Databricks CEO demonstrates AI agent's power in internal operations

Databricks recently introduced an enhanced version of its AI agent, Genie, capable of handling complex, multi-stage tasks by leveraging an internal ontology built around customer data on the platform. Databricks CEO Ali Ghodsi showcased the agent's utility by demonstrating its use in his own role, moving beyond standard product demonstrations. During the demonstration, Ghodsi showed how Genie analyzed employee satisfaction survey results to help him evaluate and discard suggested fixes. The agent also summarized a poorly received sales pitch, detailing the prospect's unmet needs, and prepared him for a podcast interview by reviewing past appearances and predicting questions. Ghodsi noted that the agent provided an unvarnished perspective on internal issues that might otherwise remain hidden. He explained that in a company of 12,000 employees, the concept of Dunbar’s number—the theoretical limit of stable relationships at 150 people—highlights why an AI agent is valuable, as not everyone feels comfortable sharing negative feedback with top leadership. Ghodsi stated that this capability benefits any organization exceeding 150 employees. Furthermore, he mentioned that costs associated with these agents can be managed through Databricks’ Unity Gateway product, which allows for usage limits and the integration of less expensive open-weight models.