The human edge in AI-driven investing: A conversation with Arnaud de Servigny, investor, academic, and author

Welcome to The Heidrick & Struggles Leadership Podcast. Heidrick is the premier global provider of diversified solutions across senior-level executive search, leadership assessment and development, team and organizational effectiveness, and culture shaping. Every day, we speak with leaders around the world about how they're meeting rising expectations and managing through volatile times, thinking about individual leaders, teams, organizations, and society. Thank you for joining the conversation.

Daniel Aghdami: Hello, I'm Daniel Aghdami, a partner at Heidrick & Struggles' Zurich office and leader of the Family Capital Practice in Europe and Africa. Today, I'm delighted to be joined by Arnaud de Servigny, investor, academic, and author, with deep expertise in quantitative investing, machine learning, and the future of asset management. Alongside his work advising investors and institutions, Arnaud teaches machine learning in investing at Imperial College London, where he has a front-row seat to the rapid evolution of AI and its implications for financial markets and investment leadership.

Arnaud began exploring machine learning applications long before the current AI wave, initially developing predictive models in entirely different arenas, before applying these capabilities to investment management and portfolio construction. Today, his work focuses on how AI and data science are reshaping investment processes, organizational models, and the leadership capabilities required across the industry. 

Arnaud, welcome and thank you for taking the time to speak to us today. 

Arnaud de Servigny: My pleasure.

Daniel Aghdami: So let's get going. Arnaud, you've been working at the intersection of investing and machine learning for many years, well before AI became such a mainstream topic. From your perspective, what has changed most dramatically in the investment industry over the last few years? 

Arnaud de Servigny: Look, there are many different things that have been changing. We've been shifting the focus from the sustainability to machine learning very recently. It doesn't mean that sustainability is of no interest anymore, but it means that, at the forefront, AI has been becoming something that is hot and topical, for different reasons. In fact, people are eager to generate additional revenues, alpha, and are also eager to be in the space of gaining efficiencies. And you can't be an investor into these new trends in the market without thinking that for yourself it is going to have an impact. So this is where we are.

Daniel Aghdami: Wonderful. And you also teach at Imperial College London. How have your lectures and discussions with students evolved as the pace of AI development has accelerated? 

Arnaud de Servigny: Well, they've evolved quite a lot. I'm going to tell you a story of how it evolved practically. Generally, students are being assessed both by the end exam and through some coursework that is teamwork, and the coursework is done by people, by students who have got access to everything. Two years ago, we told them you're not allowed to use LLMs or chatbots and things like that to do your work. The end game was that, [with] marks out of 100, the average mark was 90 out of 100. So that means that everybody has been using, and I got—the administration of the business school was not very happy with my achievements. Last year, we told them you are able to use this, but the topic is going to be more difficult, and that got us back to average mark of around 60.

However, last year, you were still—the thinking was that you would be coding things on your side. So this year, new change. You don't need to have any coding capabilities. The LLM is going to do that for us, and we're going to train you on how to use the LLM in a smart manner, so that you do this in—you have some code that is relevant and is production-grade. 

So you can see that, in three years, we've moved away from “we don't want to hear that” to now let's use it properly as part of the process. And in fact, this is something that is natural and there is no way to not use that, because the implied productivity gains are in the range of 1 to 10. That means that you can do much more with this than what people did 2 or 3 years ago. In addition to that, in fact, 2 or 3 years ago, what the LLMs were doing was not that great. Right now, the level of quality has improved. So, and it will change in another 2 to 3 years. So, in fact, we really need to—it doesn't mean that it is the machine that is going to do all the work—but we really need to think of the interaction between the person, the people, the students, and the machine to have some sort of team-like effort rather than just one or the other. 

Daniel Aghdami: And we're talking specifically about use of AI in investing today, or the investment business, and probably most often we hear about AI being used to improve operational efficiency. To what extent are firms actually looking at using AI in the investment process itself? 

Arnaud de Servigny: So I'd say to go to the first part of your assessment, asset management companies are companies like others. So all firms, I believe, are looking at efficiency in their processes related to the usage of AI. I mean more the operational type of things. The question you're asking is whether the AI type of activity becomes really at the heart of the asset management business with the investment processes that are being informed by AI technology. 

Daniel Aghdami: Exactly, yes. 

Arnaud de Servigny: What I'd say is that it's very clear when I look at the hedge fund business, that this is totally part of their activity. This is they live by it; they use it. They've had to think about two things, which is making sure that what they do is protected and is not going in the open, and secondly, there is a point which is around accountability. Who is accountable for things if the machine doesn't go the right way, etc.? How does that work? 

