Databricks has officially unveiled an updated version of its AI agent Genie, designed to perform complex, multistep jobs that require extended work periods to finish. The system operates through an enterprise ontology built by Databricks, providing the agent with essential context sourced directly from the information customers store on the platform. Corporate leaders can assign a wide variety of tasks to the agent, which then completes them completely unsupervised.
To demonstrate the practical application of the newly revised technology, Databricks CEO Ali Ghodsi bypassed a standard product demo during a recent walkthrough. Instead, Ghodsi fired up his actual internal agent to review his most recent chats and charts. He revealed how the tool analyzed poorly scoring sections of an employee satisfaction survey, and how it assisted him in rating and ultimately discarding flawed fixes suggested by various executives.
Key takeaways
- Databricks unveiled an updated version of its Genie AI agent capable of handling complex, unsupervised multistep jobs.
- CEO Ali Ghodsi uses Genie internally to evaluate employee satisfaction surveys and review failed sales pitches.
- The agent functions through an ontology built by Databricks to extract context from customer data stored on the platform.
- Ghodsi uses the tool to overcome organizational hierarchy and surface unvarnished operational feedback across a 12,000-person workforce.
Real-World Executive Applications and Testing
Beyond internal surveys, Ghodsi demonstrated how the agent summarized a major sales pitch to a prospective client that went poorly, noting the specific product features the prospect demanded before making a purchase. The AI also prepped the chief executive for an upcoming podcast interview by reviewing his previous media appearances to develop likely questions. While they talked, Ghodsi even had the system suggest talking points for the interview itself, noting that its advice proved remarkably helpful.
The underlying motivation for Ghodsi's internal use stems from the challenge of gathering unvarnished operational feedback as a company scales. Referencing Dunbar's number—the anthropological theory that humans can maintain stable relationships with roughly 150 individuals—Ghodsi explained the communication barriers inherent in a massive enterprise. At a headcount of 12,000 employees, the chief executive acknowledged he cannot know everyone personally, meaning salespeople are unlikely to confess a disastrous sales call directly to him, and human resources managers are unlikely to admit that an employee satisfaction proposal might fail.
Scaling AI Agents and Controlling Costs
Ghodsi positions Genie as an always-on solution for business leaders facing similar organizational gaps, noting that any company exceeding 150 employees can extract significant value from such agents. By practicing internal dogfooding with his own platform rather than turning to security-questionable alternatives, Ghodsi believes Genie can transform business operations in the same way coding harnesses empower software developers.
Addressing potential enterprise concerns regarding runaway expenses from autonomous AI workloads, Ghodsi pointed to Databricks' Unity Gateway product. The gateway allows corporations to enforce strict usage limits and swap in more affordable open-weight models to keep operating expenditures under control.
Comments