AI
Databricks upgrades Genie agent to handle complex multi-step workflows autonomously
The updated AI agent can now complete intricate business tasks without supervision, giving executives access to unfiltered insights on company performance.
TechCrunch — AI
Databricks announced a new iteration of its Genie AI agent that can autonomously manage complex, multi-step assignments requiring sustained effort to complete. The enhanced system leverages a purpose-built ontology designed to supply the agent with contextual understanding drawn from information stored within the Databricks ecosystem.
The company contends that corporate decision-makers can now delegate a diverse array of responsibilities to the agent, which will execute them without ongoing human oversight. CEO Ali Ghodsi demonstrated the capability by sharing his personal workflow with Genie, offering real examples rather than a polished marketing presentation.
How Ghodsi uses Genie internally#
Among its applications, Ghodsi showed how the agent analyzed weak performance areas flagged in employee satisfaction assessments and helped him evaluate (and reject) proposed solutions from other executives. The system also synthesized a poorly-received sales presentation, capturing what features a prospective customer indicated were prerequisites for a purchase decision.
Beyond operational analysis, Genie prepared Ghodsi for a podcast appearance by combing through his past interviews and predicting probable questions. During the conversation itself, he even tasked it with recommending how to respond to journalist inquiries.
Why Genie serves as a "truth-telling" resource#
The most compelling aspect of Ghodsi's use case was his reliance on Genie to surface uncomfortable truths that organizational rank might otherwise suppress. As a CEO overseeing 12,000 employees, Ghodsi acknowledged he cannot personally maintain the relationships necessary to hear direct criticism. A struggling salesperson is unlikely to volunteer failed call details to the top executive; an HR leader won't candidly admit that their retention strategy may backfire.
Ghodsi referenced Dunbar's number—anthropologist Robin Dunbar's concept that humans can maintain stable relationships with roughly 150 people. Beyond that threshold, candid feedback dries up. He suggested that any organization exceeding this boundary could benefit from an agent-based solution to recapture that transparency.
Broader market positioning and cost management#
Databricks is positioning Genie as a permanent intelligence layer for senior leadership, emphasizing that Ghodsi's hands-on dogfooding validates the product. He noted using it in personal contexts as well, positioning it as a preferable alternative to other consumer AI platforms.

