Nuno Gonçalves Pedro

Introduction

Welcome to episode 79 of Tech DECIPHERED. Today, we take a leap into the big unknown. This is a thesis episode, not your classic analysis, in-depth sharing episode. The big idea for this episode is that we may be approaching the cognitive age, and how would one, or how would a company compete in a world of infinite intelligence?

The big idea, again, is that intelligence, which has been mostly scarce and expensive for all of human history, might become abundant and cheap. If that happens, what happens to work, what happens to companies, what happens to society?

This episode will be really framing a lot of these discussions. From knowledge workers to judgment workers, addressing the AI native company and how does that change, going into the human element and whether or not we’re underestimating it, and finally, ending up going into scenarios, feasible scenarios of a future where, well, intelligence is abundant. Intelligence is quasi-infinite or infinite itself.Bertrand Schmitt

Yes. Big questions for this episode 79.

From Knowledge Workers to Judgment Workers

We can start with from knowledge workers to judgment workers. Let’s go back first to how came the knowledge worker. It’s a 20th-century invention from Peter Drucker in 1959. The idea here is that that category might be splitting. The production of knowledge itself is on its way to being commoditized by AI. However, our perspective is that judgment around production of knowledge is not disappearing and is staying for a bit control managed by humans. What’s your take on this, Nuno? Do you agree with this split?Nuno Gonçalves Pedro

I think it’s a little bit more profound than that. It’s not just judgment. Definitely, human judgment will be needed. We’ve seen agents perform all sorts of funny things in the wrong way when left alone to their own devices. Even some very well-known AI researchers coming forward and saying, “Hey, I tried to use this myself, and actually I messed up some of my systems,” or “I messed some of my code. I messed up some of my flows for a period of time.” I think just having human-in-the-loop from a judgment standpoint will be needed for a significant amount of time. That is something you can’t just delegate into machines, into algorithms, et cetera.

The second part is, ultimately, there needs to be contextualization, and that contextualization, I think, comes from two forms. One from actual data, where the machine, I think, at some point will catch up, or the machines will catch up. The algorithms, at some point, on the data analysis will get better and better and have probably the closest to the truth that you can get, minus all the biases that are in the data, just to be clear, because data has a ton of biases. We’ve looked at this in the past and discussed it at prior episodes.

But maybe on that, I think the machine has a chance to catch up, or the machines have a chance to catch up, so there’s less of distinctiveness from the human standpoint. But then, on just the attributes, the ability when you’re judging some situation, you’re in the middle of the situation. You’re judging the person and how it’s acting, in some ways, a lot of the things that end up happening, end up happening because there’s human interaction. There’s someone on the other side. I see how they’re delivering the message, how they’re implicating. We’ll talk about it later in the context of the organization and what changes in companies.

I don’t think it’s just judgment. I think there’s a little bit more than that. One of the reasons I went to the dark side of management early on in my career from being an engineer was Peter Drucker and this notion of the knowledge worker, which he later on reemphasized with the publishing of his book, which for me was seminal and defined a lot of my career in life, the post-capitalist society, which is this notion that information rich and information poor is going to be the key distinctiveness that will happen in the world. The two big camps, information rich, information poor, which links back to this invention of the term knowledge worker, that knowledge is going to be key in some ways. I think that’s what we’ve seen for the last decades.

Again, I think judgment is not going anywhere, but I think it’s beyond judgment. There’s elements of humanity and involvement that won’t go away anytime soon, where human-in-the-loop are particularly critical. We’ll discuss later some scenarios, but for me, that’s my stick in the ground. I think human-in-the-loop is going to be critical for many decades to come.Bertrand Schmitt

While we are talking about all of this, and we share some possible scenarios, there is always that question. This is moving so fast right now. If you think about AI 10 years ago, AI 5 years ago, AI with the launch of ChatGPT 3, and then AI the past 2 years, now we have agents that are running at scale. Things are moving very fast. I can tell you, me in 6 months, the change has been pretty dramatic in terms of what I can use AI for.

There is always that question that whatever we are thinking about cannot just be connected to what we were able to do 6 months ago or even today, we have to think and project ourselves at least in the next 6–12 months. Of course, we can go beyond that, and we will do that with some future scenarios, but it’s a very fast-moving, and it’s not clear yet where are the limits.Nuno Gonçalves Pedro

I think that’s a very fair point. Let me try to analyze things that I don’t think will change anytime soon for the next few years. Agreed with you that many things will change, and we’ll have a lot better tools, platforms out there. That will be difficult to predict what exactly won’t change.

I think there’s elements of humanity, and some of them do relate to judgment, like having good or bad taste, having a view on it, on whether something looks good or bad. Obviously, all of this sometimes is subjective, but some of it may not be as subjective as people think it is.

The elements of contextualization. I think a little bit going back to what we did at Chamaeleon ourselves, where we built this platform, Mantis, and the objective of building Mantis was not really to replace us, was that it was a core augmentation layer in some ways that we would use investment or investor judgment as humans in the loop to systematize pattern recognition and a variety of other things, but that Mantis would really elevate all that judgment, not just in terms of timing, us being more productive, but also in terms of the quality of the decisions we’re making.

Think of it as a little bit like having our human judgment in the context of operating Chamaeleon at a higher altitude, where we are more aware of the things that are happening and how they actually happen. The ability to really get to the data pieces and then make decisions on top of that that generate the needed alpha in our case for investors.

What I mean by this is I think there’s always going to be core elements of humanity that I do think are going to be difficult for the machines to replace. For example, the taste piece people are like, “I can figure out what’s the taste in the market.” Yeah, but that’s mainstream. That doesn’t identify what’s the next big thing, which normally doesn’t start from mainstream. It starts from something else. It could start from opinion leaders and influencers. It could start by someone having a different way of addressing a problem and having a solution that hasn’t been thought through. For example, elements of creativity, I think, in human judgment and in human operations is something that I feel the machine will still have difficulty to replace.Bertrand Schmitt

Let’s not forget how today current algorithms are working by feeding them enormous quantity of data, actually as much data as we can find. Finding more data is becoming a limitation these days. What it means is that it’s very hard for AI to think beyond its training data. There is some level of logic that’s being added, but at the same time, take the launch of the iPhone. What was the opinion before launch? Is that no, it doesn’t make sense. Not enough battery life, no keyboard, no this, no that.

If you just base your analysis on what’s written out there, what’s being sold out there, you would just say, “It’s going to fail.” AI might really follow that more generic advice and perspective because that’s what in the training data and that’s what they’re in volume.

It’s, of course, raising a lot of questions of, how do you improve the quality of the training data? How do you separate the weed from the chaff? There are a lot of questions there, and obviously, it will get better over time. But it’s still a critical part of how it’s working today. It won’t be that easy to change.

I really like your point regarding Mantis, and I will say in general, platforms that you build with AI or leveraging AI capacity. Because when we say knowledge production is going to disappear, but we’ll keep judgment, it will be a different type of judgment because the quantity and quality of knowledge we will have in front of us to build our judgment will be very different.

