00:00:00 Introduction: The reality of consulting experience.
00:05:41 Isolation felt in top management roles.
00:07:15 Difficulties attracting talent due to brand issues.
00:10:44 AI threatens consultants’ perceived expertise.
00:13:13 AI won’t displace most consultants.
00:17:35 AI mimics expert insights without real expertise.
00:20:03 Limits of AI in personal interactions.
00:21:00 AI challenges the misconception of consultant expertise.
00:22:40 Critique on Gen AI in supply chains.
00:29:49 Software issues dominate supply chain problems.
00:33:27 AI expedites market research processes.
00:38:02 Consultants often provide moral rather than expertise.
00:42:43 Demands of expertise translate into complex work.
00:44:11 Expertise facade exposed by AI advancements.
00:49:15 Consultancy future stable despite AI changes.
00:50:52 AI-driven automation alters supply chain dynamics.
00:57:11 Automating routine tasks poses existential risk.
Summary
In the world of consulting, Joannes Vermorel from Lokad exposes the superficiality behind the industry’s reliance on perceived expertise rather than substantive knowledge. Vermorel critiques the entrenched idea that consultants offer unmatched insight, suggesting that top executives seek validation over real expertise. He argues that AI, while not ready to supplant consultants entirely, challenges the myths around their expertise. With AI’s rise, especially models like GPT-4, the consulting role’s reliance on symbolic prestige and psychological support rather than genuine insight becomes evident. This shift hints at AI-driven transformation, necessitating adaptability and re-skilling for future corporate landscapes.
Extended Summary
In the storied and often paradoxical world of consulting, the intersection with artificial intelligence (AI) offers both revelations and provocations, posing questions fundamental to the industry’s existence. As Joannes Vermorel, CEO and Founder of Lokad, converses with Conor Doherty of LokadTV, a profound observation emerges: many consultants seem to thrive not on the depth of expertise, but rather on a peculiar cocktail of perception and presence. Vermorel reconstructs the consulting field as an arena where young university graduates are often propelled into roles adorned with an illusory mantle of expertise. This approach unearths a contrarian view that questions the sincerity behind the traditional mission of consulting: the pursuit of external knowledge and fresh perspectives.
The official narrative advocates for hiring consultants to tap into reservoirs of expertise inaccessible within the confines of a company. Yet, Vermorel aptly challenges this notion, proposing instead an underlying dynamic where top executives find solace and validation in external consultants—companions in traversing corporate landscapes rather than paragons of knowledge. In era-long traditions, the desire for prestige channels companies to recruit consultants, particularly from esteemed institutions, not entirely for their actual skills but for the allure surrounding them.
In a twist of fate or perhaps logic, the rise of AI scrutinizes these established practices. AI, as Vermorel argues, may not immediately replace consultants, yet it challenges the mythos surrounding their expertise. It casts light upon the rare cases where genuine expertise resides outside large firms, often with independent consultants who possess formidable knowledge. This modern technological stride discloses the hyper-reality wherein the supposed value of consulting hinges more on visibility and validation than substantive expertise.
With the advent of powerful AI models like GPT-4, a noticeable paradigm shift unfolds in the perception of artificial intelligence as more than mere automation. AI punctures the consultant’s fortress of quantitative analyses, lending itself to symbolic math and coding, areas once believed beyond the reach of text-based AI. Vermorel contends that if specialized knowledge was truly at the heart of consultancy practices, AI would conceivably replace these roles, revealing their superficialities rather than upending their functional market positions.
This discourse evolves to address concerns over AI’s integration into supply chains, where Vermorel paints a picture of a more demanding but enlightening future. Rather than simplifying tasks, AI raises the bar for executives by pushing the boundaries of technical literacy and involvement. It’s not a matter of easing their workload but redefining the necessity for ongoing skill acquisition, which AI tools paradoxically accentuate.
Thus unfolds an intricate narrative: consultants often act as intermediaries, buffering executive decisions—a role not necessarily steeped in technical prowess but rather interfaced through psychological support. Vermorel underscores this understanding by portraying consultants as fuses within the corporate circuit, absorbing volatility rather than disseminating intricate expertise, illustrating that the consultant’s role frequently transcends technical or academic brilliance.