Now, I believe that, in the common asset management business, people are lagging, and there are reasons for it, which is that this is an activity where you tend to hire people who have been successful in the past. If these people have been themselves asset managers, they have gone their own way to manage money, and they think that the best practices are—they know what it is, because it corresponds to what they've done. So, to some extent, it takes time for the firm to adapt to a new environment with people at the helm who've had experience in the past. 

If people are more on the commercial side, they're more flexible to some extent, because they'll see what is going on elsewhere and they will be—they haven't got their own way to manage money in the past and things like that—and they will be challenging their internal organization to say, “Well, you know, this and that type of firm is doing this, why don't we do it?” Now, everybody will have to move or is moving, but at a different pace. 

Daniel Aghdami: Interesting. So that leads me on nicely to my next question. I was going to ask you, you know, if we're looking at the investment landscape or the different participants in the market, which of those investment organizations are actually best positioned to make the most out of this technology? Is it more the larger ones or the small ones? You mentioned now everyone needs to, but what are the impacts? What differentiates them? 

Arnaud de Servigny: So, let me answer it in a different manner. I think that when I look at university, it's clear that all the youngsters want to use the new techniques and this, but they haven't got the experience of the market that is needed. When you have people who are with a lot of experience, who've gone through various crises and things like that, they understand, they've got a feel for what the market is. Now, they don't have the technique. So the challenge is being able to bring together the experience and the expertise with the innovation, the old with the young, where, in fact—and this is the blend that is needed, the optimal blend is changing.

Generally, the small firms are firms that contain a lot of young people. That means that they've got a lot of good ideas, but it's maybe the level of risk they might go through is quite high because they haven't seen crises. They don't happen so often and, in fact, the risk side of what they are doing is perhaps not as good and not as great or [at the] right standard.

If you look at the existing firms, you've got people who have built a name for themselves, who've got their reputation and have been always doing the things the same, in the same way, maybe feel challenged because the performance is not as good, but they already have been challenged for quite some time with passive investing. OK, so in fact they live in an environment where they have got challenged. They've got challenged so far by passive, by indices. They're going to get challenged by active AI-driven investing, too.

Daniel Aghdami: Interesting. And you've touched a couple of times already on sort of the human element, which I'm looking forward to getting to in a sec. But before we get there, if we just take a step back and look at the whole picture, maybe starting with the opportunity. What is the big opportunity here? It's more than just operational efficiency. What is the opportunity with the use of AI for the investment industry, looking at it more from an investment lens, perhaps?

Arnaud de Servigny: Right. In fact, let's face it, we are in an industry where it has been recognized that fees need to be related to performance, to alpha. If there is no performance, fees should be low, and that is what passive investing does. You have got the ETFs; they are for a couple of basis points. If you have fees that are higher, that means that there is a benefit for the client. And this has been quite challenging because fees were there, but performance has been mixed—at least not for everybody, but for some of the asset managers.

What happens with AI is the capability to enhance the performance of the active management side of the business on two grounds. One ground is this: for a long period of time, modern finance was modern, but modern finance has been built when television was black and white, a long time ago, a long time ago, and a lot of people have been doing the same thing using the same techniques. What happens with, now, AI? It reshuffles the cards, in the sense that you can use different techniques to do different things. It opens up the landscape of opportunities. And this is something positive, because that means that people are helped, supported by machine, can think differently in order to retrieve performance differently. And as a result of that, that means that the sources of alpha that can be tapped are diversified.

In addition to that, because you've got machines or agents or, you know, support by AI, you can, in fact, not only work on one approach to the market, but on several approaches to the market. And as a result of that, something interesting is that you may have the support of machines working under your control with you that deliver a little bit of alpha on their own, but because you've got several of them, by combining them together, you can create a decent alpha source, and that's something new. You don't need to have 10 people. You need to have one person that is supervising different strategy, different machines to, say, different agents in order to bring that together.

Daniel Aghdami: Like a multi-strat kind of approach.

Arnaud de Servigny: Like a multi-strat in the hedge fund space, exactly. So that is something interesting. There is another element to it, which is the change in which we are. Typically, in traditional active asset management, the work was related to fundamental analysis. You dissect the firm, try to understand everything it does, whether what it says here and there will have an impact on its balance sheet and cash flow, but the reality is that a stock is not a firm. The price of a stock combines two things: future cash flows and discount factor. Future cash flows correspond to the fundamentals. The discount factor corresponds to the appetite of investors for the firm. So it's really—the stock is really when investment meet investors.

