Databricks turns Genie into an always-on agent for CEOs and managers

Databricks on Wednesday unveiled an updated version of its Genie AI agent that can complete multistep jobs without supervision, using an ontology built from customer data stored on the Databricks platform. Chief executive Ali Ghodsi demonstrated the tool on his own internal chats, including employee survey failures and a poor sales pitch. The shift matters for workers because managers may soon rely on agents, not staff reports, for unfiltered assessments of performance.

Genie runs unsupervised tasks from company data

Corporate leaders can now assign Genie a wide range of tasks that take extended work to finish, and the agent will complete them unsupervised, the company says. The capability rests on an ontology that Databricks built to give Genie context from information customers store on its platform. To illustrate the use, Ghodsi opened his actual agent rather than a staged product demo and walked through recent chats and charts. He showed how it analyzed the low-scoring parts of an employee satisfaction survey and helped him rate fixes proposed by executives, including discarding some suggestions.

He also showed how Genie summarized a large sales pitch to a prospect that went poorly, including the features the prospect said it needed before buying. It prepared him for a podcast interview by reviewing his previous appearances and developing likely questions. During the conversation with TechCrunch, he even asked the agent what he should say to the reporter. Ghodsi said he also uses Genie in his personal life, rather than what he described as a security-questionable alternative like OpenClaw or Instinct or Muse. He believes Genie can do for the business side what AI, agents and harnesses are doing for programmers.

Why managers want unfiltered internal reports

For employees and middle managers, the practical change is how bad news travels upward. Ghodsi used Genie to find problems he said he is now unlikely to hear directly as head of a company with 12,000 employees. A salesperson is unlikely to volunteer details of a disastrous sales call to the chief executive, and an HR manager is unlikely to admit that a proposal to raise satisfaction scores will probably fail. An agent that reads surveys, sales notes and past public statements can surface those weak points without requiring staff to deliver them in person, which changes what gets discussed in reviews and planning meetings.

The story does not prove that Genie works the same way on other companies data or that unsupervised results are consistently accurate. It is a company demonstration centered on the chief executive own use, often called dogfooding, and Databricks is now pitching the product as an always-on agent for business leaders. Ghodsi cited Dunbar number, the theoretical limit of 150 stable relationships described by anthropologist Robin Dunbar, and said companies above that size could get value from such agents. On costs, he said spending can be controlled through Databricks Unity Gateway product, which sets usage limits and allows swapping in less expensive open-weight models.