The CEO who announces an AI transformation, appoints an AI owner, and permits pilots to run across the company looks decisive. Yet this is often the most conservative move at the boardroom table.
The ambition is not the problem. The problem is where that ambition is aimed. Many executives focus their attention on the use of AI, while AI is beginning to change the design of the enterprise itself. The question is therefore not whether a company deploys AI, but whether that deployment creates new value or mainly makes an existing organisation slightly more efficient. The CEO's real domain is not the AI programme. It is the way the company is organised and, deeper still, who or what staffs that company from now on. Both are, for the first time in decades, fundamentally up for debate.
Announcing an AI ambition feels like leadership. In practice, it is often merely sponsorship.
Almost every knowledge institution advises the same first step: formulate an AI ambition, define success criteria, and mobilise the organisation around it. That is a sensible start. It provides direction, frees up budget, and brings AI into the boardroom.
The problem arises when that first step is mistaken for the transformation itself. Execution then remains with individual functions, a portfolio of pilots and experiments takes shape, and the implicit expectation is that enough local successes will automatically add up to broad enterprise value. One variable is left untouched in the process: the organisation as it exists today. AI is added to the company, not used to reshape it.
This approach falls short for a simple reason: it leaves the organisation untouched, precisely while that organisation is beginning to shift. To see this, it helps to look at the origin of the role. Research by Bandiera et al. (2020) into the behaviour of more than a thousand CEOs shows that their work consists largely of bringing together information and insights from different functions in order to make directional decisions. Executives who connect across departments outperform executives who mainly operate within silos. The CEO thus sits atop a structure of management layers that originally emerged because coordination between people was costly: gathering, aligning, and passing on information took time, money, and capacity.
AI drives those costs down. As summarising, unlocking knowledge, and aligning work become cheaper, the case for certain management layers begins to erode, and with it the rationale for certain company boundaries. Recent economic research by McKendrick (2026) shows that organisations adopting AI are already beginning to flatten their hierarchies.
What becomes rare is not the sharpest insight, but the willingness to be responsible for a decision that someone else, whether human or machine, has made.
That shift plays out not only around the CEO, but also within the person holding the role. The idea that their work evaporates the moment AI takes over the synthesis of information is of course wrong, because that synthesis has two sides. AI makes the first—processing and summarising—cheap and abundant. The second, judging under uncertainty, remains scarcer. And even that judgement is less safe than it seems. In measurable domains such as the financial sector, statistical models often outperform human experts, and yet people are quicker to distrust those models the moment they slip up, even when the models demonstrably make better decisions.
What then remains as the CEO's primary role is not the best insight, but owning the outcome. Here, it is essential to recognise that the CEO of the future manages two workforces. The first is human and sits on the payroll. The second is digital: it appears on no organisation chart, never applied for a job, can multiply tenfold overnight, and increasingly makes its own calls on a customer, a price, or a risk. In Europe, a human will always sit above that digital workforce; the law requires it, and ultimate responsibility cannot be delegated. So a moment will inevitably come when an executive is held to account for a choice their digital workforce made.
If that is the true core, then the question shifts too, and this shift is reserved for the CEO alone: not which AI tools enter the organisation, but what the enterprise would look like if it were redesigned today, with machines doing a growing share of the work. Redesign alone, however, is not enough, and this is where the CEO role sets itself apart from the rest of the C-suite. The COO can reorganise the process and the CHRO the people; only the CEO owns the question of what the enterprise is for. That is not a cost question but a value question: which propositions and markets come within reach when work that was once unaffordable suddenly becomes scalable? Those who see AI merely as a cost saving miss exactly that opportunity.
This calls for a different investment logic: not a broad spread of scattered pilots, but a small number of deep, focused commitments, each tied to a critical process, to an explicit decision about how work is divided between human and machine, and to who bears the outcome. Because the outcome is uncertain, this is not a blueprint but a series of considered choices that steer the enterprise as it learns. That is what sets apart the small group that does extract value from AI.
The following strategic questions offer a starting point for the boardroom conversation:
