Databricks has rolled out a major upgrade to its AI agent, Genie, enabling it to handle complex, multi-step assignments without supervision. The upgrade hinges on a new ontology that helps Genie interpret the data stored in customers’ systems, giving it necessary context for more nuanced analysis. The company claims this lets the agent execute tasks from start to finish with minimal direction.
Ghodsi’s Internal Use: From Survey Feedback to Sales Pitches
CEO Ali Ghodsi demonstrated just how deeply he relies on Genie. He showed how it analyzed areas where employee satisfaction scores were weakest, assessed and rejected proposed remedies that leadership initially pushed, and summarized where a sales pitch failed—detailing the features a potential client felt were missing. Genie also helped Ghodsi prep for a recent podcast by reviewing his past appearances and predicting likely questions. At one point, it even suggested what he might say during the interview itself.
What makes Genie especially notable is how it surfaces insights Ghodsi might otherwise never hear: candid feedback that gets filtered out as an organization grows. At 12,000 employees, Ghodsi acknowledged he’s far removed from the day-to-day ups and downs. Genie operates like a virtual truth-teller, giving leaders visibility into problems—both internal and external—that usually stay behind closed doors.
Why Genie Matters, and How Databricks Manages Risks
Databricks positions Genie not just as a tool for executives, but as an “always-on” AI agent that delivers business value at scale. Ghodsi likens its potential to what software that wraps AI models and tools is doing for developers—aiming to streamline workflows, amplify insight, and reduce blind spots in decision-making. He believes any company with over about 150 employees will benefit, given that size marks where direct feedback loops often break down.
Cost control and responsible deployment are part of the pitch too. Genie uses the Unity Gateway product to enforce usage limits, and offers the ability to swap in less expensive open-weight AI models to reduce expenditures. Databricks argues its approach strikes a balance between value and accountability (security, cost, and performance).
Genie’s enhancements represent one of the latest shifts in enterprise AI: agents that don’t just assist, but illuminate what’s wrong, not just what’s visible or favorable. As businesses battle the growing pains of scaling culture, communication, and customer focus, tools like Genie are positioned to fill the gaps that even strong leadership often can’t reach alone. What to watch: how well Genie handles biases in reporting, how secure and private its analyses are, and whether the promise of “uncensored truth” holds up when AI agents themselves become part of the organizational structure.