Daniel Aghdami: So you're saying that all available intel or knowledge is priced into the price anyway.

Arnaud de Servigny: What I'm saying is this: in order to invest, you need to understand the dynamics of the price, and you need to understand the two elements. One is related to fundamentals, the cash flows, and one is related to the appetite of investors. And what is interesting is that historically the job of an asset manager was to focus on investment, the fundamentals, and ignoring part of it, of the equation, which was the appetite of investors.

Daniel Aghdami: Right.

Arnaud de Servigny: Now, with the new market, we need to focus on the two. And, in fact, AI will help, really, to understand the dynamic of preferences of investors, which is—there is an element here, which has been known: it's the behavioral part of the business, understanding the evolution, understanding—you know, we have been talking a lot about the Mag 7, things like that, all the quality growth stock, I mean, but it varies over time, trends change, and that is to be understood by the portfolio managers and the machines can help also on that, and not only just on looking at the at the fundamentals of the firm.

So, in a way, getting back to your point, and your question, the new techniques enable to diversify, to leverage by having more people or more agents working for you and, at the same time, broaden the spectrum of what you're looking at to be able to capture not only one part of the firms, which is how they do themselves, but understand also how the rest of the market of investors are going to look at this firm.

Daniel Aghdami: Wow, fascinating. I mean, that’s the exciting side of it, right? But there's also, I think, some growing concern over, you know, what impacts AI and machines may have on the dynamics of the market. It can go both ways, right? And if I think back to previous, you know, some massive sell-offs, even more recently, right? Some of them have been driven by algorithmic models that sort of started hitting certain triggers and then, you know, a huge sell-off. Is there not that kind of risk as well, that there could be similarities between the models, and they just start driving each other further and further down?

Arnaud de Servigny: Look, suppose that I'm going, “That doesn't exist,” maybe that in a week we have a major crisis and things like that. So this is a dangerous question.

Daniel Aghdami: I'm not going to hold you accountable. Don't worry.

Arnaud de Servigny: This is a dangerous question. What I would say is that there is a no and a yes to your question or your observation. A no, I would say, is people invest on different horizons, some are very short-term driven, some are mid-term driven, some are long-term driven and, in fact, they're not looking for the same thing, and the way they operate is not just using the same recipes. So the more diversity you have, the better. Now, that was the no answer.

The yes answer is this: it is that when you look under the bonnet at AI, look at the equations. How does an LLM work? An LLM is based on a mathematical tool called transformers. Transformers are being built on the basis of a mechanism called attention mechanism, which has been developed in 2017 by the Google team. And, in fact, the principle of this is to have a measure of similarity. So, in fact, to be relevant, you need to learn from a large universe of books, references, and things like that, from which you draw, and when you come with something new, you're going to deliver what you deliver with some similarities to what you've learnt.

So the principle of similarities is that you're not going to differ that much from what you've learnt. And when I introduce into an LLM, I won't name any in particular, and I come with an odd way to express something new, they will not be comfortable with it. If I ask them, “Please give me an idea on this or that topic,” and then I feed it back, their result to the LLM, the LLM will say it is excellent, because it's very close to what it has been delivering. So that brings an interesting thing about complementarity between people and machine. The machine is going to be very effective at bringing something high-quality, but that is a result of mimicking something that already exists.

That means that if we think that the human person is doing the same job, it's going to lose, but if the person is coming with new ideas, new way to implement innovation, disruptive type of element, that will bring something that is orthogonal to what the machine can bring.

So it's really, it's really about building something where there is a space for the machine and a space for human beings. And it is, in fact, an interesting period where it becomes valuable to have older people participate in the game, provided that they are open to innovation and to disruptive thinking.

Daniel Aghdami: Interesting, then. So it was basically, I guess, bias, to an extent. And for that bias to be challenged or eradicated, you need humans who have judgment.

Arnaud de Servigny: Yes.

Daniel Aghdami: OK.

Arnaud de Servigny: Yes. And that means to, you know, go back to your question, if we come to a world that is completely driven by machines, then it could converge to something we're not comfortable with. If we're able to have the right balance between people and machine, that means that probably we can find some sort of equilibrium that is reasonable.

Daniel Aghdami: Do you think we'll keep that balance?

Arnaud de Servigny: I think that if there is a future for active investing, it can't be just disguised passive investing. So new ideas will come from people who think differently.