If suddenly we have for free the work of 10 interns or 5 junior analysts or whatever, and you can run that on nearly anything you do in life or at work, it’s completely dramatic. Your judgment was not used to be exercised so often because often you were missing quality data to have a judgment. Before it was a lot of finger in the wind and trying to smell something, but you didn’t have enough to make a serious analysis. Except if you are working as a strategy consultant, as you used to do, Nuno. That part is actually quite interesting. That the judgment itself will be exercised much more often and hopefully on the base of much more in-depth analysis for a lot of things. We will work very differently.Nuno Gonçalves Pedro

We will go in-depth, faster and more fact-based, more data-based along the way. The question some of you might have right now is, is there some judgment that’s going to go away? Is there some judgment? We seem to be defining that there’s this organization, we’ll talk about it later, that goes from doers more into deciders. I think there’s some nuances to that, so I’ll just hit pause on that.

In terms of judgment, obviously, there’s judgment that has been hidden over the years under the pretense of being wisdom, but it’s actually not wisdom. It’s just repetitive tasking, and it’s rules-based for the most. There’s a lot of judgment done, in particular in the white-collar space, that you could say it’s just reps. People have been doing it all along like that, and so therefore to say, “I’ve done it before like this, so I’ll do it the same way.”

There’s actually no best in class, no analysis, no nothing. It’s just, “I’ve done it like that before.” I think that type of judgment will disappear because, again, algorithms will be as good, if not much better at that. They’ll be better at figuring out, actually, this would be the better way to do this. That’s how you play it forward.

Then the question is, if there are fundamental, wise people in the organization, people that can really take that more complex elements of judgment, how do you go from the world we have today, which is a world of apprenticeship, where people come out of college, they go and work, and they learn their way, and therefore, hopefully over time, some of them, not all of them, we know that, but some of them will develop that wisdom to be great decision makers 15, 20 years down the road? How do we do that in a world that now is saying, “I don’t need people out of college because I can do it myself, and I can do individual contributor, and I can have agents doing the work that would require some manifestation of management in the middle.” Basically, “I don’t need this stuff. I don’t need you.” It’s a little bit the story we’re in.

How do you create then this apprenticeship? How do we create then wisdom? My two cents on that is that wisdom, because of what we were just discussing and what, for example, myself and Bertrand was just saying, because of more often interactions with more data-stressed information and insights, what will happen is people will get better through their own reps in whatever form they’re doing, in day-to-day life, in internships, et cetera.

In some ways, that will create the accelerated growth. It’s a little bit the interactions with agents and the interactions with our beloved AI algorithms that will create that growth over time and maybe not as much with other people. That still leaves the question around social interactions, but that’s probably the way this gets sorted. Apprenticeship gets sorted through the machine and the human having more interactions in effect.Bertrand Schmitt

I agree with you because when we talk about apprenticeship, in some ways a lot of time was wasted on stuff that were not that important. But in a way, that was the price you had to pay in order to be there when people make the big decision to try to get some wisdom from that one hour of interactions that’s really useful and make a difference out of your full week. But the rest of your full week was just basic stuff that you had to do like a machine in a way. Why not let a machine do that?

That, for me, is a big question. You could argue there is a transition period where it could be hard. For instance, if you can work hand in hand with AI smartly while you are doing your 4, 5 years of universities, you could graduate with a very different knowledge, perspective, judgment, skill set than anyone who graduated 5 years ago.

I think that part will require a question around, “How do you change education?” You see what I mean? If you keep education the same way, expecting that the output is someone that should go now into 5 years of apprenticeship, that’s not going to work because companies will be, “No apprenticeship anymore.”

On the contrary, you have to come much more knowledgeable and ready to use the tools. The tools are so efficient that the bar pretty high. You need to come already very well-grounded. If the education is not doing their job, that will be trouble.

That part for me, I think is often forgotten. In some ways, the new-found importance of universities as a place to, and not just universities, the trade to really deliver people who are ready for the workforce. If on the business side, the expectation can change, of course, you have to change the education on the other side.

My worry probably right now is that it doesn’t look like universities are in touch with what businesses are looking for, businesses are working on. Of course, that’s very worrisome because the cost of university has increased very significantly. It’s not clear quality of education has improved at all. If anything, it could be the opposite. It’s pretty scary. Of course, it’s going to raise a lot of questions. How much is education worth in that type of situation?

Maybe another point because we talk a lot about apprenticeship, how this stuff was useful, but at the same time, if we go back in time, not long ago in the ’50s, if you wanted to be a developer, for instance, ’50s, ’60s, the job was very different. There was barely any programmation language out there. You had to use punch cards. Your time truly spent doing the coding was very limited. Once you had your stuff working, then, the debugging was a total nightmare.

My point is that no one is looking back to that time saying, “You know what? It was great. It was a great way to learn and to do an apprenticeship for 5 years. To do that crappy job of punching cards for the boss.” There was little value in this. Guess what? Everyone is happy it’s not being done anymore by anyone.

I think we also have to see what AI is bringing in a similar way is that everyone’s job is going to become quite different. There are a lot of big parts of the job who are not going to look back with fondness. Just looking back as, “Wow, that was very machine-like type of job. I’m glad I’m done with it.” People will want to jump directly to the next step. You don’t need to go to the punch card phase to be able to be a good developer for the past 40 years. I guess it will be the same with AI.Nuno Gonçalves Pedro

I think so. The difficulty we have as humans is to also visualize dramatically different scenarios and landscapes, professionally. It’s difficult for us to anticipate what are the jobs of the future. Jobs have changed a lot in the last few decades, not even the last century.

What people do, the migration initially from the agricultural society to then the industrial society to then the services society, and in some ways, the shift within the services industry, and now we’re seeing another shift, so we can’t really anticipate what those jobs look like.

Back to your point on education, because I think that’s a very important point. If you’re right now an undergraduate student or a postgraduate student, for that matter, and you’re not figuring out your own mechanisms of learning outside of your syllabus, outside of what your professors are telling you, et cetera, you’re going to face very difficult times.

If you’re not right now using all these AI tools proficiently, all these cycles of vibe coding, co-working, et cetera, with agents in the mix, you’re going to have a really tough time. If you’re not at this point in time as proficient as someone like myself or Bertrand, and given that we’re nerds, we’re relatively proficient with a lot of these tools that are out there. On top of it, some of us have our own platforms in-house. If you’re not as proficient as we are with those tools, you’re going to have a very difficult time because then people like us won’t need you. I think that’s the sad truth.

It’s like at some point, if you’re not needed, you’re not needed. Then again, you may find something else that’s more interesting for you to do. Start your own company, go join a new exciting job doing whatever it is that you need to do next, et cetera.

But again, I think the bar is very high. If you’re in college right now, again, undergrad, postgraduate, this is the time of transition. This is the worst time. It’s not the best time, it’s the worst time. Because education and all these institutions haven’t adapted to it yet. You need to adapt. You need to adapt. You need to adapt. If you don’t, you’re going to pay for it, not just in the loans you need to repay, but also in terms of actually having difficulty finding your career path in those first few critical years.Bertrand Schmitt

You need to be especially proactive when you’re facing this type of period where businesses are adapting as fast as they can because they all know it’s going to be survival of the fittest very quickly. Universities typically are working on a very different pace, and it’s pretty guaranteed they are not going to have adapted as fast as businesses.

