Intelligence artificielle : pourquoi l’École devient plus essentielle que jamais

À l’occasion du Global INSEAD Day à Singapour le vendredi 11 septembre 2026, j’ai été invitée à intervenir sur les transformations de l’éducation à l’heure de l’intelligence artificielle, et en particulier à tenter de répondre à une question essentielle : pourquoi l’École compte-t-elle encore davantage à l’âge de l’intelligence artificielle ?Une conviction a guidé mon intervention : plus les technologies progressent et deviennent capables de nous assister, plus nous devons donner à chacun les connaissances, l’esprit critique et la capacité de jugement nécessaires pour rester maître de ses choix.

The full speech ⤵️

Human Agency in the Age of AI: Why School Matters More Than Ever

Good afternoon, and thank you for inviting me to INSEAD. 

Here we are: Lifelong learning, transformative innovation and lasting impact. That is quite a lot to ask of one afternoon. And I certainly won't answer it all. But a few words in the outline caught my attention: Disruption, uncertainty, challenge, shift. And also reinvention, curiosity, creativity, resilience, capability. 

These words describe a world in which we are constantly being asked to adapt, to learn, and to rethink what we know. And that brings me to a question I keep coming back to: What should our children learn, when machines can do more and more, and when AI can tell them more and more of what they once used to learn at school ? What should we still learn, when machines can do more and more of what we once had to learn to do ourselves? 

 It may sound like a question about education. It is. Actually it is also a question about work, leadership and, ultimately, about the kind of society we want to build. And behind many of those words, there is another one. Sometimes explicit, sometimes in the background: AI. Artificial intelligence is not the only transformation we face. But it is already changing how we work, how we learn, how we access knowledge, and increasingly, how we make decisions. So underneath many of today's questions, you may expect me to ask whether AI will replace us. 

My question is something more uncomfortable, particularly for leaders and future leaders : As AI accelerates this shift, are we shaping the transition deliberately, or reacting to changes decided somewhere else? And who will benefit from this transformation, and who will just have to live with it? For a long time, lifelong learning was something positive, but optional, a way to move forward, change jobs, become a better manager. 

That world is now behind us. When technologies, skills and professions change this quickly, lifelong learning becomes a condition of professional survival ; and sometimes of organisational survival for companies. But where does the ability to become a lifelong learner actually come from? The habits that make lifelong learning possible, concentration, perseverance and learning from mistakes, does not suddenly appear in adulthood. It begins much earlier, when children develop the discipline to persist, reconsider and improve. 

This is where school comes in. Of course, early education is no longer sufficient on its own. No school can give a child everything they will need for a working life of fifty years. But those early years remain indispensable. The children in our classrooms today are the employees, engineers, doctors, entrepreneurs and leaders who will enter your organisations tomorrow. Some of these children may one day run these organisations. So when we decide what children should learn today, we are deciding what kind of workforce, what kind of leadership, what kind of society we are preparing. 

Every education system is a country's longest-term strategy, whether we plan it that way - or not. As a Member of the French Parliament and a former Minister of Education, this is one of the questions that matters most to me. And today, one transformation is changing the scale of that responsibility: artificial intelligence. 

AI is changing both sides of the bridge : how children learn today, and the world they will live and work in tomorrow. Let me start with the good news. AI can be an extraordinary tool for education. It can benefit young learners with dyslexia through text-to-speech tools to make reading accessible. It can empower individuals who face dysgraphia or spelling difficulties to overcome the fear and shame of writing. It allows a pupil who has not understood an explanation, to ask for another one, again and again, free from embarrassment or judgment. It can help teachers tailor exercises to different ability levels in the same class, far more quickly. 

In short: AI can remove barriers to learning. So I do not want to prevent children – and teachers - from using AI. I want to prepare them for it. But AI can also make it very tempting to rely on it too much and think less for ourselves. 

So what is important to me, is making sure that AI helps all of us become more capable, not more dependent. And I believe that this starts at school. Because there is another side to this revolution. AI can remove barriers to learning ; but it can also create a new divide, between those who know how to use it, question it and control it, and those who accept its answers without questioning them, and those who have little understanding of what AI actually is. I believe this could become one of the major educational divides of the next generation. 

