Databricks Genie agent delivers unfiltered intel to CEO
Original: Databricks’ new AI agent tells CEO Ali Ghodsi the unvarnished truth
Why This Matters
Agentic AI moving into C-suite decision loops signals enterprise adoption is deepening fast.
Databricks on Nov. 11, 2026 unveiled an updated Genie AI agent capable of multi-step, unsupervised tasks. CEO Ali Ghodsi uses it live to surface failed sales calls, low employee survey scores, and frank exec feedback at scale.
Databricks launched a major update to its AI agent Genie on Wednesday, adding the ability to handle extended, multi-step jobs without human supervision. The agent draws context from an ontology Databricks built on top of data customers already store on its platform.
CEO Ali Ghodsi skipped the polished demo and opened his actual Genie instance for TechCrunch, walking through real recent sessions. The agent had analyzed low-scoring sections of an internal employee satisfaction survey, evaluated (and discarded) executive proposals to fix those scores, summarized a failed sales pitch including specific missing features the prospect demanded, and prepped Ghodsi for podcast interviews by reviewing past appearances.
Ghodsi frames the tool's value around Dunbar's number—the theoretical 150-person ceiling for stable human relationships. At 12,000 employees, no salesperson is confessing a blown deal directly to the CEO. Genie, he argues, fills that gap. "Anyone who has more than 150 employees—a lot of companies—can get value out of this," he said.
On cost concerns, Ghodsi pointed to Databricks' Unity Gateway, which lets companies cap agent usage and swap in cheaper open-weight models. He also said he uses Genie in his personal life instead of consumer AI tools he considers security risks.