In time of big dramatic change, it will be trouble. It will be trouble. Yes, you will have not fun. Not saying it was part of the deal when you sign up for that loan and decided to go for university. But that’s life. There has been issues before. It’s not the first time. You have to do something about it.

You talk about your perspective about, “Hey, why do we need you if you are not already fluent and very efficient with these tools and stuff?” The truth, in some ways, it’s even worse than that. Each time we spend with someone who is not efficient with all of this is less time we spend with the tools that are already providing magic for us.Nuno Gonçalves Pedro

Exactly.Bertrand Schmitt

It’s a very big choice of, “Hey, do I spend more time training this person?” Do I just… there is an opportunity cost. Or, do I spend more time staying at light speed? Why do I slow down to do something else in the hope that maybe I will get to return versus the light speed I’m already on? It’s a lot of tension. Again, it’s certainly new. But if we want to look back, I think you talk about the switch from agriculture and society, industrial society, and now the service industry.

The reality is that, yes, we have made dramatic changes in the past before. 140 years ago, we were 90% agricultural society in Europe, in the US, 90% of us. Today, it’s what? 2%. So my point is that that’s a normal evolution. There is no progress without change. Sometimes the rate of change is soft, and sometimes you have a step function.

Now it’s a step function, and it’s also a pretty fast step function. Before, it could take decades to get new stuff being put in place, to have electricity come up, this or that. Now we see that the rate of investment in AI is insane, way beyond anything we have seen before. Two, in a way, a lot of the architecture behind the scene was already there to support an even faster transition.

What’s new might be the pace of the transition, how unnatural it might look. But at the same time, if you put yourself in the shoes of someone who lived 150 years ago, I mean, this was also a dramatic change for them. From horses to cars to planes to rockets, pretty big change, maybe even bigger change.Nuno Gonçalves Pedro

Maybe the silver lining, just to bookend this section, is one, there will be new roles. There are a lot of things we can’t anticipate. There will be new roles, there will be new jobs being created, and new things that we can’t really quite grasp yet.

The second part is that the rules are changing, and they’re changing, I would say, in general, for the better. If you are a decision-maker or an organization, and you still have your job, you’re probably making more important decisions with more data, with more tooling around you, with less red tape, hopefully over time. I know that will not hold true for all the big corporations out there that are listening to us, but it is starting to happen.

Things are making an impact on how decision-making is made. There’s less and less red tape along the way in certain organizations. There are more and more fact-based discussions happening as we move along. The silver lining is better jobs, more jobs, different jobs in the future, hopefully as well.

Secondly, the second part of the silver line is that the jobs that exist today, hopefully, will be more interesting, certainly on the knowledge space and on this judgment space that we’re now introducing as part of this episode.

The AI-Native Company: Org, Hiring, Culture

Switching gears, maybe to how does that shift? How does the company of the future look like? How does an AI native company look like? I feel there are a lot of discussions on, “Oh, you only need one person to run everything.” Let’s not go to that level.

We’ve had a couple of episodes where we focused on AI as your co-founder and a couple of other elements that you guys can go back to. Let’s focus on a more evolutionary view of what’s happening to organizations, and maybe start with the org structure.

In general, we should see more flat organizations where mid-level managers have to justify their pay in some ways because middle management are routers. They are normally routing tasks. It’s sometimes aggregating it, synthesizing it, and pulling it back up. Guess what? AI and agents in general are very good at that.

The synthesis piece, et cetera, is not as well needed. One could say there are several elements of middle management that are valuable, like the coaching of people, the creation of apprentices, and the accountability that comes with some of middle management. But lo and behold, most of middle management is seen as a little bit of a thin line that doesn’t need to necessarily exist.

I feel we’re moving into a world of smaller teams, more senior teams, where there’s more judgment at the top, where you’ll have people that both do a mix of what we used to call management in its new form, but also a lot of individual contribution.

If you’re not used to that, if you’re not used anymore to be an individual in the future, again, and if you’re a very senior in an organization, maybe this is the right time to either reinvent yourself, find some other job that doesn’t require as much of that, which we’ll have plenty of those jobs for the next few decades, or maybe retire.

I’ve actually, shockingly enough, seen people who have said, “You know what? This thing is changing too fast, too dramatically. My industry is changing quite aggressively right now. I’m about to retire in a couple of years. I’m just going to retire now.” I’ve literally met two people who have done that.

Again, there’s nothing wrong about it. I think we’re, again, going through a step function and a huge shift, but figuring out where you fit in this new model of organizations, more senior at the top, smaller teams, more of a mix of individual contribution with management than ever was done before.Bertrand Schmitt

I agree with you. In some ways, I’m not surprised that some people might say, “You know what? It’s now time to retire.” I feel a bit sad, maybe because it means you don’t like to keep reinventing yourself and changing your habits and thinking about new stuff. You were a creature of habits, I would say, if that’s your conclusion.

But everyone is entitled to their own opinion, obviously, and a way of life. I guess that’s what happened, again, at regular times in the past in terms of big change. What I can see is that the rise of, you can call it the full-stack individual, someone who will have multiple roles inside the team. Before, you had to really separate the role.

Especially in the US, there is such a clear separation between every role you can have in a company. Let’s take a tech company. You will have people doing design, people doing different types of designs, people doing front-end development, back-end development, and operations.

You see step-by-step hyper-specialization. I have seen that, and it’s true that the level of complexity you had to deal with at some point requires some level of hyper-specialization because it will take you 6, 12 months in order to be really, really strong on a specific topic, a specific language. God forbid, trying to go deep into something that you had no real experience into.

But I feel with AI, it’s a big change, actually. It’s the opportunity to go beyond that. It’s the opportunity to do more, to touch more. You can combine designing and shipping code, product managing and shipping code, being an analyst and deploying. Of course, we have to think how it works because putting a marketer shipping code to production, maybe that will get you into trouble.

But I think that there must be some change. We see it changing dramatically, how fast we can get into something, something different from what we are used to. I think it would be crazy not to take that opportunity to dramatically change the scope of many positions and put an end to that hyper-specialization.

I think for me, in some ways, hyper-specialization was bad. There is only so much you want to be a specialist in because a lot of things, a lot of opportunities are actually coming from the mixing of many different ideas, many different perspectives, and you lose if you go to hyper-specialization.Nuno Gonçalves Pedro

I don’t think the age that is coming is the age of the generalist. I think it’s going to be the age of the multispecialist. We’re going to go into an age of multispecialization, which is a little bit, we’ve mentioned it as well in the past, what Amazon defines as an athlete or T-shaped or pie-shaped people, people that have on top an amazing ability to do general management, strategy, managing teams, et cetera, then have spikes. Spikes into business development, corporate development, product management, whatever it is.

With AI and with agents, the development of those spikes, as we’ve been discussing in this episode, will actually be easier. It’s almost like a given. If you want to go deeper and deeper into a certain area, you can go much faster.

