So I have a warm-up question about how AI is changing the entry point into professional employment. But researching why we should speak to you, there are just so many things. So what I'd love for you to do is just begin with giving a quick introduction to who you are and what you have going on right now.

I am a professor at HEC Paris. Before that, I was a management consultant for a quarter century. And my focus in my research and my teaching is basically decision-making. What I study and what I teach is how to make better decisions, strategic or otherwise.

One question that obviously has emerged as a very important one recently is, how does AI change those decisions? How should it change how we decide, and what should we not let it decide? What does it mean to let AI decide? That's what I talk about in my latest book, which I co-authored with my friend Eric Hazan.

You know that this series is on how AI is changing the entry point into professional employment and what universities need to understand. Give us a quick picture of what thoughts come to mind when you are thinking about this entire kind of change, this entire revolution, just free-form.

First, the claims we hear about a revolution in the short term are overstated, in my view. I hear a lot of people saying companies stop hiring because of AI, they're replacing every junior position with AI. You see posts on LinkedIn every day saying, you don't need McKinsey anymore because here's a Claude prompt that is going to replace a team of three consultants worth 100,000 pounds or dollars for five euros. That is nonsense, of course.

I'm not disputing that some companies might be hiring less, but it's absolutely not clear to me that AI is the sole, or even the main reason. At best AI is an excuse for companies that are reducing their hiring for entirely unrelated reasons. At worst, AI is one of the reasons for companies to reduce their junior hiring, alongside another culprit, which is underappreciated and underdiscussed, which is the working-from-home trend.

There is some new research, not yet compelling, but pretty suggestive, showing that the companies that are hiring less are indeed using AI, but are also companies where working from home has been standard for a long time. Using some statistical weaponry, the authors separate the effects of one and the other, and it looks like working from home is actually a stronger contributor to the decrease in employment than AI is.

Another reason why I'm not convinced there will be a major shift in the short term (and again, I stress "in the short term") is because in most large organizations, managers tell us that they are very enthusiastic, that they do a lot of stuff, and that they get absolutely no tangible bottom-line benefits.

Now, that's not the story that Anthropic and OpenAI are telling you in their IPO prospectuses, but that's the story that keeps coming out over and over again, whenever managers are asked about this. Bear in mind, by the way, that managers who are asked about this have had every incentive to claim the opposite. So if they are telling you in survey after survey that they are not seeing the bottom line results yet, it's probably because they really aren't seeing them just yet.

And the reason for that, I would venture, is because they haven't rethought the work processes in a way that could let them capture the benefits from automating certain activities. So, to oversimplify a bit, but not too much, what is going on is, yes, you're using AI and because you're using AI, Alfred and Narek can go home at 4 p.m. instead of 5 p.m. But you are still paying the salaries of these two employees. For the company to be able to have every third or fourth or fifth person taken out of the workforce as a result of productivity gains, you would need to reorganize in a pretty significant way.

Take a simple example: I hear a lot of people say AI is going to revolutionize strategy consulting. I'm sure it will. But does that mean jobs are disappearing? If a consulting team has three people, can you achieve such productivity gains that you're going to reduce it to two people? That seems unlikely. So what you're doing instead is achieve more with those three people, work faster on certain activities. Whatever activities they perform faster leave time for them to do things that they otherwise wouldn't have had time to do. Obvious example: spend time with their clients to make sure that they understand what you're recommending and then actually do it, as opposed to producing slides that explain it in beautiful diagrams.

This is why we wanted to do this project. There is someone else saying the complete opposite of what you're saying. And all of it is extremely reasonable, extremely science-based. But I think that speaks to the time that we're in. We are in a completely messy, ambiguous time.

I'm not making a forecast. I'm talking about the interpretation of what we're seeing right now. Now, we can have different interpretations of the facts.

The facts are that there is less hiring in some places, a clear trend downward in hiring of graduates. The time for people to get a job out of university seems to be longer. Some significant companies have reduced their hiring, and others have let people go. So something is going on.

