Databricks on Wednesday introduced an updated version of its AI agent, Genie, which can now execute multi-step jobs that require extended time to complete. The company says the capability relies on an ontology Databricks built to provide Genie with context from customer data stored on the Databricks platform.
Internal demonstrations and use cases
Rather than a staged demo, Databricks CEO Ali Ghodsi demonstrated the agent using his actual internal instance, showing recent chats and charts.
Examples Ghodsi walked through include:
- Analyzing areas of an employee satisfaction survey with poor scores and rating (or discarding) fixes proposed by executives.
- Summarizing a large sales pitch that went badly, including features the prospect said it would require before buying.
- Preparing for a podcast interview by reviewing prior appearances and drafting likely questions; during the conversation Ghodsi even had Genie suggest what he should say to the reporter.
Ghodsi emphasized that Genie helped surface candid information about problems he, as CEO, is less likely to hear directly.
Scale and the limits of executive visibility
Ghodsi referenced Dunbar’s number — commonly cited as around 150 people — which suggests a practical limit to the number of stable relationships a person can maintain. In a small startup a CEO often knows everyone and hears blunt feedback, but at Databricks’ size — Ghodsi noted the company has 12,000 employees — that direct line of candid reporting disappears. A salesperson is unlikely to tell the CEO about a disastrous sales call, and an HR manager may not admit that a proposed satisfaction initiative is likely to fail.
"Anyone who has more than 150 employees, which is a lot of companies, a lot of people can get value out of this," Ghodsi said, speaking both about AI agents broadly and Genie in particular.
Dogfooding and personal use
Ghodsi’s own use of Genie (often called dogfooding) is one reason Databricks now markets it as an always-on agent for business leaders. He also said he uses it personally instead of other alternatives he views as more questionable from a security perspective, naming OpenClaw, Instinct, and Muse as examples.
Cost control via Unity Gateway
Addressing concerns that agents could drive up corporate AI spending, Databricks points to its Unity Gateway product as a control mechanism. Unity Gateway can enforce usage limits and enable companies to swap in less expensive open-weight models to manage costs.
Why this matters
The Genie update shows how enterprise AI agents are becoming capable of coordinating complex, multi-step workflows and of surfacing uncensored feedback that executives may no longer receive directly in large organizations. At the same time, managing costs and ensuring data governance remain central considerations for adoption.