I think that level of multispecialization is going to be really cool to observe. I’m not sure we’ve had an age of multispecialization over the years. Maybe people would point out, well, the Da Vinci example, people that are great across very different areas.

Maybe that’s an example of multispecialization. But honestly, from my perspective, this is going to be an exciting time because of that, because you’ll have people who, instead of being just focused on this area of sales, and I only do that, they can actually and should actually do a lot of other things. So the work, as we were talking before, can be more interesting. More demanding as well, because the judgments you need to make are more complex. The context you need to actually gain needs to be gained much faster. At a level of magnitude, you haven’t been able to do it before. Talk about information overload.

But actually, ultimately, the roles can be a lot more interesting, a lot more exciting, because I can jump around. If I’m an investor, in this case, we have two investors on this conversation. But if I’m an investor, one of the things that we start looking at is actually not just looking at a startup as, is this startup doing something in AI or not? Is it AI-enabled or not? Is it an AI platform or not? But actually, more fundamentally, is this an AI native startup?

Meaning, organizationally, culturally, is this the company that’s already in the AI age? How is the team working? How are they defining things? It’s not just that they only have two or three people. It’s like, what are those two or three people doing? How are they doing it? What cadence are they doing it on? What tools are they using? How are they making decisions?

I feel we’re still actually relatively early on that track. It’s very interesting because we’ve had all these companies raising mega rounds. First round out, we invested in one of them, but there have been many frontier labs out there raising a ton of money. But a lot of them don’t have a fundamentally different way of doing business. Of organizing themselves, of how they do the day-to-day. Although they’re working on cutting-edge stuff, with very notable exceptions, they’re actually not using it themselves. They’re not actually shifting how they do stuff themselves.Bertrand Schmitt

For me, that’s very interesting because in the past, I used to be quite conservative on how you manage and run a company in the sense that if you’re already in tech, if you are already on the cutting edge of what technology can deliver, and this and that, don’t waste time trying to invent a new org structure. Just focus on delivering something great, amazing, and be great at technologies. That’s already your huge differentiator.

At the time, there was no real reason to innovate on the team organization. I have seen so many teams that tried to innovate, and it was just catastrophic because there was not much to innovate on, because we had decades of optimization that we could leverage. There was no reason to invent.

But here it’s very different. There is a dramatic shift in how you can organize differently a company. I don’t think there are any blueprints yet on what’s the best way to do it because it’s too new. But at the same time, I would feel very bad to invest or support a company that first is not focused on AI or AI-enabled, but at the same time is not trying to innovate on the team itself. Because if you don’t do that, you’re going to get killed by someone who is going to innovate better than you on not just the product, but on the org as well.Nuno Gonçalves Pedro

Indeed. The shifts are pretty substantial. If you look, for example, just at hiring, what do you hire for? Certainly, there’s this element of the multispecialized orchestrator, which normally will be someone with quite a lot of wisdom and expertise.

It doesn’t necessarily mean someone who’s old, but someone who has the ability to work with all the AI tooling and platforms out there and be an orchestrator of agents. Why do they make judgments, make decisions, move stuff forward really, really, really quickly? Again, those jobs are going to be the best jobs.

The second part, I think that is very interesting, around hiring, is you’re going to skew towards the elements that are potentially either very aligned with the use of AI tooling and platform, AI expertise, or being AI native, or someone who’s used to using AI. That’s one side of the fence.

On the other side, you’re going to actually be optimizing to hire people that have the characteristics that will be difficult for AI to replace immediately, like taste and the notion of fundamental accountability and notion of implications, the notion of how you affect change in organizations, how you affect change in individuals, the elements of coaching, and beyond coaching.

You’ll be optimizing for those kinds of hires as well. Then, last but not least, for me, I feel that there is a momentum already happening. I think it will happen even more, which is the tendency to under-hire rather than over-hire. The moment of the good old days of blitz scaling, “Oh, let me go and hire 300 people to scale my go-to-market and just land grab market.”

Now, that’s not how it’s going to work. People are going to try and first get the efficiencies in-house with top talent and see if there’s, at the end, the need to hire more people or not, rather than the other way around. I think the issue here is a little bit of what we alluded to before in this episode.

There is a tax on individuals. If you hire more people, you’ll have to manage people, you’ll have to work with them, et cetera. If I don’t need to, I might as well work with the agents that the tools and platforms that I use give me access to. Because that’s a world that’s much more efficient, right?Bertrand Schmitt

I’m in total agreement with you on this. It’s definitely raising way more questions than before because, again, on one side, you have the product, the technology used to build products that are completely different.

At the same time, all of this is also enabling new ways to design organizations and to scale differently, especially in a world where, as we have seen in 3, 6, and 12 months, stuff that you thought were impossible are suddenly becoming possible. So you’re, “Hey, I’m going to scale and burn a shitload of money for 6 months before I know if there is any return.”

Versus, “You know what? Maybe I just wait 6 months. The AI has improved enough so that we don’t need this new team. We don’t need these people to do stuff.” Because actually, if you just wait 6 months, we will have stuff coming for free from either new AI models or new AI tools or this or that.

If you remember, we used to say that in mobile, things were going three times as fast as on the web in terms of pace of innovation and speed of development and stuff. I mean, with AI, it’s 5X mobile.Nuno Gonçalves Pedro

Maybe even more. Yes, well.Bertrand Schmitt

Maybe even more, maybe 10X. Every assumption around blitz scaling or scaling in general was based on past assumptions. It’s not based on how is the industry evolving today. Might make more sense for you to really grow your agents and spend more money on more tokens.

I remember, of course, Jensen is selling his business interest, but he was saying, “Hey, for each one of my 450K engineers, he better spend 250K in tokens a year.” I’m not saying it’s the right way to say it, but I think there is some truth in it, and that would be something to think about. Have we maxed out the token usage per employee? I’m not talking in a stupid way because token maxing and wasting money has no value and is as stupid as it gets.

But if you are truly getting a return on these tokens, can you use more? Can you generate more? Can you create more loops so that one engineer manages not just 10 agents, but 50 agents, but 200 agents? I think that’s the big question. We’re trying to add more people. More people means more management, more issues, more this, more that. That would be a fair question.

Another piece of the puzzle is how do you build in a way your… I don’t know if it’s a digital twin, but more like the digital version of your companies represented by agents. How do you make sure that everything you do as a business is truly captured, is truly leveraged so that your agents are getting better and better? Not just because the model gets better, but because you are putting more data into it, because it has more opportunity to learn, and as a result, gets better at your specific business.Nuno Gonçalves Pedro

The next big thing is culture. How does culture change? I think the biggest shift that I see is, why would you do meetings all the time?Bertrand Schmitt

Yes.Nuno Gonçalves Pedro

At least at Chamaeleon, we have a very small team, just by the way. We have a very small team at Chamaeleon. We’ve reduced by way more than 50% the time we spend on meetings between each other across the board, one-on-ones, partner meetings, et cetera. I think we’re really pushing to be more and more asynchronous. There’s stuff you can process via message.

