Databricks' new AI agent acts as a truth-teller for CEOs

Databricks recently updated its AI agent Genie, giving it the ability to execute multi-step tasks using a custom-built enterprise context ontology. During a live demonstration, CEO Ali Ghodsi showed how he relies on the agent for internal operations, including evaluating low scores on employee satisfaction surveys, summarizing poorly executed sales pitches, and prepping for public appearances.

The core value of this tool lies in solving a classic corporate challenge known as Dunbar's number. As companies grow past 150 employees, CEOs lose direct personal connections, and employees become hesitant to deliver negative news upward. Genie acts as an autonomous sounding board that bypasses corporate filters, surfacing operational issues that leaders urgently need to address.

To address enterprise adoption concerns, Ghodsi noted that Genie can run safely on internal company data without privacy risks. Furthermore, companies can manage the rising financial costs of running AI agents by utilizing Databricks’ Unity Gateway, which allows organizations to set strict usage limits and swap in more affordable open-weight models.