Looking to the horizon, Vermorel forewarns of AI’s impending transformation, likening it to a mass extinction event for routine white-collar roles. While elite consulting retains its prestige and complexity, AI-mediated automation promises sweeping changes across back-office operations, hinting at a new era where adaptability and re-skilling become indispensable. Vermorel urges consideration of alternate paths and manual skills, as AI simplifies coding and technological fluency, heralding an epoch of incessant industrial metamorphosis.
Full Transcript
Conor Doherty: So Joannes, thank you for joining me. The topic we were going to talk about today is AI and its influence on consulting firms. But before we get into what I’m sure will be another customary hot take, I’d like to take just a step back for a moment and preface the conversation by asking you a question. I am absolutely certain everyone who’s ever met you or watched anything with you has thought: where does the contrarianism come from? Because, I mean, I like it, you know that I work here. I’m the same way, I recognize this, but for anyone else who doesn’t understand or has never met you, where does this rebelliousness or pugnacity, whatever term you like, come from?
Joannes Vermorel: Uh, I was probably born with it. But the reality is, when I was a student at the École Normale Supérieure decades ago, I started doing consulting missions to make extra money. There were several things that were very surprising to me. First, it was relatively straightforward to get consulting missions. Why was it surprising? Well, because I knew nothing but math. I was incredibly ignorant on pretty much anything that wasn’t mathematics, algorithmics, or computer science, which was my passion for a long time. But when it came to business, I was very ignorant, and the interesting thing was, it was never an obstacle to getting a consulting mission.
I quickly got involved in bigger initiatives with big names in consulting, and realized that most of their teams were just people like me, two years older, fresh out of university, with zero expertise or experience. They were smart, driven, nice, polite, certainly educated. No problem. But once you’ve done something yourself, you see what goes behind the curtain. My first experience with consulting was just smoke and mirrors, and I was part of the charade. That gave me this contrarian perspective: if I could do it at 21 with 0.01% of what I know now, is it really something that is so challenging? Should you be paid that much? As a student, I enjoyed consulting, it was very well paid, shockingly so considering my expertise at the time. But now, with more maturity, I wonder if it’s money well invested by companies.
Conor Doherty: Well, on that point, because I don’t want to segue straight into what will doubtless be a critique—not necessarily criticism—but a critique. Before we do that, could you take as much time as you need to steelman what you see as the value proposition for consultancy firms now? So, we’re talking 2025, pre-mainstream AI. This is April 17th when we’re recording. What is the unbiased steelman version of the value proposition for consultancy firms today?
Joannes Vermorel: There is the official one and the unofficial one. The official line is, “We need this pool of expertise that doesn’t exist internally. We need the best of the best and we go outside to bring in the experts.” That’s, in my experience, the public reason given for almost all consulting missions. I believe that is rarely the real reason. I’m not even sure if 1% of the actual missions are truly justified by this. But this reason sounds good, plausible, acceptable, and that’s what gets written and advertised.
Now, the unofficial reason: in many large companies, top management often feels very alone. If you are a top executive, the game is very political. The people above you can fire you for any reason. You have full precarity in the job, it is tough and extremely competitive. The amount of work, effort, and sacrifices to land a high position in a large company is enormous. The people under you are not your friends, they might be reports, but you also might be in charge of firing them. So, who is really in your team? You might feel very alone.
Moreover, the problem can be compounded if the company doesn’t have an appealing employer brand. If a large company, say a multi-billion dollar company doing something traditional, has been around for a century and doesn’t have a good employer brand, super talented young people might not flock to you. If the company needs a VP of something, there will be plenty willing to do that. But if it’s to be below, more like an executant, not an executive, it becomes harder to fill those positions with top-tier talent from Harvard, MIT, or European equivalents, especially if you need something out of ETH Zurich or places like that.
So you can reach out to consultants and pay them directly. Good consultants are friendly, supportive. Their employers’ brand, like McKinsey, is quite good, attracting young talented people.
Conor Doherty: Well, that’s the first time you’ve said ’talented’. That’s the first time you’ve even acknowledged the concept of expertise, skills, talent, know-how.