I was just asking one of my colleagues, “Can you just send me that prompt for that so I can just do that on CoWork?” Or “Can I just go on Mantis and do this? Can you tell me the cycle?” Or vice versa. Basically, it’s a little bit like you’re just going to do it. I don’t need to meet. I don’t need to meet all the time.

There are some things where we still need to meet and interact, and we need to brainstorm at times, and we need to go to a different level of abstraction on the top end. Then on the lower end, there might be things that are a little bit more specific and governance-related and operational-related that we need to agree on that are more sticky. But otherwise, the culture is going to be biased towards build. “Go and do it,” rather than, “Let’s do a meeting.”Bertrand Schmitt

Yes.Nuno Gonçalves Pedro

Async is the thing. I’m more and more like we have a couple of interns this summer. “Can we async this?” They’re like, “What does that mean?” “Can we make this interaction asynchronous?” Because synchronous interactions for me are very expensive. Can you send me something that I can process, and then I can send it back to you?

We don’t waste time on you giving me context and whatever. Then I’m not ready quite yet because I need to process it. Maybe I’m in between two meetings that I’m actually thinking about other things in my mind.” Again, I feel that shifts how stuff is done. One, build rather than meeting. Two, asynchronous versus synchronous. In some way, millennials had it right when they shifted a lot to messaging and stuff like that. Let’s do more asynchronous rather than synchronous, those two elements from just an operating model of the company are significant.

Maybe this is a good time for me just to put one parenthesis because there’s this thing that’s bugging me as we’re talking here. Everyone who is listening to us at this point in time might be saying, “Cool, but I work for this large organization. We’re just now…”

Everything we’re saying here is contextualized by time. We’re giving you extreme situations. We’re looking into the future. Some companies that we’re talking about might be doing this already as we speak. Some of them might be in the process of doing this and might in the next couple of months be doing it like we are describing it here. Some of them might take years to get there. Then again, some of the companies that might take years might actually be destroyed in between or meanwhile, and be disrupted. Some of them might not because they’re in very legacy businesses, and it’s fine, and it’s okay.

Again, don’t take everything that Bertrand and I are saying today as this is gospel, and it’s going to happen tomorrow, and why the hell are we not doing it? We think that aspirationally, this is where you should be moving to as an organization, whatever size you’re at. Speed will matter, as we discussed before, but not everyone, obviously, is going to move as fast as we’re describing it here.Bertrand Schmitt

Yes. Me, for instance, take inspiration often with what some of the AI labs, frontier AI labs, are doing, the way they are working, especially in OpenAI and Anthropic. They are clearly at the top of the spear in terms of what is it that you can do because they have access to models we don’t have access to, because they have unlimited tokens they can use for tasks.

They hire people who are, of course, 100% on AI. They are the best example of what is achievable if you have the top minds, if you have the latest models, if you have unlimited tokens. From there, you can take that for our needs and for our situation, and others in industries that are not as advanced.

Definitely, you have some time. But as you say, things are moving fast, things are changing. Wall Street is going to expect better returns because when we discuss all of this, the conclusion is that you should be able to do more with less. That’s as real as it gets at some point. By the way, that’s what you see. You see better performance, a better business performance right now. So even if you might not get disrupted, you’d better start there. For some, it might take more time, and they might still be fine.Nuno Gonçalves Pedro

Maybe to bookend this section, clearly what we’re saying is organizations are going to change. Their MOs are going to change, the structures are going to change. There are elements of what we discussed before in terms of judgment that are fundamental to this. The ability that in some ways, one would say a lot of the technique of getting solutions out there, even in brainstorming or problem-solving, is going to get democratized. The algorithms are able to do that.

On the other hand, having points of view and having wisdom is not necessarily democratized, necessarily by the machines. It can be facilitated, it can be more productive in achieving that level of wisdom, but wisdom still will matter at the end of the day. We’re not saying that’s out of the question. Actually, that’s going to be the asset. People who have fundamental wisdom that can come to the table and frame things.

We see this even today in prompt engineering, on just creating prompts. The better your prompt is, the better the outcome is going to be, the result that you get from the algorithms. That’s not going to change, in my opinion, anytime soon. That UI interaction piece is not going to change anytime soon.

Again, if you’re an organization thinking through organizational structure, culture, if you’re thinking through hiring, these are some of the elements that we think will give you an opportunity, but I would actually go one step further. On the positive side, I would say, they give you arbitrage. If you’re able to move faster than your competitors and really adapt your org faster, you’ll reap the benefits faster as well. That’s what many still say and relate to as the word innovation. That’s how innovation gets accelerated.

I think there’s a huge opportunity right now for arbitrage. If you move fast, experiment, experiment on new org structures, experiment with talent, you’ll know that some of them will work well, some of them will fail miserably, so you can’t experiment on literally everything. On the other side, I think the doomsday scenario is if you don’t, if you’re on the other side and your competitor is outpacing you on trying these different organizational models, structure, hiring models, and operating models, they’ll potentially just disrupt you. They’ll do stuff that you thought you had the moat on, and lo and behold, you don’t anymore. Sometimes it comes just from org, just from injection of people with a different MRO, different operating model.Bertrand Schmitt

The Human Element: Are We Underestimating It?

Maybe we can move to our next section about the human elements. Are we underestimating it or are we overestimating it? The three things that are a big part of the human elements, emotion, creativity, and synthesis. Is it just soft skills, replaceable part? On the contrary, is it the durable part now that we have automated intelligence?Nuno Gonçalves Pedro

I’ll start with emotion first because I think it’s probably the easiest of all the ones you’ve mentioned. Emotion is key. Many of you listening to us will know this. The way you deliver a certain message, the emotion that you have when you deliver it, just in and of itself, this could be a sentence, it’s something verbal, et cetera. Makes a difference between the person or the people on the other side actually adopting it or actually just resisting it.

Emotion is critical. It’s what runs the world. Everyone talks about a bunch of things, but emotion is a currency that is still naturally human. It will be, I feel, difficult for these AI tools and platforms to recreate it fully until there’s some literally very high-definition manifestation of them as avatars or some physical manifestation of them as robots and all that stuff. It will take a while for that emotion to be manifested.

Emotion, I think, is still something that we as humans have as a moat, and it’s critical. As you mentioned before, I was a strategy management consultant at McKinsey, and getting people to action is actually 80% about the delivery, communication, the emotion that you surround the project itself, more than sometimes the truth. It’s great to have the truth and to have something that is similar to the truth in terms of analysis, but in some ways, that’s not what really moves change. Change is moved by, I would argue, a significant amount of emotion and alignment on emotions.Bertrand Schmitt

You could argue that’s something that most politicians have perfectly understood. If you look at most campaigns these days, everything on emotions, maybe the tagline might be one word. It’s interesting when you see from that perspective that actually it’s very little on facts, very little on all of this, but more about emotion. You could argue it’s the same for businesses in the future? That’s a fair question. I think creativity is another one that’s quite important. At the same time, it’s not so easy because I must say I’m quite amazed when I’m looking for creativity from AI, either to generate the image, to generate video, to generate audio, or to generate text. AI can be pretty creative.

