Databricks on Wednesday announced an updated version of its AI agent Genie. This allows you to run multi-step jobs that take a long time to complete. This is done through an ontology built by Databricks to provide Genie with the necessary context from the information customers store on the Databricks platform.
Company leaders can assign agents a wide range of tasks, and agents can complete those tasks without supervision, the company said. To prove the point, Databricks CEO Ali Ghodsi showed me how they use it internally. Instead of a product demo, he fired up a real agent and walked me through the latest chats and charts.
He taught me how to analyze where a company scores poorly on employee satisfaction surveys, and how it can help you evaluate (and discard) fixes to low scores that some executives have suggested. He tells me how he summed up a major failed sales pitch to a prospect, including the features the prospect said they needed before making a purchase.
I was also able to prepare for the podcast interview by looking back at previous appearances and creating questions to anticipate. While we were talking he told me what to say. (And the advice on that one wasn’t bad.)
What’s interesting is that Godi was using Genie to find out things he wouldn’t be able to hear now as the big boss. It was a source of unvarnished truth about what was going wrong that he needed to address.
“Have you ever heard of Dunbar’s number? The magic number is 150,” he said when I pointed it out. Named after British anthropologist Robin Dunbar, it refers to the maximum number of people a person can theoretically maintain a stable relationship with. In small startups, the CEO knows everyone, has direct relationships with everyone, and feels comfortable communicating negative information openly.
The company has 12,000 employees, and Godoshi is the first to admit that he doesn’t know everyone. At that size, it’s unlikely that a salesperson will confess to the CEO that they made a disastrous sales call. Human resources managers are unlikely to confess that their ideas for increasing employee satisfaction will probably fail.
“If you’re a company with more than 150 employees, a lot of companies, a lot of people can get value out of it,” he said of AI agents in general and his product, Genie, in particular.
Ghodsi’s own use of Genie (often referred to as dogfooding, from the phrase “eating your own dog food”) is one reason Databricks is pitching Genie as an always-on agent for business leaders. He says he uses it in his personal life as opposed to security-challenged alternatives like OpenClaw, Instinct, and Muse.
He believes Genie can do for business what AI, agents, and harnesses (software that wraps around AI models and makes them use tools) do for programmers. Godici also has an answer to the risk that agents will ratchet up a company’s AI costs. He said costs can be controlled through Databricks’ Unity Gateway product. This allows companies to set usage limits or swap to a cheaper open-weight model.
If you make a purchase through links in our articles, we may earn a small commission. This does not affect editorial independence.