Joannes Vermorel: I say talent, but if you hire right out of university, as is the case for the vast majority of consulting firms, at 22 you don’t have expertise. I was passionate about software, had some expertise in computer science, algorithms, some software engineering, although never managed any team at the time. That was very thin, paper thin. Most students, even talented ones without a specific passion, come out with almost zero expertise. They might have raw talent, intelligence, but genuine expertise is very thin.
However, elite consulting groups do have attractive employer brands and succeed in attracting talented people, even if expertise is super thin when they start.
Conor Doherty: Alright, well then, to get more towards the crux of the discussion, do you contend that the consulting firm position doesn’t work well or is broken due to AI developments?
Joannes Vermorel: Developments, so I think AI, in a way, is putting a spotlight under this pretense of expertise.
If consultants were truly hired for expertise, then my position would be that they are going to be obliterated by AI tools. But as I said, it’s only the official reason. This is not the unofficial reason. So yes, the official reason is expertise. My take is that it’s extremely rarely the case.
There are a few exceptional situations where you bring in someone who has truly a unique skill. By the way, those consultants with expertise are almost always on their own. They don’t work for a firm like McKinsey. I met, for example, a person more than a decade ago who had developed incredible expertise in manufacturing over 30 years, focusing on the proper setup and calibration of very complicated machine tools.
This person could, in a few hours in specific types of factories, halve the number of defects thanks to a lifelong experience in calibration of complicated machine tools imported from Italy. Just to give you a picture, this is someone who can come to your factory, charge half a million for four hours of work, and reduce the amount of defects from one out of 100,000 to one in a million.
This person had, in the distant past, worked for a big name firm, whether it was McKinsey or Bain, but as soon as people started to recognize his expertise, he was on his own. If you have this sort of completely unique expertise, you don’t need to be under a brand; you’re better off on your own.
That case of expertise is rare. If you accept the idea that consultants mostly don’t bring expertise, then why should AI pose any danger to consultants? Because AI will bring almost 100% expertise. True AI, at least as it stands, isn’t going to be your comforting buddy, patting you on the back.
My take is that AI is a complete non-threat to most consultants. It clarifies that what executives hiring those consultants were looking for was not expertise. It makes the field a little bit more transparent.
Conor Doherty: The field of AI has been around for decades. Your line of reasoning really took a turn around mid-2023. You had earlier mentioned a change by the end of 2022. We did two conversations, one after I joined Lokad around late November or December 2022, and then we revisited generative AI and supply chain in spring 2023. Your attitude pivoted between those recordings.
Infamously, you had originally referred to LLMs as being basically like a cat and were somewhat skeptical.
Joannes Vermorel: I was stealing this line from Yann LeCun.
Conor Doherty: Yet by March 2023, you had updated your stance to seeing it as a game-changer. We’re now in spring 2025, so what was it about spring 2023 that was such an aha moment for you?
Joannes Vermorel: At the time, the LLMs available were GPT-3.5, which by present-day standards is pure crap. If you didn’t invest several days in improving your prompting skills, what you got out of the LLM was pure garbage. It took hours to get something valuable from the tools.
Where I really pivoted was with the release of GPT-4. Suddenly, the results were valuable. I understood what prompts worked and realized that if used carefully, GPT-3.5 could already yield very interesting results.
It took me time to understand how these tools had to be used with their limitations. Most limitations no longer exist. The field has progressed enormously over the last two years. It’s a new form of intelligence or sub-intelligence. Even if it’s not yet a general intelligence, it is incredibly useful.
So if we put the debate on: is it intelligent or not—we say, okay, doesn’t matter. Yeah, it’s not even relevant to this conversation. There are entire classes of stuff that it does incredibly well. And it is just so massively useful. And yet, maybe it is just a stochastic parrot. Doesn’t matter. In fact, the fact that people say, “Oh it merely copies what an expert said online,” and I say, “Well, that’s what you’re paying for. That’s what you want fundamentally.”
I mean, there are plenty of situations where if you just parrot what a reasonably competent expert said on the subject, that’s already pretty good. That’s already pretty good. And that’s so incredibly useful. And free, depending on your access point. Free, yeah. I mean it’s essentially free even if you pay—even if you go crazy and pay $20 or a few tens of dollars a month—from a corporate perspective, this is pennies.
