Capgemini's CEO Has a Blunt Message for Bosses Rushing Into AI: Slow Down Before You Waste a Fortune

Aiman Ezzat does not think you should panic about AI. He also does not think you should ignore it. What he thinks, specifically, is that the current scramble to adopt the technology as fast as humanly possible is going to cost some companies a great deal of money and deliver very little in return.

Ezzat runs Capgemini, the French IT consultancy that employs roughly 340,000 people across 50 countries and pulled in €22.5bn in revenue last year. He is not some academic theorist or a VC partner talking his own book. He is a man whose company gets paid to implement this stuff, and he is telling clients to slow down. That should probably give people pause.

'You don't want to be too ahead of the curve,' Ezzat said — a line that sounds almost quaint in a boardroom culture where being behind the curve is treated as a sacking offence.

The Metaverse Lesson Nobody Learned

Ezzat has watched this film before. He brought it up himself: the metaverse.

Two years ago, the metaverse was going to change everything. Companies built labs, hired specialists, committed budgets. Capgemini opened its own metaverse lab, which is the kind of detail that tends to get glossed over when a CEO warns about hype cycles — he is not exempt from them either, and to his credit he did not pretend otherwise. Adoption stalled. Consumer interest never materialised at the scale the forecasts promised. The money, for most firms, was wasted or written down. A few niche applications survived; the grand vision quietly evaporated.

The parallel to AI is not exact, and Ezzat was careful to say so. AI has demonstrable, immediate uses that the metaverse never had. But the pattern of corporate behaviour — the rush to invest before understanding what you are investing in, the fear that competitors will get there first, the pressure from investors to show you are doing something — is identical. Strip away the technology and what you have is FOMO with a procurement budget.

His argument is that AI adoption will be incremental. Not one enormous leap, but a series of smaller ones, each building on the last, each requiring a slightly different kind of readiness. The companies that do well will be the ones that pace themselves; the ones that overcommit early will find themselves locked into capabilities nobody needs yet, burning cash while the market catches up.

Or to put it more bluntly: if you spend £50m on an AI infrastructure that your customers will not be ready to use for three years, you have not invested. You have gambled.

What Capgemini Is Actually Doing

The company runs what it calls exploratory labs — small-scale research operations in quantum computing, robotics, 6G mobile and, yes, AI. The point is to learn without betting the firm. Ezzat described the approach as continuous learning cycles rather than big flagship projects, which is consultant-speak but also, in fairness, a sensible way to stay current on technologies whose timelines are genuinely uncertain.

Quantum computing is the clearest example. Nobody knows when quantum machines will be commercially viable at scale. Could be five years, could be fifteen. Capgemini is not building a quantum division; it is running a lab, keeping people trained, watching the science. If the technology matures, they are ready. If it does not, they have not lost much.

Apply the same logic to AI and you get Ezzat's core message: experiment, learn, do not overcommit. The firms making headlines with massive AI investments might look visionary now, but so did the firms making headlines with massive metaverse investments in 2022. Some of those firms no longer talk about the metaverse at all.

The Bigger Point, and the One Most Companies Are Missing

Here is where Ezzat said something genuinely interesting, or at least more interesting than 'don't rush.' Most companies, he argued, are using AI to make existing processes faster. Automating invoices. Speeding up customer service. Trimming costs in finance and operations. All useful. None of it is transformation.

Transformation, in Ezzat's framing, means letting AI change what a business fundamentally does — not how quickly it does the things it already did. The question he posed was not 'how do we use AI to be more efficient?' but 'how will AI disrupt our business model before a competitor does it for us?' Which is a rather different question, and one that most leadership teams are not yet asking because the efficiency gains are easier to measure and easier to sell to a board.

He is right about that, probably. The companies that treated the internet as a faster fax machine in the late 1990s are not, by and large, the companies that dominate now. The ones that rethought their entire model around what the internet made possible — Amazon, Google, eventually every bank — are. The analogy is imperfect; all analogies are. But the principle holds: if you are only using a new technology to do old things faster, someone else will use it to do new things entirely.

Ezzat also pushed back on the idea that sticking a 'human in the loop' is sufficient for responsible AI. The phrase has become a kind of corporate talisman — say it in a board meeting and everyone feels better. His argument was that human oversight needs to be embedded throughout the design process, not bolted on at the end as a compliance checkbox. Whether most companies will actually do that, given the cost and complexity involved, is debatable. But he said it, and from someone running a company of Capgemini's scale, it carries weight.

No product launch date, no earnings guidance, no dramatic pivot. Just a consultancy CEO telling other CEOs to calm down. In the current climate, that might be the most contrarian position available.