My point is twofold. First, what is going on is not nearly as massive as what you hear in the big proclamations that OpenAI and Anthropic and Microsoft and so on are making. When I hear things like "every white-collar job on the planet is going to disappear in the next few years," I simply say: we're not seeing that. The magnitude of what we're seeing is very, very far from that.

And the second thing that I'm saying is that there are alternative explanations for the decline that we're seeing in the short term. One is basically the economy, and the other one is working from home. So the narrative that AI is going to destroy employment at the junior levels at a massive scale is one that I'm simply not buying. I hear that there are people who disagree. I'm happy that I am not just stating the obvious!

That gives me a lot to think about. Let's play back to me what the people who have the opposite point of view are telling you, so I can see if that should budge my mind, or if I have evidence that backs up what I'm claiming. I could paraphrase. I don't want to put words into someone else's mouth, but just for the sake of the discussion, many people are saying, and researchers have been writing books about this since the 90s, that organizations are usually a pyramid, and what you're seeing right now is that the entire bottom part of that pyramid is getting slashed. And you have PwC putting out a report in 2026 talking about how that pyramid is disappearing, how it's turning more into a diamond shape. Now, that doesn't really prove that it's actually AI doing that, just that that is actually happening. So to your fair point, I see. But for me personally, I would say that it is too big of an event to just see it as circumstantial.

The main argument to say that the jobs at the bottom of the pyramid are going to disappear is essentially that AI is enabling massive productivity gains in the activities that are performed by some people, and therefore you're going to have fewer of them. That's the claim.

Imagine you apply this logic in the 1990s. If you had looked at me as a young junior consultant in the early 1990s with a stopwatch in hand, you would have seen that my job as a junior consultant at the bottom of the pyramid at McKinsey consisted in tasks that would all soon be automated.

I want to make this concrete, so let me tell you what my first project was about. My job was calling electrical wholesalers around the country, on the phone, a landline of course. I would get the phone slammed in my face ten times before I could actually get someone to speak to me. Then I would try to get permission to come and visit them to interview them, schedule a meeting, get on a train to go to the other end of the country, and finally spend 45 minutes with the manager of a mid-sized electrical wholesaler. All of this to find out a few things, but mostly one piece of information, which was how many employees he had. And at the end of that three-month project, we came up with a brilliant segmentation of electrical wholesalers. It had three segments: the large ones, the medium ones, and the small ones.

Now, in the process of arriving at that earth-shattering insight, I would do lots of calculations by hand with an HP 41 calculator, produce a lot of drawings, on paper, with a pencil and a special piece of plastic that McKinsey gave me. My sketches would then be given to a team of visual aid specialists, artists, really, who would turn this into beautiful slides. It would take a week to produce a presentation.

Once the slides were produced, you needed to make copies of them. And since copy machines at the time were not very good at assembling, you needed to make copies of page one, copies of page two, and so on. Then you would assemble them by hand. But of course you needed page numbers, too. So I had a set of stickers with numbers on them, and my job was to stick those page numbers exactly one centimeter from the bottom-right-hand corner of each page.

So you look at this in 1990 and you say, I know that Word, Excel and PowerPoint are coming. 95% of what this guy does is going to be eliminated. Add to this the internet, and it's 99% of what I literally did all day that was eliminated.

Did McKinsey divide its number of junior consultants by 20 or by 100? No, it multiplied it by 20. What is going on here? What is this mystery? This mystery is a very simple economic mechanism. It's called productivity. When you become more productive at doing something, you do it cheaper, and demand grows.

And so McKinsey grew its business because with one consultant like me, it could deliver more value to clients. Because guess what? Sticking page numbers on slides is not the most productive use of a young graduate's time. Therefore, McKinsey was able to grow its business massively because what a junior consultant would do 10 years later was massively more productive. And the work that I did in three months was later done in three weeks, in three days, or in fact in three minutes, meaning that you would give it for free to a client as part of a proposal as opposed to making a project out of it.

The point I'm getting at with this long story is not just that I'm very old. It is a very basic and yet constantly overlooked point: just because you eliminate an activity does not mean that you eliminate a job. It can actually mean that you create jobs.