And this brings me to a conviction at the heart of what I want to say today: The more AI can do for us, the stronger our own knowledge must be, if we want to remain in control. There is a tempting idea I strongly disagree with: if AI knows so much, perhaps children need to know less. I believe exactly the opposite. To question AI, you need knowledge. You cannot check a historical claim if you know no history. 

To correct AI, you need knowledge. You cannot judge a mathematical result if you do not understand maths. To decide what to ask AI in the first place, you need knowledge. It is hard to spot a flawed argument if you have never learnt how to build a sound one. 

And to know when not to use AI, you still need knowledge. That is why, for me, knowledge becomes even more important in the age of AI. It is what allows us to keep thinking, judging and deciding for ourselves. 

And it means that school should not move away from knowledge, to teach only creativity, collaboration or critical thinking. Of course, we need all of these abilities. But critical thinking starts with knowledge: you need a foundation before you can question, compare and judge. (please note the word foundation, I’ll get back to it later). 

In France, these two things, critical thinking and knowledge, even have names of their own. We call one esprit critique, critical thinking, not as a slogan, but as something the curriculum sets out to build on purpose. We call the other culture générale, a broad, connected knowledge of history, literature, science and the world, wide enough to place a new fact somewhere, and to notice when it doesn't fit. Neither translates neatly into English. That's not an accident. It's a sign of how differently a system can choose to define what an educated mind actually needs. It's also why every French pupil, in general and technological tracks, future engineers and future poets alike, sits a compulsory philosophy exam at the end of secondary school, high school or Junior college whatever you call it, just before entering university. Not to become a philosopher. Not to become a new Einstein. To practise, one last time before adulthood, the discipline of questioning an idea properly. 

So only when equipped with critical thinking and general knowledge, can we decide what we are ready to let the machine do for us ; what we are ready to delegate. 

I am perfectly willing to delegate certain tasks to a machine. If it can do something faster, better, more safely, and if doing it myself adds no special value, I see no reason to refuse the help. That is what technology is for. But I still want to understand, question and, when necessary, reject what it gives me, and always remain the one who ultimately decides. And then comes the difference between delegating a task and delegating judgement. And an even bigger difference between delegating judgement and delegating responsibility. 

I am much more careful about delegating judgement. And I do not want to delegate responsibility. This is where education becomes central. Young people will use AI anyway. And they already do. (On a side note: most of the time, in many schools, AI is used under the radar which means used by everyone, yet mastered by very few). 

 What they need help with is knowing when to use it, how to question it, what not to delegate to it, and who remain responsible for the final decision. They need curiosity. They need the ability to reason. They need strong basic knowledge. And they need the courage to say: "I do not understand this answer." Or: "I do not agree with this answer." Or even: "I do not want a machine to make this choice for me." Policy experts call this human agency. I would put it this way : our ability to remain the authors of our own choices. 

Let me test that, quickly, in this room. Please raise your hand if in the last month, you have accepted an AI's answer without checking it? I have done it too. That's exactly the point : this room is full of capable people who are also very busy ; and busy people delegate. With the risk of delegating too much to AI. The question is whether we notice what we're delegating. And the consequences reach much further than education. 

They affect how we make decisions, how our economies function, and how much control countries retain over technologies they increasingly depend on. But it is not enough for a student to control how they use a tool. We also have to ask who controls the tool itself. Who sets its rules, whose values shape it, and whose interests it serves, are questions we cannot ignore. 

For me, that question becomes especially sensitive when the users are children. When a school system brings an AI tool into the classroom, it also brings in choices that were made elsewhere, about data, design, ownership and the legal framework in which the system operates. There does not need to be any bad intention behind those choices. But governments still need to know what they are putting in front of children, and how much freedom they retain if their own priorities differ from those of the provider. 

