One chief executive, shown a new AI "board bot," had a blunt reaction: "I'd immediately replace a couple of my board members with AI." He then asked whether the tool could be priced at half what he pays his directors. "I want to make them nervous," he said, according to Brian Stafford, chief executive of Diligent, the software maker whose platform directors use to share documents and run meetings.

The remark was half a joke. The question behind it isn't. How far can, and should, AI go in the boardroom?

How are boards actually using AI?

Mostly as a supercharged assistant. AI today summarizes lengthy board papers, tidies up minutes, tracks market trends and answers directors' questions. Joanne McEnteggart, global head of debt, capital markets and corporate at IQ-EQ — a fund-services firm administering around 30,000 board meetings a year across 25 countries — says the technology is changing how directors prepare. One director she has known for years now asks questions they never would have before. "That's good," she says, "because it challenges thought."

Some boards push further. Executives are creating digital replicas of colleagues to rehearse difficult boardroom discussions, Stafford says. One chief executive uses AI to play the role of an activist investor, providing a dissenting voice. At one acquisitive company, a director asks the AI to argue against every deal. With no human having to play devil's advocate, Stafford says, boards avoid "awkward" arguments. Some boards have even used AI to break stalemates — a role usually played by the chair, McEnteggart adds.

In April, Lloyds Banking Group became the first FTSE 100 firm to acknowledge deploying an AI agent for boardroom work, The Times reported. Its executives and directors use the bot for agenda planning, pre-meeting analysis and to help mitigate human bias.

What are the claimed benefits?

McEnteggart lists the strengths plainly. "AI can assimilate information quickly and dissect years of data," she says. "It removes emotion, stress, and bias, and can call out a director who seems to be deciding against the facts."

Stafford also sees AI cutting boards' reliance on external advisors, with AI specialists filling knowledge gaps. "Clients are asking for a Middle East expert, a sustainability expert, a futurist," he says. Pippa Begg, chief executive of board-software firm Board Intelligence, goes further: AI could eventually sit in on meetings and be "able to almost interrupt and say: 'Hang on, I think you're falling into this trap.' Or: 'I disagree.'" Giving it a formal legal vote, though, would be "a dangerous leap."

Where has AI actually been given a seat?

Two companies have crossed the line. Kazakhstan's sovereign wealth fund, Samruk-Kazyna, appointed an AI board member called SKAI last year and gave it voting rights. Its chief executive says SKAI "voted on all the items" at its first meeting — but the country's financial regulator says the appointment breaches national law. SKAI remains on the board as a pilot, according to a recent filing.

Abu Dhabi's International Holding Company, the UAE's largest listed firm, gave a non-voting board seat to an AI called Aiden Insight, which has issued statements describing itself as "a C-level AI entity."

Mark Freebairn, head of the board practice at search firm Odgers, knows of other boards that have given a bot a seat and a vote, though it remains "very uncommon." Doing so, he says, means "accepting a robot adds as much value as a human."

What's holding it back?

Law, for one. "A bot cannot be held personally responsible in law nor can anyone vouch that it is fit and proper to sit on a board," McEnteggart says. Directors carry personal legal duties; software can't.

Then there's trust. In a test by Wharton and INSEAD researchers, a board of AI chatbots outperformed six human boards on decision quality and use of evidence — but it struggled to build trust. The researchers concluded AI is better used to stress-test a board's thinking than to replace it. Freebairn worries about a subtler risk: groupthink. If directors lean on a tool few understand, AI becomes a new source of consensus. "I sometimes feel like I'm the only person that's seen Terminator 2," he says. People assume AI is impartial, with "zero emotion, zero human risk of bias," he adds. "How can you test what an algorithm tells you is right? That feels like one hell of a risk."

Deanna Strable, chief executive of Principal Financial Group, who also sits on the board of Elevance Health, doubts AI directors will go mainstream — but won't rule it out. "My gut answer would be no. But so much has happened in the past year." If it happens, she expects an advisory vote, not a decisive one.

Is there a governance gap?

A big one. Half of U.S. public companies lack a board-specific AI policy, and only 14% give directors company-approved AI tools for board work, according to a survey by Deloitte and the Society for Corporate Governance. That matters because board papers are among a company's most sensitive documents — unannounced deals, unpublished results, succession plans. Many directors are retired executives without enterprise AI access, Stafford notes, so they fall back on consumer chatbots with weaker data protection. Upload papers to ChatGPT, he warns, and it will have them "forever." Boards need clarity on "what company data is going where," McEnteggart says.

The knowledge deficit is stark. Two-thirds of directors and executives say their boards have limited or no knowledge of AI, per a 2025 Deloitte survey of nearly 700 respondents in 56 countries. "Most board members do not understand the complexity of rolling this out," Strable says. Seven in ten large European companies have defined board oversight of AI, according to proxy advisor Glass Lewis — but that covers corporate use of AI, not directors' own habits.

Nearly half of directors (49%) say their board is actively reviewing which decisions should stay human as AI improves, according to a Board Intelligence survey of more than 400 directors, CEOs and CFOs. Freebairn frames the stakes: "If you take humans out of judgment, there's nothing else we offer really."

Even Stafford, the man selling the board bot, urges restraint. "AI slop is a big issue, so you don't want that creeping its way into the boardroom," he says. "More is not always better. It is just more."