Note that this is different from the Schumpeterian argument that creative destruction creates jobs in other parts of the economy. That argument goes: yes, jobs of consultants are going to be destroyed, but they're all going to become video game designers. That is another thing you could have said in 1990, which is also correct, by the way. A lot of people who would have been consultants or could have been consultants became hedge fund traders and video game designers and AI developers like you. And these are jobs that we couldn't have imagined would exist at the time.

But the point I am making here is an even more basic one: there are also going to be more jobs in the existing organizations. Because the people whose activities are being automated will find other higher-value activities to perform.

A classic example of that in another field is the prediction that Geoffrey Hinton was making a few years ago. Geoffrey Hinton doesn't make that many mistakes. But he was saying, in essence, "we should stop training radiologists. They are going to be completely useless five years from now because AI will be better than them at reading X-rays."

He was right. AI is in fact way better than them at reading X-rays. But he was also wrong. We haven't stopped training radiologists, and we're very happy that we haven't, because we need more radiologists, not fewer. Why? Because radiologists, when a machine is reading the X-ray, still have a lot of very important, interesting, high-value stuff to do. Like talking to their patients. And before they talk to the patients, talking to their colleagues to make sure that they interpret the X-ray in the context of an overall patient perspective.

So the basic point here is, don't imagine that the automation of activities mechanically results in job elimination. That assumption is explicitly made by most studies you read on the effects of AI on white-collar jobs. They all do it in different ways, but every time, they measure the activities, they estimate how exposed the activity is to AI, and they conclude that a certain fraction of that activity is going to disappear. I think I've just told you why I disagree with that logic.

I do accept that there is a limit to this reasoning. The limit is: demand must be elastic. The reason McKinsey did not shrink when it automated my page-number sticking and my chart drawing is because demand for advice is elastic to price essentially, and there was latent demand that was waiting to be satisfied by people like me.

There are probably activities for which that is not the case. I imagine that if you are in charge of translating user's manuals in a consumer electronics company, the demand for producing translations of user manuals is limited by the number of products that you issue, and just because you do it more economically, you are not going to have more demand.

There are arguably a lot of things in organizations that meet that description. If you are an insurance company, you have people processing claims. A lot of this can be automated. On paper, you can let those claims processing people go, and the job will be done faster, better and obviously cheaper.

Does that create a "jobs apocalypse" in the short run? I doubt it, for another reason we have not discussed yet. It is quite simple: just because something is possible does not mean that it will happen. And specifically, just because a company can save money, and just because it's economically rational for that company to do it, does not mean that it will.

Again, I've been a consultant for 25 years. My job for many of those 25 years was essentially to go see clients and to tell them: "I have an idea for you. If you did X, you would get Y." X was usually something they could do with some consulting help, and Y was usually some savings they would get or some benefits they would get in other ways. Do you think that worked every time? Of course not. One of the things you discover when you're a consultant is that just because a company has the opportunity to save money does not mean that it does. There is viscosity in the economy. There is friction. There are people who leave money on the table. There are dollar bills on the floor waiting to be picked up. The economy is not a perfect frictionless simulated universe.

Why were my prospective clients not interested? The projects I was pitching were profitable, and way less difficult than changing all of your processes because of AI. I'm talking about tried and tested, low-risk, safe ideas. But, quite understandably, many executives didn't want to capture savings when that meant letting people go. It's certainly different in Europe than it is in the US and certainly different in some industries than in others, but it matters.

In Noise, you have a chapter on structured hiring, and you claim that rudimentary algorithmic decisions outperform human judgment. You touched upon that question. You know why?

In order to have a structured interview or simple algorithms, you are forced to define what you want. It forces you to actually list your criteria, prioritize them, give them a weight. As soon as you do that, you improve your decisions massively.

The promise of AI done right would be to do that. The promise of AI systems that most AI vendors are proposing is a radically different one that says, don't bother doing that, we'll figure it out based on your past decisions. The best you can say for this approach is, it won't be worse than the decisions we're making using our gut right now, and it will be cheaper and faster.

What are the skills that graduates are going to need in the future that universities are or aren't teaching today?

