For years, most board meetings have followed the same ritual: Two weeks before the meeting, directors receive a thick digital packet—often 100 pages or more—then do their best to find the “needle in the haystack” before showing up in the boardroom. Florin Rotar, group CTO at Atos and Avanade’s former chief AI officer thinks that era is ending fast.

“In terms of getting ready before the board meeting and asking the right questions during the board meeting, and then making sure that there is proper follow-up, I think going away from this old-school approach of a big deck, or a mountain of material, is key,” says Rotar. “It’s really helping with digestion of what’s important.”

Rotar, a veteran technologist and advisor to boards, sees AI not as a replacement for directors, but as an amplifier—one that will soon be table stakes. “I don’t think anything should be automated—I think it should be augmented,” he says. “Using AI to empower board directors with a richer perspective, to be able to see the needle in the haystack, to help them with a diversity of thought—that is quite powerful, and I think that will become standard.”

From board books to “virtual board members”

The most advanced boards Rotar works with are already using AI to summarize sprawling pre-reads, surface anomalies in KPIs and financials, and flag trends that might otherwise be buried. “Being able to digest information and synthesize information quicker, get to the bottom of what really matters through data—that’s the important point here,” he says.

Some are pushing further, using AI as a simulation partner. “This is more experimental for now, but I’ve seen it used as a sparring partner for simulation scenarios,” says Rotar. “A little bit like tabletop exercises for how cyber used to be done. Using AI to simulate and prepare for scenarios like activist investors or unsolicited acquisition attempts, and use AI to provide a contrarian perspective.”

In his view, it’s not far-fetched to imagine AI playing an even more visible role in deliberations. “We are going to start to see AI being used for augmentation of board directors to provide an additional perspective, to help with a bit of role playing and what-if analysis,” he says. “Who knows? Maybe we’ll have AI agents as virtual board members at some point in time.”

He notes that some sovereign wealth funds in the Middle East have already publicly experimented with “virtual board members” to support investment decisions. “I’m pretty sure it’s a little bit of marketing and hype,” he says with a laugh. “But I think it’s not entirely out of the art of the possible.”

Sovereign AI as the board’s “corporate cortex”

Behind the scenes, the technology itself is also changing. Rotar says serious work is moving away from generic, public models and toward tightly controlled, highly customized systems.

“For serious stuff, you would normally use a customized version, because otherwise the risks are too big,” he says. “You’d have hallucinations, and you should worry a little bit about the security and privacy bit of it as well.”

At the frontier, some companies are going further still. “For the really most sophisticated scenarios I’ve seen, it is more than customized—it is custom, purpose-built, sovereign models,” Rotar explains. “Things which companies build almost from scratch nowadays and it’s part of their secret sauce, the most important asset in the company. It’s the know‑how, the ‘corporate cortex,’ if I can use that word.”

Just a few years ago, he notes, this would have been out of reach for all but the largest firms. “Things which used to cost $1 billion five years ago, $100 million two or three years ago, are suddenly in the $10 million‑or‑so range,” he says. “It’s rapidly becoming more affordable and realistic. I wouldn’t say it’s trivial for an average mid‑market company, but it’s feasible.”

A new kind of board work

If boards want real value from AI, Rotar says, they’ll have to change how they personally work with it. That starts with using systems that are actually capable of deep reasoning—and then pushing them hard.

“Nothing beats hands‑on experience,” he argues. “You can go to all the school programs in the world. Nothing really beats experience of using AI on a day‑to‑day basis.” The catch: “Unfortunately most people are using free or dumbed‑down versions of AI. My recommendation would be: Just invest to buy access to the best AI money can buy.”

He also urges directors to treat AI less like a magical oracle and more like a very unconventional colleague. “The way you get really good results with AI, especially reasoning AI nowadays, requires one to almost engage with it like you would engage with a human, but some aspects of it are very atypical of what a decent human would be doing,” he says. “The more assertive you are with AI—and to some extent bullying it and second‑guessing it and questioning it—the better it gets.”

Rotar even uses multiple models on the same question and lets them critique each other. “The best results I get is when I ask them for almost the same thing and then I pit them against each other and ask them to critique each other’s work and synthesize,” he says. “You would never do that with people—it would show a lack of trust. But it’s actually the way AI works.”