Databricks said on Nov. 11 it has unveiled a new version of its AI agent, Genie.

TechCrunch reported that Genie supports multi-step tasks that take a long time to complete.

Databricks also built an ontology to provide Genie with context from information customers store on the Databricks platform.

The company said executives can assign Genie a range of tasks and it can finish them without human supervision.

Databricks CEO Ali Ghodsi (알리 고드시) opened the agent he actually uses instead of giving a product demo, showing recent conversations and charts.

Genie analysed low-scoring items in an employee satisfaction survey. It also supported evaluating and filtering improvement proposals from some executives. It summarised how a large sales proposal to a potential customer failed. The summary included features the customer demanded as purchase conditions. TechCrunch reported that Genie also reviewed past broadcasts featuring Ghodsi and generated expected questions for a podcast interview.

Ghodsi said he uses Genie to find things that are hard to hear from the CEO seat. He described it as a channel for hearing unvarnished accounts of what is not going well at the company.

At a small startup, a CEO knows all employees and staff can relay bad news immediately, but Databricks has 12,000 employees. At that scale, failures in the sales field are difficult to reach the CEO.

Ghodsi said, "There are many companies with more than 150 employees," and added, "Many people can get value from Genie."

Databricks is pitching Genie as an always-on agent for executives. Ghodsi said he also uses Genie in his personal life instead of OpenClaw, Instinct and Muse, which he said have questionable security. He said Genie can play the role in business that AI, agents and harnesses play for programmers.

On concerns that agents could increase corporate AI costs, he said, "You can control it with Unity Gateway," adding, "Companies can set usage limits with this product and switch to cheaper open-weight models."