I don't know. I don't think anyone knows. But here's my thinking about this.

Essentially, what does AI do very, very well? AI is artificial intelligence. It does a very good job of simulating what we call intelligence, when we talk about human intelligence. And what is that? There is an old saying among psychologists, which is that intelligence is what intelligence tests measure. It sounds like a joke, but it's actually profound. Intelligence is what IQ tests measure.

What IQ tests measure is your ability to give the correct answer to questions to which there exists a correct answer. The question may be verbal or mathematical or other, but a test, any test, consists in asking you questions and making sure that you have the right answer.

Artificial intelligence does just that. It gives perfect or increasingly perfect answers to every question that has a correct answer.

So if our definition of what we learned at school, and our definition of what we hire people for, is their ability to give answers to questions that have a correct answer, we are training and hiring people whose skills can be replaced with AI.

The main debate that we see in schools and universities about AI these days is, how do we make sure that students do not use AI to cheat on their exams? But the question we should be asking is, if what we're testing them on, what we want them to have learned, can be done in three seconds by a piece of free software, are we teaching them the right stuff?

Now, there's an argument to be made that if you were talking about six-year-old kids, the fact that AI can do it does not mean that they shouldn't learn it. You should learn to think and you should learn to write and you should learn to count, because these are basic skills that you want people to have. But when you're talking about accounting tests for students in a business school, is it the same thing?

Some people will say: yes, it is the same thing, because if we do not keep learning these skills, we are going to lose them, we are going to let our thinking skills atrophy, and it's terrible. We hear about "anthropological breaking points" and "the end of the human race" and "cognitive surrender" and so on. Scary stuff!

My take on this is: yes, some skills are going to atrophy. No question about that.

But lots of skills have atrophied already. Can you extract a square root by hand without a calculator? Probably not. That's something that people used to learn. But rightly or wrongly, people have decided that it's a skill that is not so important that we want people to possess it. We can outsource it to machines. Does Google Maps contribute to the atrophy of my sense of direction? Yes. Is it a problem that I don't have any sense of direction whatsoever at all? For me, probably not a big deal, but if I was a Marine parachuted behind enemy lines, or just an occasional hiker, it is a problem.

My point here is, the question of what skills we think are truly essential, for whom, and when, is a question that deserves careful consideration and rational thinking. Just because a skill is a skill that we've always had and that was difficult for us to acquire and without which we couldn't live ten years ago does not mean that we should teach it to our kids. Can you guys build a fire without matches or a lighter? No. Well, guess what? Without that skill, you would have been dead 2,000 years ago.

One final question. Some of my friends who are much younger than me are still in university. Some of them are saying, my university just bans AI or my professor doesn't let us use AI. And we're having other conversations about places like 42 Abu Dhabi that have completely changed how education looks. There are no professors anymore. Students create their own curriculum. Also, people we're talking to are saying that universities need to drastically change how they look. So I just want your comment on these things. What should universities do about AI? Second, what is the role of the university going forward? And I want you to speak directly to a student right now that's thinking about studying something. What should they do?

Again, AI is about getting the right answer to questions that have a right answer. So if that's what school teaches you, it is wasting your time. Unfortunately, that's frequent. Sometimes they dress this up as, oh but no, we're not teaching you this subject, we're teaching you to think. Perhaps, but I'm not certain this is true of everything we teach in every educational institution in the world. So we have a problem. I don't have a ready solution, but we have a very real problem.

Some universities are experimenting with lots of different things. There is one that I know quite well because I sit on its board, that is truly trying to reinvent education for a world of AI. What characterizes it is that it's not about AI. It's about all the stuff that AI will not be able to do.

Because if you think sensibly about this, what you will want to learn is the stuff that AI cannot do and will never be able to do.

And what's that?

At present I see six things.

Number one, relationships. You want to be the sort of person who builds relationships, who has true networks, who has a personal reputation, who can get someone on the phone. Why? Because when people tell you AI is going to replace every salesperson, that's nonsense. What AI is going to do is, it's going to generate millions of emails landing in your inbox, which another AI will need to clean up. And you, the buyer, will never buy anything based on AI. Because it will be an arms race between their AI and your AI. But what you will do is, you will pick up the phone from the guy you know.