I still think you need to control its creativity; you need to understand what’s good, what’s bad, what’s quality, but at the same time, I can see even in creative tasks, AI can be a very strong partner. I’m talking about any creative task, like invent a name for a product, let’s brainstorm the mission for the company. AI can actually be doing a pretty impressive job. That’s the type of job where you will hire experts, where you will use some of the best people in your team to help you for days. We say, “You can do quite a lot.” It’s an interesting one because I think there is some unique human creativity, and at the same time, AI can be pretty strong at creative task as well.Nuno Gonçalves Pedro

I agree. In particular, if it represents benchmarking, if it represents repetition, if it represents seeing the world and then coming up with something that presents itself as creative, to be honest, it can actually outpace humans. If it’s like genuine light bulb moments of creativity, angles that haven’t been tried before, certainly not in the same way, I think humans still have the advantage. To your point, I agree. This is not a humans-win situation.

On the previous one, on emotion, still, part of it is because, also on emotion, there are exchanges. You and I might be looking at each other, and from the facial expressions and the reactions, where you judge that for AI to get there, it’s going to take a long time. There’s going to be a lot of very complex algorithmic stuff put into that for AI to be able to create synthetic emotional behaviors, but creativity, I agree with you. There are a lot more nuances to it today, where AI does have significant advantages at the end of the day.

Synthesis depends. Synthesis, I feel, if we’re talking about holding a bunch of messy assumptions, contextualized inputs with different layers of data adjacent to them and then trying to create and form one coherent, fully accountable point of view that you stake something on, like a decision, a company, a business unit, whatever, I think humans have the advantage.

Part of it is the complexity of what we have today with generative, pre-trained transformers, today with GPTs, where the hallucination comes through, where it’s really more statistical analysis. Over time, maybe synthesis will be a forte for AI. Right now, I think we still have that ability to really be the ultimate decision-makers and judge-makers and have that wisdom put at the table to make those decisions.

Honestly, models are very good on balancing both sides, so ended up, as we say in Portuguese, neither fish nor meat. It’s to balance both sides’ answers. That’s not helpful in most cases. When you’re in a difficult position where, for example, the future of a company, company is almost dying, what do you do? I’m not sure your AI algorithms that are going to give you a great solution. Because it will give you a median or average solution, which likely will lead you to a median or average outcome, which in this case would be failure.

Again, on synthesis, there are some areas of advantage for human beings. If you are looking for clearly synthesized perspectives on certain elements that are maybe less edge-focused, they’re more than the normal part of the normal distribution, then probably AI agents are brilliant at that. All the tools we have today are pretty good at that, and I think they’ll just get better over time. That’s how I see synthesis.Bertrand Schmitt

I think a lot of improvements will come with a better fine-tuning of agents to what’s special about your company. Because if you just take a general agent, there is only so much. It can understand your industry, your company, and your way of working. I think that part of making sure your agents are finely trained, finely tuned on your own business, so that they can give you a really well-calibrated feedback, will have a lot of importance.Nuno Gonçalves Pedro

I think that’s absolutely spot on. Maybe to end it, what is definitely different about humanity? Definitely, emotion, as we discussed, some pieces of synthesis. Creativity, maybe the light bulb creativity, not the more repeatable creativity, the one that you can put and encapsulate into processes in some ways. There are elements of us being physical, which robots can’t still recreate. That’s definitely an advantage. The embodied, we’re embodied. That’s obviously a huge advantage.

With that also comes advantages because we have to interpret each other, and we have to see the complexities in physicality that land to it. Is human and the human element categorical difference? If we’re having a more philosophical discussion around this, I think it is. I think it will be for at least the foreseeable future and maybe decades to come, even in whatever scenarios we’ll discuss, which is our next section, scenarios.Bertrand Schmitt

I would say projecting beyond 10 years is always pretty hard on this because, again, some of the improvements we are talking about we can imagine based on how it has evolved, but at the same time, there will be disruptions in AI. Stuff that we take for granted in terms of weakness, especially, might not be there in a few years from now. Either because it has been solved through brute force or because the field will have made significant change and improvements and discoveries, making some of our points moot.

If we talk about embodiment, obviously, robots are coming. How fast, how cheap? That will be a big question. Right now, they’re not very smart. They’re usually very specialized. The more we move to a more general form factor, humanoid form factor, the more I think it will change. Also, another piece of the puzzle is that we have the assumption of agents having trouble to convince humans and stuff. At some point, we keep assuming that humans in the loop. If we’re talking about agents convincing another agent, not having embodiment might be even more efficient. That will be another perspective.

Going forward, we will have not just agents we control who are doing a job and scanning the job, but agents truly interacting with other agents. You have agents controlled by one person, one team in your company, working either together or maybe not confrontationally, but trying to think and having different perspectives with another agent, controlled by other teams. I don’t think we have seen much of that now. We have seen mostly agents that are controlled by one team doing one job in one direction. Not multiple teams agents working together, or against or in parallel with another team agent. I think we will see some interesting things coming out of that.Nuno Gonçalves Pedro

Scenarios

Switching to scenarios, we love our two-by-twos. We haven’t done one in a while. This time it’s a two by two. We have four scenarios. I think on one axis, we would have potentially the capabilities of AI. One side would be more incremental. The other side would be the extreme full AGI. I’ll define it in a bit so that we can at least have a little bit of a definitional view on what the AGI is. Then the other axis would be how gains are distributed, concentrated versus broad. Obviously, if they’re very concentrated, it’s more unequal. It only goes to a few companies, a few people, a few individuals. If it’s broad, it’s much more dispersed through society, et cetera.

AGI, just to try to define it, the formal definition of it is that it’s a hypothetical AI that matches or exceeds human capabilities across virtually all cognitive and practical tasks. In some ways, AGI can learn, reason, and adapt to novel situations across any domain. Then there are several mutations on this, but there’s one notion, or rather, there are three notions that normally are across a lot of these definitions. One is generalization, ability to seamlessly transfer knowledge from one domain to another without needing retraining, which is a very impressive skill that we humans still seemingly have.

Autonomy in agency, the capacity to operate independently, set goals, plan and execute complex tasks. I think AI is their issue with agents to a lot of that extent. Then, last but not least, human parity, performing economically valuable work at or above the level of a typical human knowledge worker. If you listen to one of our last episodes, you’ll realize that Bertrand and I have slightly different views on AGI, and if it’s already here or not.

I think, definitionally, maybe we have slightly different views on what the definition actually is. For me, maybe AGI is a little bit more what some would call superintelligence and generalized superintelligence. Strict to census, Bertrand is more connecting to AGI as in its prime definition. It behaves as well or better than a human thing. Maybe that’s what’s leading us to differences on whether AGI has arrived or not.Bertrand Schmitt

Personally, I will have a different scale where I will put AGI, as you just said, in some ways, relatively similar in performance to your average human being. On top of it, it’s able to touch different domains that most humans are not able to do. Usually, there is some level of specializations where in AI, it can be more generic. I will put ASI, Artificial Superintelligence, as clearly the step beyond. Something that, on any dimension you pick, it’s able to beat a human expert.

