Research on artificial intelligence and organizations has followed a revealing trajectory in just three years.
In 2023, researchers from Harvard, Wharton, and MIT tested 758 consultants from Boston Consulting Group, with and without access to GPT-4. The results were striking: an average 40% improvement in work quality. At the time, the key question was whether AI actually worked. This study made the answer clearly yes. But there was also a subtle signal in the data that deserved attention: lower-performing consultants improved by around 43%, while top performers improved by only 17%. The use of generative artificial intelligence leveled the playing field. It democratized technical competence. And it quietly shifted the central question. After all, if this new tool reduces individual differences, what truly distinguishes people and teams?
Two years later, the McKinsey Global Institute attempted to answer that question. By 2025, almost every organization had adopted some form of AI, yet only around 1% considered itself to have achieved real maturity. The biggest obstacle identified was not employee resistance or the cost of the tools. It was leadership that “is not moving fast enough.” The question had evolved from doubts about the usefulness or effectiveness of the technology to: why is it not working at scale?
More recently, a new Deloitte publication has added further momentum to the discussion around AI adoption in business contexts, showing that the organizations driving genuine transformation and impact are those redesigning processes and redefining how decisions are made. In other words, operations and leadership, traditional management competencies. Three years after the initial excitement, the conclusion now seems unavoidable: as we have seen with previous technological transformations, the central issue is not the technology itself, but the quality of management.
Faced with data supporting what experience had already suggested, when a student asked me, in a tone somewhere between concerned and genuinely curious, “I read today that AI is going to replace CEOs. What do you think about that?” I had little hesitation in my response. I simply said: “Certainly not. But it will require different skills.”
This is not the first time we have seen this pattern. With CRM systems, ERP systems, and every digital transformation since the 1980s, the sequence repeats itself: technology arrives, organizations adapt the narrative, and eventually it becomes clear that the real challenge was never implementation. It was always leadership.
The scene is familiar. A forty-five-minute meeting, ten people in the room, three decisions to be made, and none actually made. AI can produce the analysis, organize the information, and anticipate scenarios. What it cannot do is replace the manager who reads the room, understands what is not being said, and makes a decision. That remains, and will continue to remain, irreplaceable.
This is where AI acts as a mirror that tells the truth. The teams that benefit most from these tools are not necessarily the most technologically advanced. They are the ones with clarity of purpose, the confidence to experiment, and the ability to learn. These conditions are created by managers, or constrained by them. In a cohesive team, AI amplifies what was already working. In a disorganized team, it amplifies what is already there too: confusion, lack of direction, and ultimately wasted talent. Technology will never fix weak leadership; it will simply make it more visible.
Some will argue the opposite: that AI will actually shield poor management, generating polished reports, flawless presentations, and the appearance of productivity without substance. It is a serious argument. But it ignores a deeper pressure: speed. When decision-making cycles shorten, when competitors move faster, and when teams have access to the same information as their leaders, managers who hide behind appearances are exposed by outcomes. AI accelerates everything, including evidence of who is not capable of keeping up.
The good news is that this exposure may force organizations into a conversation they have long postponed. The question will no longer be “how many jobs will AI eliminate?” That question is already becoming secondary. The more uncomfortable question is this: when we remove from management the tasks machines now perform better, what remains? As Erik Brynjolfsson, the Stanford researcher who has extensively studied AI and productivity, said: “Managers who know how to work with AI will replace those who do not.” A deeper interpretation points in a complementary direction: what remains is the ability to create conditions in which others can do great work, the courage to make decisions with incomplete information, and the consistency that builds trust over time. Managers who know how to lead people will survive any technological cycle.
In summary, in my view, AI will not replace managers. But it will make it much harder for managers who never truly knew how to lead to hide. And in the current context, that is good news.
Silvia Almeida, Professora da CATÓLICA-LISBON