Conor Doherty: You mentioned the phrase ‘stochastic parrot.’ What it repeats is very impressive and this has only gotten truer as time has passed. If consultants listen to this, they might argue that their work involves numeric and quantitative analysis beyond text-based analysis.
Joannes Vermorel: I would say, first, LLMs are very good at symbolic math. So they are very good at composing a formula where they take ten factors and give you a formula that you could cut and paste in Excel, for example. They are excellent at doing that. I mean, their coding skills are pretty high. And let’s be realistic—the sort of numerical recipe that a consultant, even one that comes from an elite group, will conjure is on the very low end of what LLMs can do. LLMs can do nowadays a lot more in terms of coding challenges than what the sort of coding recipe that consultants used to deliver.
So the LLM is perfectly capable of giving you the formula, and then you cut and paste in Excel or compose a Python script and it will compute whatever you want. So I would say this argument doesn’t really hold. When when they say, “Oh the LLM doesn’t do things.” Yes, I mean technically, for example, an LLM cannot interview your colleagues. Yes, that is true. That is true.
But again, what it can do is very well compose the interview plan. So now the consultant will say, “Well, what I can do is conduct the interview with the interview plan that ChatGPT gave me.” But then we can argue—is it really expertise that you’re bringing to the table? If the interview plan was composed by the LLM, the mission and the way it unfolds is composed by the LLM, and you just happen to be the person executing the thing in the meet space where you can meet with people and do the thing.
But again, my take is that it would be a mistake to think that what consultants bring is expertise and thus they have been made obsolete by LLMs. Again, the mistake is that if you accept that what consultants bring is not expertise, then the fact that LLMs bring to the table tons of expertise is a little bit irrelevant.
Conor Doherty: Just to note, you’re a man of philosophy, so two things can be true simultaneously. LLMs can bring expertise, and consultants might still bring expertise, but it becomes redundant or obsolete due to the level of expertise available from LLMs. You’ve made the claim that you’re not getting expertise.
Joannes Vermorel: If you consider the type of expertise you would get, such as training sessions from elite consulting groups, they typically start with a few weeks where you’re trained to produce memos, PowerPoints, and the like. However, if you’re aiming to replicate the form of deliverables, LLMs are incredibly capable.
Conor Doherty: This leads to the next point, but I want to tie together a comment you made earlier about elite groups. Let me stitch these two together because when talking about elite groups in this space, we’re discussing AI. You’ve written some rather ascerbic, forthright, and wholesome reviews about the work of Yan Lun and Harvard Business School when it comes to applying generative AI in supply chains. What do you see as the point of contention between your understanding and everyone else’s? You have a bit of a contrarian take here.
Joannes Vermorel: The reason I mention those elite consulting groups is to address the counterargument that you’re dealing with incompetent consultants who don’t matter. Let’s say the baseline isn’t my own consulting missions when I was younger. I’m talking about elite consulting groups, the best representation of consultants. I have a divergence in view.
First we had a very specific point for this article on the Harvard Business Review, where fundamentally the claim that was being made implicitly in this article was: LLMs can autonomously write very complex and very extensive pieces of software. And I don’t think the authors even realized what they were saying. And when I say autonomously, because they were making the claim that this thing could be happening under the supervision of a person without technical expertise.
And so for me, that was a problem where, okay, I fundamentally disagree. I am—even if now I am, I think, a very proficient user of LLMs, including the latest one of OpenAI and their peers—no, you do not get that. Even considering state-of-the-art LLMs, you do not get that. And you certainly, as a result—as a deliverable—you certainly do not get that if the person babysitting the LLM doesn’t have tons of expertise in software engineering.
So that was a specific point where—I think the people who were writing that were themselves not very well-versed into software engineering. And so they were not even realizing that the claims they were making was essentially: we can have software that writes itself in 2024. And no, we’re not there yet. We’re not there yet. So that was something. Another one was—I would say—so that was specifically for this article.
Conor Doherty: Just jumping in, you talked about taking an exemplar, not cherrypicking or nutpicking a weak example. You’re discussing a paper on how generative AI improves supply chain management, reviewed in December last year. MIT, McKenzie, Microsoft, and Harvard Business Review were involved. Even if contrarian, you’re still punching up.