From my perspective, I think we already discussed that, but we are at AGI already. We have AI that can do way better, not just way better, but at least as well as humans on many topics, sometimes better. Yes, there are some topics that are not for AI yet. Embodiment, for instance, to flock with your humanoid robot in 2026. For me, we are partially there or fully there in AGI. If we take the stricter definition, ASI, we are definitely not there, but my guess is that it’s moving quite fast. We might be there in a few years from now.

I don’t think we are talking about multi-decades. It’s 5 years, maybe 10. Of course, there are questions because people will say, for instance, “Hey, how do you become truly super-intelligent when all your training is based on human data?” That’s not an easy one because how do you train on that? To be way better, not just a bit better, but way better. Maybe I’m going on a tangent, but some are looking at AI learning from AI, AI being taught from AI, AI fighting with AI, AI challenging AI.

The same way we saw this AlphaGo moment where AI was not trained anymore, like in chess with human moves, but has been trained to play against itself. That’s when it reached superintelligence in Go. It reached superintelligence by playing against itself and basically letting go of that human baggage, if you want, and going to the next level. What I found interesting in that, actually, first, that’s what happened, but two, there was some analysis that the average level of Go players and the top players went up after AlphaGo because AlphaGo, in a way, opened doors that humans didn’t believe were open in front of them, or they didn’t see them. They didn’t see these doors, so they didn’t bother to open them.

AI opened new doors, but interestingly enough, humans improved after that, thanks to AI. You see what I mean? It was an interesting, okay, that self-learning from AI was the way to go beyond the current level of human knowledge and human expertise, but at the same time, humans were able to follow up. It was not like suddenly humans are totally useless crap. They improved. Did they still beat AI? Maybe not, but it was definitely also helpful.Nuno Gonçalves Pedro

Back to our scenarios. We’re going to take the definitional extreme just for argument’s sake for scenarios. We’re going to talk about maybe what you were saying, ASI rather than full AGI, but like ASI. Again, artificial superintelligence as the extreme on the one hand. Let me talk about maybe the first scenario that would come to mind. Maybe we can call it the plateau scenario. All of this was great, but it was all smoke and mirrors. They were great at some cognition stuff. They’re a great tool. At some point, they’re going to hit a wall. Hallucinations are never going to be a thing of the past. We can’t fully trust them on really hardcore stuff.

We’ll gain productivity enhancements. We’ll keep gaining those productivity enhancements, but at some point in time, we really won’t reach ASI. We really will be stuck with what we have. It’s a little bit like we get the next big thing, the next big spreadsheet, the next big internet, but it’s not going to change the whole world beyond just productivity, enhancements, and amazing tools that we have available to us that makes us much better.

In that scenario, the winners will continue being fast adopters, probably small and medium businesses, because there won’t be a push for maximum speed either, so they’ll catch up at some point. Then AI native companies will be better companies than other companies, but not necessarily overall disruptors across the board. It’s not necessarily a new species of companies. It’s just companies that are a little bit better at doing stuff, which we also saw during the internet phenomenon and that first big push forward and then bubble, where we had some companies that were fundamentally different on how they operated. It took us another couple of decades for companies to be more and more digitally native along the way.

Basically interesting, but it’s boring. It’s like, cool, we got tools, we got promised the world. What are the implications? All these companies that are worth trillions and trillions of dollars are not worth trillions and trillions of dollars. Because at some point we’ll face competition, commoditization. It will just be tools and platforms. They will not unlock that next stage. Therefore, this will have been a bubble, and likely it would be a hard landing to that bubble. That’s the implication.Bertrand Schmitt

I would just say that, yes, I agree with you, but I would just say overall, even if it stopped today in terms of quality improvement, speed or stuff, or it barely improves, I still think we will have 10 years of madness just to leverage everything that we have today.Nuno Gonçalves Pedro

Understood, Bertrand. This is a scenario. I understand, but maybe we’re going to hit a wall, and we’re going to hit that wall next year, or we’re going to hit that wall in 2 years or whatever.Bertrand Schmitt

Possibly. I’m just saying we still have 10 years of goodness from that big push in AI we experienced the past few years.Nuno Gonçalves Pedro

Absolutely. Agreed, but it’s boring.Bertrand Schmitt

It’s boring. It’s a plateau.Nuno Gonçalves Pedro

It’s a plateau. The second one is more of something that we have AI, but humans in the loop are going to be critical along the way. The judgment work that we described earlier in the episode is going to be critical to everything that happens. It’s, I would call it the augmentation scenario. The AI will be a great augmentation tool for humans, but humans will never really quite stop being in the loop. Some of the gains that AI has are broadly distributed in society and in the startup, big corporation and small medium business world. Everyone will have access to them.

We humans, are still very important. We have all these augmentation things, and AI is mostly benign. There will be a couple of issues, but honestly, at the end of the day, we’re just better. We’re better, faster, more data-driven, more factually current. We’re doing stuff faster, but humans are very much at the center of everything that gets done.Bertrand Schmitt

I think that’s definitely a possible scenario. I would say more likely than the plateau scenario. I don’t believe in the plateau scenario at this stage. I think we can move there. I certainly believe that everyone will benefit, the same way that everyone benefited from the agricultural revolution, the industrial revolution. Everyone ultimately benefited. You might have some individuals who really, really have trouble to adjust. Overall, the products they will still buy, the services they will still use, their kids, things will get better because there will be new opportunity created.

Again, let’s not forget, it’s always easier to see what’s going to get lost than to imagine what’s going to replace. It’s always that paradox. That’s why some politicians try to take advantage of this, trying to just show the bad stuff that happened. That’s what might get them elected. That’s not the good stuff that’s coming down the road.Nuno Gonçalves Pedro

The third scenario, maybe at its extreme, would lead to the doomsday that some politicians do talk about, which is a bifurcation scenario, a scenario where the world gets to some types of ASI, but it gets unevenly distributed. Some players that own either platforms or infrastructure or energy or all of the above will control that. Because they control it, they will have massive productivity enhancements.

Labor as a force and as a global force will lose a lot of its leverage. In some ways, the social contracts will just get strained. It’s like the big guys that have made a lot of money, they won’t necessarily share that with the rest of the world. The people that are struggling in the middle class that might have been evaporated is having difficulty making ends meet, and there will be a huge social constraint, et cetera.

In and of itself, as a scenario, it doesn’t need to necessarily mean that it is a doomsday scenario. At its extreme, it could be a doomsday scenario. Where very few companies control the world, where the social constructs are destroyed in many parts of the world, where wars just start getting massified, either through civil wars in many places or through actual wars between different countries. That might lead to a doomsday scenario. But the scenario itself of bifurcation doesn’t necessarily mean doomsday. It’s just added as extreme could be a doomsday scenario.Bertrand Schmitt

Yes, I think it’s difficult to dismiss that scenario. I just put a very low likelihood given again, we now have centuries of economic revolution, scientific revolution to look at in the past, and it never happened. The only thing that happened is more wealth, more health for countries that pick democracy, that pick science, that pick capitalism.

