Skip to content
Advertisement
Tech · AI

Databricks’ new AI agent tells CEO Ali Ghodsi the unvarnished truth

Most bosses hear good news first and bad news last, if they hear it at all. Databricks CEO Ali Ghodsi says he found a workaround. On Wednesday, Nov. 11…

By · 4 min read
Databricks’ new AI agent tells CEO Ali Ghodsi the unvarnished truth
Share

Most bosses hear good news first and bad news last, if they hear it at all. Databricks CEO Ali Ghodsi says he found a workaround. On Wednesday, Nov. 11, his company unveiled a revised version of its Genie agent, and Ghodsi uses it to learn what his 12,000 employees might never tell him face to face.

This matters beyond one company. The agent is not a staged demo. It is the same tool Ghodsi runs on his own work, and Databricks now pitches it as an always-on agent for business leaders.

Why Big Companies Hide Bad News From the Top

Ghodsi used Dunbar’s number to explain the problem. “The magic number is 150,” he told TechCrunch. Named for British anthropologist Robin Dunbar, the idea is that 150 is theoretically the most people with whom a person can keep stable relationships. In a small startup, the CEO knows everyone, and people feel safe giving him negative news straight.

At Databricks, with 12,000 employees, that closeness is gone. A salesperson is unlikely to admit a disastrous sales call to the CEO. An HR manager is unlikely to admit that their plan to lift employee satisfaction will probably fail. Ghodsi says any company with more than 150 employees can get value from agents like Genie.

What Ghodsi Actually Asks His Agent To Do

TechCrunch reporter Julie Bort wrote that Ghodsi skipped the product demo and opened his real agent, then walked her through his recent chats and charts. Genie analyzed the parts of an employee satisfaction survey where the company scored poorly. It then helped him rate, and discard, fixes that some executives had suggested.

It also summarized a big sales pitch that went poorly, including the features the prospect said it needed before buying. Before a podcast interview, it reviewed his earlier appearances and drafted likely questions. Bort adds that it even told him what to say to her, and its advice was not bad.

How the New Genie Works Without Being Asked

According to Databricks, the second generation of Genie is an always-on, autonomous agent for enterprise work. A user gives it a goal. Genie then takes ownership, delegates to subagents when needed, tracks progress, follows up on missing inputs, and reaches out when a person must decide. A new priority inbox shows what needs attention, and Genie emails users when they are away.

Context is the engine. Databricks says every action is grounded in Genie Ontology, which captures how a company operates, from definitions and metrics to the people and processes behind them, and keeps that picture current. Ghodsi says Genie keeps up with his inbox and preps him for meetings. He says Genie has made him “a 10x more productive CEO.” That figure is his own claim, not an independent measurement.

Databricks gives examples beyond the CEO’s desk. Marketers can ask Genie to monitor performance and suggest content. Customer success teams can have it watch usage patterns and alert the account team to meaningful shifts. Executives can ask for a weekly brief on revenue trends and top customer feature requests. HelloFresh’s Alan Khalil says Genie helped reduce the time to identify ad winners on Meta by 75%. That is a customer claim published by Databricks, not an independent test.

Why Cost and Control Still Matter

An agent that works all day raises an obvious worry about the bill. Ghodsi told TechCrunch that Unity Gateway can control costs by letting companies set usage limits and swap in less expensive open-weight models.

Databricks adds that admins decide which tools Genie can use, set approval rules and spending caps, and trace every action. Genie also follows permissions already set in Unity Catalog, so it sees only what each person is allowed to see. Pricing is not per seat, and customers pay for what they use.

Availability is the part to read closely. Databricks says Genie and Genie Ontology are generally available. The new always-on Genie, along with Genie Search and Ontology Manager, is coming soon to preview and beta programs. Buyers should treat the always-on part as an upcoming preview, not a finished product.

What Leaders Should Take From This

The lesson is not that every CEO needs an agent. It is that bad news travels badly up a large company, and a tool that reads the data can surface it without anyone feeling blamed. Leaders should still check what an agent says against what people tell them, because it only knows what its data holds. Start with one real question, such as where employee scores fell, and see whether the answer changes a decision. Before rolling it out widely, ask who can see the answers, who approves its actions, and what it will cost each month in total.

Staff Writer

Adriana Nelson is a staff writer covering technology and how it is changing the way businesses work. She writes about artificial intelligence, automation, and the software tools used by small and mid-sized companies. Her stories explain complex technology in plain language, with a focus on what it means for business owners and their teams. Adriana also follows cybersecurity and data privacy. She is based in Austin.

More from Adriana Nelson →

More on Tech

See all Tech →
Entrepreneurs Edit
Read like a founder. Every story, every morning.

Cover stories, founder interviews and The Daily Edit newsletter — the stories founders talk about.

Subscribe Get the newsletter