Joannes Vermorel: Exactly, the point I’m making isn’t about incompetent consultants. Yes, plenty exist. I’m discussing the very best, as that’s what matters. Autonomously writing software remains science fiction, especially without competent supervision.
If you have a competent software engineer in the loop, it’s possible, called vibe coding, possible for over a year now. It differs greatly from having LLM supervising supply chain software without effort at mastering technicality. I insist—we aren’t close yet.
More generally, use cases by consultants tend to offer a very soft message to supply chain executives. Let’s discuss use cases for other fields. For supply chain directors, LLMs can serve as incredible teachers for learning technical skills. Use LLMs to teach yourself technical knowledge without books or courses.
Is that because when you look at large companies, very frequently when you look at supply chain problems, 90% of the supply chain problems are software problems. It turns out that many supply chain executives are not very competent or very well-versed in software technicalities. Well, the interesting thing is that LLMs provide an incredibly useful remedy to these gaps in their skills.
Conor Doherty: Not only that, but when you use it, I like the analogy of the teacher. So, for example, if you were to use deep research within a chatbot, essentially an AI agent, and give it a task like, “I want to learn about the paradigms of enterprise software development, how my ERP runs.” You can say, “Okay, cool. Go away, spend 30 minutes researching that by yourself, computer. I have to sit in a meeting.” I like that you can parallelize tasks. From a productivity perspective, that’s how I’d sell it.
Joannes Vermorel: But you see, where I say it is a tough sell, because here the sort of things that get pushed about generating AI by consultants are things where they say, “The executives, your life is going to be easier. This thing is just going to improve things for you.”
Yes, and where I would have a contrarian message. However, I’m not a consultant anymore, and I’m not trying to be the best buddy of those executives. The reality is that due to this availability of those tools, the bar has gone higher. Now you have even fewer excuses not to acquire this technical baggage that you should have. It is a tough message.
If I compare that to the message from the Harvard Business Review article, it was saying you can give high-level instructions. You know nothing about software, and the LLM will auto-compose the supply chain optimization logic end-to-end, holistically, for you.
My take is that no, it’s not going to happen. What you can do, however, is use the LLM so that it teaches you how to understand those technical elements. You will reach a point where you have sufficient technical expertise to drive the LLM to do tasks. But you see, it is a much more demanding exercise, as opposed to the idea that magic will make your job easier. On the contrary, I believe it will make your job overall more demanding. The LLM brings expertise, not your buddy. It’s a different mindset what works best.
Conor Doherty: You have also argued, not argued but discussed, in that very seat with Meinolf Sellmann, the value of AI agents. Not overstating the value but one of the points of value with AI or Agentic AI is the ability to do things like market research. So, let’s say you’re a supply chain executive. You need to shortlist companies for whatever it is you want. Okay. Well, I could hire a consultant for that, do it myself, give it to an intern, or turn to the open window on my laptop.
Joannes Vermorel: But you see, again, the LLM is making your job as a supply chain executive more demanding. Why?
Let’s consider the previous situation. Market research used to take weeks, and you would have to hire consultants. The whole process would be slow. You’d spend 90% of your effort talking to a consultant, briefing them, getting updates, challenging them, etc. It takes three weeks, and at the end of the three weeks, you get a 10-page report. Now you get it in 30 minutes and can repeat the process. This means you have more responsibility.
In one day, you can end up with 200 pages of super high-quality, super dense information that you have to ingest. That’s why I say the job is getting harder; LLMs make you realize that your bottleneck is how fast you, as an executive, can process all this information and make sense of it. If there are things you don’t understand, you’ll have to go back to the LLM, asking it to explain this and that again.
You don’t get the luxury of a slow process with consultants where you can have a nice lunch with them and take a few weeks for the whole task. The LLM takes the fun parts away from the job and pushes you directly to your screen to ingest hundreds of pages of carefully created documentation that answers all your questions.
