The Autonomous Boardroom: The Radical CEO Replacing Their Executive Team with AI Agents

When I first learned about it, I thought it was a joke. According to reports, a Singaporean founder replaced the majority of his executive team with a stack of AI agents operating on a customized orchestration layer. Not a CFO. Not a COO. Sitting in a glass-walled office above a noodle shop is a skeleton crew of humans, primarily engineers and lawyers. It sounded like the kind of proposal you make to attract attention at a conference. The industry stopped laughing when the financials were released.

Speaking with executives who have witnessed this unfold, there’s a feeling that something has changed over the past eighteen months that nobody can quite pinpoint. In short, according to IBM’s most recent survey: Compared to 26% a year ago, 76% of organizations now have a Chief AI Officer. It’s not a trend. It’s a stampede. What happens when the CAIO becomes the layer above the CFO instead of a peer is the more intriguing question that no one at Davos seems to want to ask aloud.

The CEO at the center of this experiment, whom I will refer to as “the operator” because that is how he prefers to be described, is in charge of a mid-sized logistics platform that operates on three continents. He continues to maintain a small finance team for regulatory filings, a human general counsel, and a head of safety. He maintains that agents handle everything else. decisions about prices. negotiations with vendors. shortlists for internal hiring. escalations from customers over a specific threshold. He recently told a reporter that his work now resembles that of a referee rather than a chief executive and that his agents make about 4,000 decisions every day. It’s more difficult to determine whether marketing or that is true.

The unglamorous appearance of the actual setup is striking. I’ve seen pictures. There are no Bond villain dashboards or humming server racks. Just a row of green, occasionally amber monitors that occasionally flag a decision for human review. The design is more akin to an air traffic control tower than a boardroom. Workers pass by carrying coffee. A partially completed crossword puzzle has been left on a desk. It’s difficult to ignore how ordinary it is.

Many skeptics cite Gartner’s prediction that insufficient safeguards will cause 40% of agentic AI projects to fail by 2027. They bring up the CEO of Wayfound’s recent finding that AI agents drift rather than crash. That’s the part that stays. A drifting agent doesn’t make a big noise when it fails. Three quarters later, it quietly begins approving incorrect invoices or favoring one supplier for reasons that are impossible to fully reconstruct. Todd McKinnon of Okta has been direct about this, stating that AI agent governance is a fundamental security concern rather than a future one. He contends that the majority of businesses are unable to even compile a list of the agents operating within their own premises.

Nevertheless, investors appear to have faith in the wager. The operator’s business raised a new round at a valuation that appears aggressive by any conventional measure. When it comes to hype cycles, Wall Street has a long memory, but when it comes to margins, it has a shorter one. Anyone observing this space is aware of the similarities to the doubts Tesla faced years ago. The promise of quicker decisions, lower salaries, and no boardroom politics is precisely the kind of narrative that attracts buyers.

It is also uncomfortable from a philosophical standpoint. Human executives claim that AI is incapable of handling stakeholders, making moral decisions, and understanding the unspoken guidelines for when to object on a board. They may be correct. Alternatively, they may be saying what those whose jobs are in jeopardy have always said. Which is still unknown.

As we watch this develop, the truth is that no one can say with certainty whether autonomous boardrooms are the way of the future for corporate governance or if they are just a warning sign awaiting their first significant lawsuit. Most likely both, in varying amounts based on the business. It is evident that the experiment has left the laboratory. Right now, someone is running it with real money and real repercussions.