Daniel Aghdami: So, actually, I mean, by the sound of things, this is a really big opportunity for the active investing space that has been suffering quite a lot over the last, say, 5 to 10 years, right

Arnaud de Servigny: Yes.

Daniel Aghdami: So this is a huge opportunity for them. They should be regarding it as an opportunity rather than a threat to their business models.

Arnaud de Servigny: I completely agree with you. In addition to that, there is this wave coming from the US, mainly for tax reasons, which is the actively managed ETFs. In the US, it's largely related to a tax treatment that is favorable to ETF versus phones, but that means that that gives some sort of new momentum to delivering things a bit differently and taking advantage of these new vehicles to do things there a bit differently too.

Daniel Aghdami: Interesting. And if we go back to what we were just saying a moment ago about judgment and the important role of human in this dance, in this setup. If we're looking at how the boards and leadership of asset managers, family offices, hedge funds may be thinking about how their leadership team should be looking going forward to embrace this change and make the most out of it, what do you think are the key things they should be bearing in mind in the formation of those leadership teams or renewal of those leadership teams? What are the key skills that we should be looking for in those leaders?

Arnaud de Servigny: I think, Daniel, this is a very good point. When I look, I can't say I know all the boards, at all, far from that, but what I've seen over time is that, given the regulatory complexity, you've got quite a lot of people who have specialized within boards, making sure that there is compliance from a regulatory perspective. And it's not something easy because there is client money at stake and there are significant responsibilities.

I feel that, in board, there are people who have the knowledge, the experience of managing money and being able to assess the quality of the investment teams, but what is going to be difficult to find is people who understand that there is a needed reform in the investment processes, and the productivity there is—the productivity gains are going to be quite substantial. So what is needed is perhaps not people who have done that because there are very few, which are very few people who have been in that space, but people with the judgment to be able to understand what is at stake and assessing whether this is getting to the right direction or not, is going to be important.

Daniel Aghdami: Is that mainly at board level or would you say the same applies also to, say, the CEO and CIO, chief investment officer?

Arnaud de Servigny: What we've said is, and what I've said, sorry, is that the difference between the machine and the person, in that perspective—the person are bringing value, provided that they have got some sort of not disruptive thinking, but some sort of out-of-the-box thinking, OK?

Daniel Aghdami: Yes.

Arnaud de Servigny: And the question is: how do you create this out-of-the-box thinking? It can be that board member challenged the CEO; it can be that the CEO is coming with innovative views and is presenting that to the board, but that conversation here between the board and the CEO is exactly the place where these challenges should be handled. Because, in fact, when you think of people underneath, the CIO and other people, they need to have the operation work, and it's sufficiently complex so that it's difficult to handle everything at once.

Daniel Aghdami: Yes.

Arnaud de Servigny: But people who've got a longer-horizon perspective should be able to help and guide the process with some sort of interactive discussion.

Daniel Aghdami: And if we focus for a second on the on the chief investment officer, where today, of course, we look for long-standing, very deep financial market expertise and experience. Do you think we might find ourselves in a world, sooner or later, where the CIO no longer brings that expertise, but actually is coming from more of a data-driven or technical background?

Arnaud de Servigny: So I've been a CIO for many years.

Daniel Aghdami: You've evolved.

Arnaud de Servigny: Yes, I've evolved. There is now more than 10 years, but what I would say is, so we've moved away from star portfolio managers, in a way that doesn't mean that they don't exist, but if they're really truly star portfolio managers, they should work in a hedge fund. So that means that what is at the heart of a robust asset management company are the investment processes. The CIO, in my view, is the gatekeeper and the promoter of a variety of investment processes, working with his portfolio managers to make sure that they are state-of-the-art and that they are effective, drawing from the right data, and to be able to create performance for clients.

Now, there is a difference between some asset class space. Typically, when you're dealing with equities, there is more diversity at play because there are more instruments than on the fixed-income side, where you've got basically rates or yield curves and things like that. So, you can see that if you are more in a fixed income shop, there is still this bottom-up—this top-down approach, the view of the market, the understanding of what the central banks are going to do, etc., that is really playing a role. When you are more on with an asset manager that is working from a diversified pool like equities, then the diversity of robust investment process makes a difference.

What I would say is that, with this trend of the usage of AI techniques, is that the big, big revolution to come, in my view, it is that a firm doesn't anymore rely on a few very talented people, but relies on a large diversity of modestly talented investment processes, agents, and people, bringing them together to create superior performance and alpha for clients. So the activity of bundling sources of management and applying the judgment upon them is going to make a difference.

