Eric Schmidt explains how AI helpers could change top executive jobs

Eric Schmidt, who used to run Google, says the CEO of the future will have an AI assistant that can see across the whole company. He shared the idea on the "In Good Company" podcast, as reported by Business Insider. His warning is that people whose job is keeping a large business coordinated could be the ones most at risk.

What the assistant would do

Schmidt described AI agents, which are software that can carry out tasks on its own. He said they could check whether company activity matches a leader's values, follow up on projects, and help find ways to raise profits. He argued this could make companies faster and flatter, with fewer people needed to plan, pass along information and move numbers around.

Early examples

Schmidt said some of this is already happening in small ways. The examples he gave:

• Two founders he knows ask an AI assistant every Friday to flag anything at their company that conflicts with their values.
• One executive built an AI version of himself that checks an employee's progress, gives instructions and answers requests for help.

The second example worried Schmidt. He said the exchange sounded potentially abusive, though he saw that it sped up the executive's decisions.

The coordination tax

Schmidt said his own days as CEO were eaten up by meetings, 12 hours a day, five days a week, listening to people explain projects and decisions. He calls the work of gathering updates, passing them upward and keeping teams aligned a "coordination tax." He thinks AI can remove much of it. For a company with 10,000 employees, he guessed it could leave "a couple of thousand people who you don't need."

He pointed to a friend whose fast-growing company cut 20% of its staff, including the "glue people" who handled planning and coordination. The friend later told Schmidt he had received death threats over the cuts.

Why he says time is short

Schmidt said AI competition is fiercer than the PC-versus-Mac or Android-versus-iPhone fights. A model can take about six months to build and two more to test, then be overtaken within five months of release. "Time compresses every cycle," he said. He also rejected the idea that AI is a bubble, citing Anthropic's fast revenue growth as "real money with real customers."

What stays human

Schmidt said companies will still need people who build products, make sales, speak with regulators and lead teams. For workers whose roles disappear, he said executives should move them out "legally and humanely." These are his views and guesses, not a forecast backed by data in the interview.