So the value of having relationships is going to be much greater. Now, in the short term, that's an advantage to the incumbent, it's a barrier to entry. In the long term, what it means is that you need to be the sort of person who knows how to build a network and to build a reputation over time. That's why it's valuable. That's one skill set that is glaringly absent from the curriculum or the testing criteria of most universities, including, surprisingly, business schools.

Second, you want to have initiative and entrepreneurship. AI can give you the answer to questions and can tell you, yes, this is a good business. It cannot start the business for you. It cannot actually do stuff. It cannot drive change in an organization. It cannot convince people that the decision the AI has suggested is the right one. Someone needs to take responsibility and to put their neck on the line and to say, this is my project, I do it.

That's the initiative of the entrepreneur who is an individual entrepreneur. But we need intrapreneurs in organizations who do that stuff. Otherwise, we have what we see, which a lot of people complain about: young graduates who, every time a question arises, ask the AI, give you the answer, and don't feel or take any responsibility for it. We need people who take the initiative.

Third, we need people who have curiosity. AI will give you the answer to any question you ask, but what question do you ask? Thinking of better questions is going to be a huge skill. That's one of my weak points, so I am constantly amazed by people who tell me, here's a question I asked Claude. And I think, wow, what a great question, I wish I had thought of that. That sort of creativity and curiosity and innovation in general is going to be super valuable because AI gives answers, it does not create questions.

Four, human skills. Human relation skills, caring for people, coaching people, motivating people. I told you earlier the example of, AI can tell you what pushups you should do and how many reps and so on, but it doesn't actually get you to do that. That's what a good coach does, because you sense that the coach cares. You don't want to disappoint her. You have a one-on-one relationship with someone, and that takes a certain type of skill, social skills and human skills.

Or you're in a meeting where a decision is being made. You could ask the AI, what should we do? But you, as someone with that sensitivity, sense that most people in the room aren't happy about the decision that is emerging. You read the room. You feel the signals. No AI can do that.

Five, ethics. Ethics means knowing when to use AI and when not to use it. And what AI to use. It means refusing to use AI when it's ethically unacceptable to do that. Which of course means that you have developed a strong ethical backbone to have your own sense of what is ethical and what is not. It might be stuff that your company disagrees with. Your company might ask you to use AI to do something that you think is inappropriate.

By the way, when you use AI, one ethical obligation you will have is to be accountable for whatever decision you make using AI. And to resist the very easy temptation of saying, oh, it wasn't me making the decision, it was AI, not my fault. So having a truly ethical backbone means knowing when to use AI, when not to use it, and taking responsibility for the outcomes in both cases.

And finally, the sixth thing is common sense or being street-smart. When I give this talk in conference rooms, there is a conference organizer. Apparently, most of his or her job can be done by AI. Sending an email to the people in this room to have them come, briefing me that the talk would be 45 minutes plus 15 minutes of Q&A. But there is one thing that the AI cannot do which the conference organizer did, and that has saved my life today: putting gaffer's tape on the electrical cord of the projector so I don't trip on it. That kind of last mile, the practical sense of people who actually do stuff in the physical world, is irreplaceable.

Note that these six things spell the acronym RICHES.

RICHES?

RICHES! There are RICHES to be captured for the managers who will possess those skills. They won't be automated out of a job by AI because they will have a more interesting job, a more rewarding job. A job that cannot possibly be automated.

And we need to figure out a way for universities and schools to teach those kinds of skills. Right now, all I can say is, management schools are slightly better at that kind of stuff than engineering schools or economics schools because there's more practical stuff, because there's projects, because there's associations and non-profits and your students' life in which people actually do practical stuff. So I can see how some of this is tackled by some of what people do on the side. But we can do better.

So that's my view. I could be completely wrong on any and all of those dimensions. I have been wrong before! But that's my thinking at this point.

*This interview has been lightly edited for clarity and readability. The interviewee reviewed and approved the transcript before publication.