it told him what to say to the reporter. TechCrunch's assessment: the advice was not bad.

The pitch: candid information at scale

Ghodsi framed the use case around Dunbar's number, the theoretical maximum of about 150 people with whom humans can maintain stable relationships. With 12,000 employees, he said he cannot know everyone, and a salesperson is unlikely to confess a disastrous call to the CEO. An HR manager is unlikely to admit an idea for raising satisfaction will probably fail. Genie, he indicated, surfaces information he is now unlikely to hear directly.

"Anyone who has more than 150 employees, which is a lot of companies, a lot of people can get value out of this," Ghodsi said of AI agents generally and Genie specifically.

Databricks is now pitching Genie as an always-on agent for business leaders. Ghodsi said he also uses it in his personal life, rather than alternatives TechCrunch described as security-questionable: OpenClaw, Instinct or Muse. He believes Genie can do for the business side what AI, agents and harnesses (the software that wraps around a model and lets it use tools) are doing for programmers.

Cost controls

On the risk that agents raise corporate AI costs, Ghodsi pointed to Databricks' Unity Gateway product. It lets companies set usage limits and swap in less expensive open-weight models.

What this means

The update moves Genie from answering questions about data toward finishing extended, multi-step work, with the ontology as the mechanism that supplies business context from data already on the Databricks platform. The strongest evidence offered is Ghodsi's own use, which is a vendor's internal example, not an independent evaluation. The stated approach to cost is to cap usage and substitute cheaper open-weight models through Unity Gateway, which suggests Databricks expects spend on long-running agents to be a buying concern. Pricing and details of the underlying models were not given in the announcement coverage.