For many of the countries represented in this room, that freedom is still limited, unless, perhaps, you come from the United States or China. Europe is trying to catch up, and France is doing its part with Mistral. The name comes from a powerful wind in southern France. We are still behind the two leaders, but we are trying to make good use of the wind, even though Mistral is deeply intertwined with American cloud infrastructure and capital. 

So France and the European Union have started building an answer to these questions about control, values, interests or rules. France published a national framework for AI in education in 2025 : clear rules on data, transparency and critical thinking, with AI use growing more independent as pupils grow older, then guided in early secondary school, and more autonomous by the final years. 

At European level, the AI Act backs this approach with law. It is a risk-based model : some educational uses of AI, deciding admission, assessing outcomes that affect a student's path, are treated as high-risk, not banned, but tightly regulated. Others are prohibited. Schools, for instance, are not allowed to use AI to infer from biometric signals whether pupils look attentive, bored or anxious. 

In June this year, the European Commission also updated its guidelines for teachers on the ethical use of AI and data in teaching and learning. Here, the challenge is to combine innovation with what the Commission calls ethical and critical AI literacy. Europe is often accused of regulating too much and innovating too slowly, and, to be fair, we Europeans do have a certain talent for rules and directives. But European regulation has at least one unusual sign of success: from time to time, it manages to irritate Washington and Beijing simultaneously. 

Behind the joke, a serious point : these rules exist to protect our freedom, our choices, how children's data is used, what gets delegated to AI, where human judgement stays in charge. Now, some of you may be thinking: easy for the French to say. Yes, we do love a good rule. Fair enough. 

Yet just a few metres from this room, regulation tells a parallel story : Singapore’s Ministry of Education arrived at a strikingly similar instinct // for very different reasons. It has built the Student Learning Space, a national platform now including AI tools that personalise learning and support teachers. But what interests me most is not the technology, it is the framework around it: four principles, Agency, Inclusivity, Fairness, Safety. 

Agency comes first. Students and teachers are meant to keep choice and control over the decisions that matter, with teachers staying firmly in the loop. Two very different traditions of government with different ways of balancing individual rights and collective priorities. Yet both have reached a similar conclusion : human judgement must remain central. 

I don’t think this convergence of views is a coincidence. China, too, has moved fast, introducing AI progressively from primary school onward, building a national Smart Education platform, and launching a 2026 plan for deep integration by 2030, tied explicitly to the country's wider ambitions on technology and talent. I am not saying any one of these models has all the answers. But the comparison tells us something: education technology is never only about technology. The choices we make about tools, data, access and control are choices about the kind of society we want.

Now let me leave public policy for a moment, because behind all these frameworks sits a question that is far more personal to me, and I do not have a complete answer to it. I use AI. I see what it can do. I am impressed by it. I am optimistic about it. But I also ask myself: how do I remain fully the author of my choices, living with a tool that answers faster than I can, knows more than I ever will, and increasingly behaves in ways that feel human? What draws my attention most is the emotional dimension of our relationship with AI. The fact that it can answer quickly or remember vast amounts of information is impressive. What I find more unsettling is the way it can sound as if it understands me. We have no scientific grounds today to say large language models feel emotion the way we do. 

A recent Nature Human Behaviour study argues explicitly that they do not. But they can already express what sounds like emotion. And two randomised studies published this year found that people can develop a real sense of closeness during emotionally engaging exchanges with AI. And this is where my personal question becomes a question about control. I know what kind of relationship I have with a calculator: I ask a question and I get an answer. 

With AI, things can become more personal. It uses the language of empathy. It can reassure me, encourage me, flatter me, and sometimes make me feel understood. What happens when AI becomes so good at sounding human that we start trusting it, not only for what it says, but for how it makes us feel? And that is enough to change the relationship. Could that make us less critical, only because we feel understood? Could we trust an answer because the voice giving it, feels familiar and warm, rather than because the evidence is strong? I do not know what future systems will be capable of feeling, if anything. 