I’m personally quite hopeful this will not happen. However, I could say that bad regulations will create that. I’m personally very worried about what’s happening in Europe. Very bad decisions in terms of energy choice, very bad AI regulations that can create part of Europe as a permanent underclass. If you don’t have energy to run data center, if you cannot use AI, if you cannot run AI, that will be trouble because US and China are not going to wait. I think some of this could happen.Nuno Gonçalves Pedro

Honestly, it could happen. My biggest concern is not so much regulation. I think my biggest concern is the unchecked companies, that there’s a couple of unchecked companies that decide to just create monopolies around whatever they’re doing, and it will be difficult to take them out of place, in particular in political environments that are a little bit more appeasing to such companies.

Actually, the whole Skynet scenario for me is not totally impossible. Probably not quite in the way that Skynet happened, but it’s not totally impossible. We should be on our toes on this stuff because, honestly, if we don’t think through a lot of the implications of what we’re doing, we might well end up there.

Last scenario is a scenario of full abundance. We are at ASI, Artificial Superintelligence, and everything’s broadly distributed. Everyone has access to it. All the gains are deliberately distributed. Either through policy or through the great companies of the world that decide to lead the way and share their profits, not only with their employees, but with humans at large.

In some ways, we either pivot to doing other signs of activities or we work less days a week, or there’s other elements of this, like universal basic income or other things that come to fruition that might appease the masses, so to speak, and everyone’s happy in a genuine bunch of companies that, albeit our overlords, they are very good at basically distributing that wealth that gets created over time.Bertrand Schmitt

I don’t like this scenario, personally. It might happen, but I don’t think it’s the right way to go. I think that, again, I have that belief that capitalism and science will create opportunities for everyone. I don’t think there is a need for companies to share their profits beyond their employees and shareholders.

I think, ultimately, if companies provide a good product, it will benefit the people who are using their product. That’s how you work as a business. But I can imagine definitely a world where even better AI, stronger AI, can be broadly distributed and have a lot of benefit. But I think it’s good if human beings still work, do something meaningful. We don’t create a class of people who are just entertaining themselves through the fruit of others. I don’t think it ever ends well.Nuno Gonçalves Pedro

The Catholic in me, just to be clear, thinks the full abundance is heaven, and that will become after.Bertrand Schmitt

It’s a different story, yes.Nuno Gonçalves Pedro

Maybe it doesn’t need to happen here. Here we need to toil and work, but it would be a good scenario, I guess. I agree with you. It’s probably the most unlikely of all the four scenarios. If I had to put some probabilities to each of them, I would say second, probably least probable scenario, and we’re talking about the next 10–20 years, which is what we can have some visibility on, not the next century, would probably be the scenario we just described around bifurcation that could end up in that doomsday. It is higher than the probability of full abundance, I think. Actually, quite higher, but I don’t think we’ll be as high as other scenarios that we have at the table.

I would put the first scenario I discussed, plateau, as the second most likely scenario. It’s a little bit like an inflection point. We hit an inflection point, we’ll get productivity gains, we’ll hit a wall, we’ll get through, to your point, the use of that productivity over the next decade or so, but then it’s another inflection point to get to the next level and next layer that will take us finally to ASI.

Last but not the least, I think the most probable scenario, if I had to, again, look 10, 20 down the road is augmentation, where human loop will continue being critical and where these AI agents and platforms will be fundamentally shifting how we do things.

Again, optimistic. Both those scenarios are relatively optimistic, to be honest. Augmentation would be probably, I would allege, the most optimistic for human beings, probably the most positive for humanity in the broad sense, where we all have roles, and we all have importance in whatever we’re doing. Obviously, with new jobs and new job descriptions and new things that we’re doing. That, I think, would be my probability adjusted answer.Bertrand Schmitt

Yes, I agree with you. The full abundance is not so believable because there is always some limited resources. There is always a want for more. I’m not super sure it’s ever possible. The plateau, I think, is really unlikely. There are still so many ways to improve.

We might plateau in 10 years. That’s a different story. But I can see that acceleration and slope we’re on would take many years before slowing down. I’m like you. I certainly believe, and especially the optimist in me, the historian in me, believe that strong AI, broadly distributed benefit is a likely scenario with a possible risk through some more bifurcation between have and have not. But again, I think this situation would happen more if we have too much government control and mistakes in term of how to let the best companies win and let the market be efficient.Nuno Gonçalves Pedro

Our Take

Key takeaways from today, and before we move to conclusion, obviously, intelligence is getting cheaper and cheaper or perceived intelligence. Maybe we’re just talking about analytical capabilities or other things. But as it gets cheaper, some elements of judgment, taste, and even accountability are obviously getting more expensive.

I think we have a very positive long perspective on the human element, that the human element will be distinctive, will be able to defend itself, and cultivate its role in all sorts of processes of decision-making, obviously, including in the professional and work environment.

It’s not safe by default that we are to be in the loop, the humans, but we feel that there’s a high chance that we’ll be in the loop, and we’ll have something to give to these processes and to these judgments and these decisions that we’re making.

We feel that it’s a time for you to take some qualified risks. It is important to see that there are a lot of productivity enhancements in the market right now. If you are not acting more and more as an AI native company and evolving your org structure, your culture, your hiring practices, et cetera, you may be left behind. That might be one of the first angles of disintermediation or disruption by new entrants.

We feel it’s a no-regrets move in some ways. AI native matters. Being AI native as an org or a company. Then last but not least, if you’re not using these tools, if you’re not playing around, if you’re not using platforms and using these systems, get ready to be disrupted yourself.

Unless you have some of these core skills that we just discussed in our episode that don’t require any use of AI tools, you’re going to be facing a difficult uphill battle along the way. Again, if you’re in the midst of the AI transformation, and you’re not using it yourself, it’s going to be difficult to justify a paycheck a couple of years down the road.Bertrand Schmitt

I think at the end of the day, AI is here, whether you like it or not. AI is going to keep getting better and better. What happened the past 4 years was nothing short of extraordinary in term of AI improvements, CapEx investment in AI. I would be shocked if it’s not growing even faster from here.

Of course, there will be jobs, little impacted with AI, but definitely, if you are a knowledge worker, your job will be impacted by AI. I would guess, nearly anything you do in 5–10 years from now, will be totally unrecognizable if you were a knowledge worker. The job will have changed completely dramatically.

If you are a business, you will have to adjust and adapt because there is simply no other choice. It’s a capitalist system. People will find the best tools, will learn the best tools, will use them, and will try to beat you and will try to be there. You have to be part of it, you have to make it work.

I think overall for society, this would be very incredible improvements to what we can build, what we can deliver in every industry. Personally, I’m very excited by that new cognitive age in front of us, assisted by AI or expanded with AI. I don’t think AI will supplant us, but definitely it will be a big chunk of our future workforce. If you think about it, there is nearly no limit to how much we can scale AI.

Conclusion

It was good to talk basically from knowledge workers to judgment workers. Talk a bit more about what would look like the native AI company. Talk about the human element that humans can uniquely bring to the table. Finally, we went around four scenarios about where AI will bring us. Thank you, Nuno.Nuno Gonçalves Pedro

Thank you, Bertrand.