Conor Doherty: Well, to get a bit concrete here, let’s say a company has a supply chain issue. Imagine now the company wants to re-evaluate: should we, as a company, switch suppliers? A textbook supply chain example. I’m not satisfied with the performance of my suppliers. I can hire a consultancy firm for that, or I can use an AI. Would you please explain how those two would differ, how you see them being different, and positively different in the case of the AI?
Joannes Vermorel: If you go with the AI, it will immediately ask you, “How do you diagnose that your supplier is not up to standard?” Do you have a supplier scorecard? If not, it will provide you with one. Do you have the relevant KPIs? Give the AI your KPIs, and it will review them and offer an improved list. Can you compute those KPIs automatically from your ERP? If not, give the AI a schema of your ERP, and it will provide you with the SQL queries.
You see, the whole thing will, at full speed, let you execute the thing. Now that is expertise, expertise, expertise, expertise. So you see, just piling up on the expertise.
Now, the reality might be, “I know already that we have to kick out this supplier because it’s bad.” The problem is that a lot of this supplier’s employees are ex-employees of this company, because it frequently happens. In companies, you have like the mothership, and you have the suppliers, and a lot of people go there. Okay, so this supplier is not good. The problem is, it has a lot of people who have a lot of ties. And a lot of people under my teams really appreciate this supplier. Yes, maybe they’re not good, but we have excellent relationships. A lot of people appreciate working with this specific company. So there’s a lot of entanglement in terms of interests. And this supplier is not very good, but they have been very loyal. So I am very much afraid of what it would imply for me—for my career—if I were to, you know, cut them out. What if the replacement supplier is not better? What if, etc.
And so there’s a lot of, I would say, fear. Let’s say you’re a supply chain executive. You kind of know what you have to do. You already know the numbers. Yes, you can have like 20 more KPIs, but fundamentally you’re already like 99% confident that this is correct. And now what you want is consultants that will support you. That will produce this aura of authority that says: “Look, I’m doing that, but it’s not just my decision. I brought experts. Experts agree. This is what we have to do. This is important for survival.”
But is it truly expertise you’re buying? You already know the conclusion, which will be immediately shared with the consultant, leading them to conclude what you want to conclude. Still, it’s that sort of support where they have your back and help you. The human touch. So you see how it compares. The expertise mission, where “Let’s give you the thing to do your diagnosis, where I will let you compose a scorecard, I will let you figure out analytically who are the true bad suppliers, if any” Versus: “I already know the conclusion. I already know what I want to do. I already know what should happen. But I feel extremely alone in doing that.” “I feel—and it’s a little bit terrifying, it is exhausting, and I just need backup that will be on my team to do it.” “And that will be the consultant.” And that’s—so there is very little expertise in the sense of this, you know, raw expertise skill here. But that’s exactly what the consultant can bring.
Conor Doherty: Actually, that ties back to a point that Eric Kimberling made. We spoke to him last week. You recall one of the points he made was that often when consultancy firms are brought in, the primary reason for bringing in a consultant isn’t necessarily needing access to their expertise. Instead, as a board or supply chain executive, you simply want an intermediary to deliver certain messages. For example, when we have to get rid of Joannes as a supplier, it’s not me saying that, but rather Connor from Mackenzie, whom I just hired. It’s a way to channel the feedback through someone else.
Joannes Vermorel: Exactly. A consultant can play the role of the fuse. They are willing, for a price, to play this game, which has value. Again, it is the unofficial value proposition, not related to expertise. Playing the role of the fuse doesn’t require super in-depth expertise.
Conor Doherty: Well, not the technical kind of expertise. It’s a different skill set.
Joannes Vermorel: Yes, but do we need someone who was in the top 10% of graduates from MIT to do that? Do we need top-tier talent for the optics? Yes, but for the actual execution, no. When I discuss this with consultants who studied sciences at university, they were doing complex calculations, yet as consultants, they often find themselves just computing percentages and creating PowerPoints. If genuine expertise was required, they’d be doing work more demanding than at university.
And by the way, this is the case—for example—software engineers at Lokad, they work on stuff, and usually the sort of software tidbits that they compose are way more challenging than whatever they ever did at university.
So if you are really in a job where it’s your expertise that matters—when you go and work—if it’s really expertise that is being purchased by your employer. Then, as a rule of thumb, what you do in your day job is vastly more demanding than anything you’ve ever done at university.