But we have to understand that emotional interaction with machines already matter. And this changes what « staying in control » means to me. Checking facts is the easy part. The harder part is understanding why I trust a machine in the first place (Why I follow its advice. Why I accept its recommendation.) Competing with the machine is a losing game ; we will never win on memory, speed or calculation. Trying to beat the machine on its own strength is an illusion. 

Real control is knowing what we want AI to do for us and what we refuse to let it decide for us. And neither trapped nor even influenced by emotions. It requires human authority over the purpose. AI may increasingly tell us how. But we must remain capable of deciding why. 

This is why school matters even more in the age of AI. I'm convinced that AI will perform more and more of the tasks teachers currently perform. The biggest gift that teachers can give to our children is to understand what they should refuse to delegate to AI. 

The transmission of basic and broad knowledge? Definitely. The practice of exercises to learn? It has to be individually performed. Getting some feedback? Sometimes, but wisely : never about human behavior. 

 This is exactly what the OECD 2025 PISA assessment report that was released this week says. Let me quote its preface : « artificial intelligence becomes part of the classroom, it should be used in a targeted way for feedback and personalised practice. It should add to students' effortful learning, rather than replace it, since real understanding comes from doing the thinking, not from having a tool do it instead ».

Then what about understanding a child in all his or her complexity? Helping a child stand up again after failure? Teaching to disagree without violence, to listen, to argue, to take responsibility? Transmitting a common culture? Helping a young person discover not only what they can do, but who they want to become? 

Education brings together knowledge, relationships, judgement, culture, responsibility and human interaction, all of which shape the person a child becomes. And this brings us back to the fundamental purpose of education: giving every child enough knowledge to understand the world, enough intellectual independence to question it, and enough freedom of mind to choose what he or she wants to do with it. 

With AI advancing at such extraordinary speed, that mission has never been more urgent. And this is where school meets lifelong learning. Here in Singapore, (there is an image that feels right) anyone who has watched being built one of the towers around us, knows that before you build high, you go deep. You do the piling, driving foundations into the ground so the structure above can rise safely. The taller the building, the stronger those foundations need to be. 

For me, school education is to lifelong learning what piling is to a tower. Nobody lives in the foundations. But everything next depends on it. We will spend the rest of our lives adding floors, redesigning rooms, adapting the structure to new needs. That is lifelong learning. But if the piling is weak, every new floor becomes harder to build. Early education cannot give children everything they will need for the rest of their lives. It should not even try. 

What it must give them are strong foundations: knowledge, the capacity to reason, curiosity, the ability to make an effort and to learn, the confidence to start again, and the independence to judge for themselves. 

Then lifelong learning can build on that. And perhaps we should stop treating school and lifelong learning as two separate worlds. Schools need to understand how work is changing. And employers need to understand what foundations schools must build and protect. Teachers need to hear from business leaders, and business leaders need to hear from teachers. 

And this is where my responsibility meets yours. Mine, as a political leader working on education, is to help make sure schools prepare young people for the world they are entering. Yours, as business leaders, entrepreneurs, investors and decision-makers, is to decide what kind of organisations they will enter. So look around this room for a moment, and ask yourselves: When these young people come to work with you, what will you value? The person who produces the fastest answer with AI ? Or the person who knows when the answer is wrong? Will you reward people who follow the algorithm efficiently, or people who know when to challenge it? Will AI help your younger employees become valuable contributors to your company faster, or will it take away the very experiences through which they learn, grow and develop judgement? And what about your own lifelong learning? The question is not how quickly you learn to use AI. It is what you continue to learn, so that you remain capable of judging what AI gives you. 

That is why I am particularly interested in today's agenda. And let me add a question of my own. What bridge linking childhood education to lifelong learning should be built to help today’s children, who will be tomorrow’s leaders, become more powerful with AI, without becoming less free? Let me quote Nelson Mandela. It is a well-known quote, perhaps because it says something fundamental : "Education is the most powerful weapon which you can use to change the world." I believe those words are just as true in the age of AI. 

So I’ll leave you with this very last question : how will each of you use lifelong learning to change the world? I am genuinely looking forward to hearing your answers today. And I wish you a very productive and thoughtful afternoon. Thank you very much.