Databricks CEO Uses AI Agent to Hear What He Can’t

When Databricks CEO Ali Ghodsi asked his own AI agent what he should say to a reporter, the answer was a candid reminder that even the top executive can’t hear everything. Genie can perform multi‑step jobs unsupervised, thanks to an ontology that gives it context from the data customers store on the platform.

The agent analyzed an employee satisfaction survey, rated suggested fixes, and summarized a failed sales pitch, all while preparing Ghodsi for a podcast interview. Ghodsi used Genie to sift through the survey, evaluate fixes, and distill the key points of the sales pitch. He also let it draft the questions he might face in the interview and even suggested what to say to the reporter.

With 12,000 employees, the CEO can’t know everyone, so Genie surfaces the unvarnished truth that would otherwise stay hidden. The benefit is limited to companies with more than 150 employees, the theoretical cap on stable relationships. Databricks says the cost risk is tamed by Unity Gateway, which lets firms set usage limits and swap in cheaper open‑weight models.

This means the agent can stay always on without blowing budgets.

Our read: Genie’s ontology gives it a contextual advantage that could reshape how leaders get unfiltered insights.

As AI agents become the new sounding board for executives, the question is whether they will replace human intuition or simply amplify it.