Databricks Updates Genie AI Agent to Give Ghodsi Unfiltered Feedback

Databricks on Wednesday released an updated version of its AI agent Genie, built to carry out multistep work without supervision, and the company's chief executive says he uses it inside his own company to surface problems employees rarely bring to him directly, according to TechCrunch.

Ali Ghodsi, Databricks' CEO, demonstrated the agent to TechCrunch instead of running a standard product pitch, opening his own recent chats and charts with Genie rather than a prepared demo, the outlet reported.

What changed in this version of Genie

The new Genie can perform multistep jobs that take extended work to complete, drawing on an ontology Databricks built to give the agent context from information customers already store on the Databricks platform, according to TechCrunch. Corporate leaders can assign the agent a range of tasks and it will finish them unsupervised, the company told the outlet. The update moves Genie from a query tool into something closer to a standing assistant that runs in the background on a leader's behalf.

How Ghodsi says he actually uses it

Ghodsi showed TechCrunch several real examples rather than staged ones. He said Genie analyzed the sections of an employee satisfaction survey where the company scored poorly, then helped him weigh and discard fixes that some executives had proposed. He said it also summarized a sales pitch to a prospect that went badly, including the specific features the prospect said it needed before it would buy. Separately, he said the agent prepped him for a podcast interview by reviewing his past appearances and drafting likely questions, and that during the TechCrunch interview itself, he had it suggest what he should tell the reporter.

Why a 12,000-person company needs an AI agent for honesty

Ghodsi framed the tool's value around a limit on human relationships rather than a technology gap. "Have you heard of Dunbar's number? The magic number is 150," he told TechCrunch, citing the theory from British anthropologist Robin Dunbar that humans can maintain only a finite number of stable relationships. In a small startup, Ghodsi said, a CEO knows everyone personally and employees feel safe reporting bad news directly. At 12,000 employees, he said, he cannot know everyone, and a salesperson is unlikely to volunteer that a call went badly or an HR manager that a proposed fix will probably fail.

"Anyone who has more than 150 employees, which is a lot of companies, a lot of people can get value out of this," Ghodsi told TechCrunch.

What Databricks is selling beyond its own use case

Ghodsi's internal use of Genie — a practice sometimes called dogfooding — is part of how Databricks is now positioning the product externally, pitching it as an always-on agent for business leaders generally, not just engineers, according to TechCrunch. Ghodsi said he also uses Genie in his personal life rather than what he called security-questionable alternatives, naming OpenClaw, Instinct and Muse. He told the outlet he sees Genie extending to business functions what AI agents and the software "harnesses" that wrap around models have already done for programmers.

How Databricks addresses the cost question

Asked about the risk that agent-driven work could drive up corporate AI spending, Ghodsi pointed to Databricks' Unity Gateway product, according to TechCrunch. He said it lets companies set usage limits and swap in less expensive open-weight models rather than defaulting to the priciest option for every task. The answer positions cost control as a configuration choice for customers rather than an open question Databricks has yet to solve, though TechCrunch's report did not include independent figures on how much agent usage typically costs or how much Unity Gateway can reduce that spending.

Databricks did not respond through the TechCrunch article with additional detail on pricing for the updated Genie or a rollout timeline beyond Wednesday's announcement. The original report is available at TechCrunch.