Daniel Aghdami: That makes so much sense. And if we look at the investment leader roles in those organizations today, and I'm sure there will be some investment leaders listening to this today, thinking, OK, how safe is my job going forward? I mean, you mentioned earlier judgment is a key thing, right? That's something that these people bring to the table. We've touched on this, you know, in the past—I know we've been catching up about how important expertise, experience in the markets is, to be able to have that judgment to oversee these kind of models. But if we're looking specifically at the chief investment officer role, the CEO roles, what advice would you give those people to kind of future-proof themselves within this industry going forward?

Arnaud de Servigny: My advice to them is, first of all, don't be worried. Secondly, the challenge is not—if you are open to innovation, you shouldn't be in trouble. But strategically, what you have to do is to work with young people and bring them working with you. The challenge is more going to be with young people who will find—will get it more difficult to find their place in the organization. And you need to have behind you people who are not just users of techniques, but develop their own judgments.

And so you are the person who has the judgment and the experience. Make sure that you don't work on your own and that you're surrounded by a few talented junior people who are over time going to learn from you, going to bring you what you perhaps miss in terms of understanding of the new techniques, and to bring them over time to a senior role. And if you're able to do that, it's a win-win situation.

Daniel Aghdami: I love that. I think that's probably one of the biggest concerns for most firms at the moment, right, is being able to future-proof their business and get people up to that level of developing sound judgment, even though they don't have as much opportunity to do so anymore as they once did. So it's a give-and-take, basically, right?

Arnaud de Servigny: Exactly.

Daniel Aghdami: You're saying, “Right now, I, as a senior person, can provide the judgment thing. I know that you, as a younger person entering the market in our business is going to be bringing more nimbleness and fluency in new technologies, and let me then give something back to make sure we support—”

Arnaud de Servigny: I think we should almost try to think, in terms of organization in these firms, to try to have some sort of, you know, one senior guy with one junior guy, to have this type of work taking place and to make sure that each brings to the other something.

Daniel Aghdami: It’s like a two-way mentorship.

Arnaud de Servigny: Yes, exactly, exactly.

Daniel Aghdami: I like that. And yes, lastly, if we focus on the future, what's the thing you're most optimistic about, and what's the thing that concerns you most about this rise of AI?

Arnaud de Servigny: From an optimistic perspective, we are in the middle of something that is just beginning. OK, it's not something that is behind us, all these techniques, we've had some waves. For instance, one of the waves that we had is given what I know, what can I say? We then moved to, given the history of what I know, taking into account how the history evolved, what can I say for the future? We've had given all what I know not only among the numbers but among what is being said or visualized on the firms and things like that, what can I say? We've had—so, in fact, it is a space where we're learning to think about things differently on a year-by-year basis.

So what makes me positive is that we're not in a situation like we've been for many years, where there was an established standard and good practices that were shared by everybody. So that is good because that means that there is diversity, there are different opinions, different views, different ways to manage money, etc., etc., which is sound for the market.

What worries me is the fact that, and I think that what people who've been in the business for many years will say is, “Gosh, this business is getting further and further away from concrete value creation.” There are a lot of young people who enter into this business having no idea on how a firm operates, and it's true that the market has become technical and therefore requires some effort to adapt to it, but if we do not understand what the reality of the firm is, it is a weakness.

Because, in a way, a market has proven effective in terms of optimal allocation of capital, but it's optimal allocation of capital to deliver growth, to deliver in the real world, and getting, you know, making sure that we keep a contact with the real world is important. So after having talked about all the things that happen with enthusiasm, making sure that we keep an eye on the firm as it is, with its fundamentals and things like that, it remains something important.

Daniel Aghdami: Super. So there will still be a place for leaders in the future in these asset management firms. 

Arnaud de Servigny: Yes. 

Daniel Aghdami: And the asset management firms themselves. 

Arnaud de Servigny: Yes. 

Daniel Aghdami: Wonderful. Arnaud, thank you so much for joining today and sharing your insights. I know that many of our clients will be fascinated to hear them. There's so many questions that come up about this topic, so thank you. 

Arnaud de Servigny: Pleasure. Thank you.

Thanks for listening to The Heidrick & Struggles Leadership Podcast. To make sure you don't miss the next conversation, please subscribe to our channel on your preferred podcast app, and if you're listening via LinkedIn or YouTube, why not share this with your connections? Until next time.

About the interviewer

Daniel Aghdami (daghdami@heidrick.com) is a partner and leads the Family Capital Practice in Europe and Africa; he is based in the Zurich office.