If this is the opposite—meaning that you were doing fancy equations at university, and then you end up doing PowerPoints at work—then most likely, we’re not talking of expertise. This is not what is being brought to the table. This is not where there is value-add. It’s something else. Important—but else.
Conor Doherty: Well, this actually comes back—or brings me back—to a point that you made earlier when you were talking about consultants softening the message for supply chain executives. And you just mentioned university students. And if I can tie that—university students—and a previous comment you made before about the mass extinction event that is coming or possibly has already arrived due to AI. I want to ask you—when you say AI, there’s a mass extinction event coming for white collars.
Joannes Vermorel: Yeah, back office—especially back office.
Conor Doherty: Back office white collars—including, obviously, in terms of expertise, consultants. And then you talk about, well, top 10%.
Joannes Vermorel: Again, consultants are not back office white collars.
Conor Doherty: But the expertise that you’re saying that they can provide is already gone.
Joannes Vermorel: Yes, but as I said, the expertise was never—it was only the pretense. It was never the real thing. So the pretense is kind of busted, but you know, the charade can continue. Again, the thing is that it looks silly to advertise: “I need to bring this elite consulting group because really, I feel lonely. I need more support.” “Yes, it’s going to cost 500k to the company, but I need it.”
Okay, that sounds silly like that. So I say, “I bring in the experts.” Okay, fine. But your other—your peers—you know, other executives, other VPs and whatnot—they know the game that is being played. So they are not fooled. You know, they know that consultants are being brought in not for their expertise. They are very much aware, because that’s what they did when they also used the very same consultants. So again, the fact that—I think the optics can remain. The optics can remain: “We are bringing the experts.” Yeah, fine. The optics will stay unchanged. The AI is just making it more obvious that it’s just a pure matter of optics. But that’s it. You know, fundamentally—because the value-add is completely different—it will continue.
Conor Doherty: When you say there’s going to be a mass extinction event affecting a lot of jobs, but people already know that, what do you want to achieve by drawing attention to it?
Joannes Vermorel: So first, let’s clarify a few things. First, consultants are clearly not a back office office job. I’m aware of that, because you are always in a position of selling. So a good consultant will do the mission, and while they are doing the mission, they sell the next mission. So for me, that’s the archetype of the front office white collar, which has a sales job. And they are—as they do their work—they do their sales job. So that’s the sort of thing where I believe that, again, AI will have really a hard time automating that. Because automating enterprise sales is very much, again, a human thing—to sell the mission. So that’s not really in danger of being automated.
Now, when I say this mass extinction, I’m talking about all the clerks. And large companies have thousands and thousands of people who are doing clerical tasks. So there is information coming in, possibly through a mailbox, and they have to do a series of steps and then they will forward a slightly modified version of this information to somebody else. And then do something. And when you have those white collar workers who are in fact blue collars—because what they are doing is that they are pretty much like automatons, you know—it’s just that it’s information that they process. The job is extremely repetitive. That’s where, I would say, extinction is coming.
You see, and the reason why I’m pointing out this—drawing the line on that—is that I believe markets are not really good educators. They’re filters. So what will most likely happen is that a majority of companies will do exactly nothing about that. They will not improve their process. They will not try to automate 90% of the jobs. And that’s fine. That’s what happened in the previous industrial revolutions.
What will happen is that some other of their competitors will do—and those companies who do not will just disappear. A small minority of companies will actually do the upgrade and survive these waves. But again, it will be a minority. So what I’m shedding light on is that—I think it is a pivotal moment where some companies can actually do this transition and be part of the next wave of companies. But all the companies who do not will end up with a bloated cost structure that is not competitive anymore compared to their peers who have done this transition. And their peers also include newer, younger companies who will just appear in this process, in this transition.
Conor Doherty: Winding down, considering recent technological strides in the last 20 to 24 months, projecting into 5 years, how do you see the future of supply chain consultancy?
Joannes Vermorel: Business of elite consulting groups will not be radically altered by AI. The key reason is that expertise is mostly irrelevant. The fact that there’s a technology that commoditizes expertise, like the LLMs, is not what is being provided. I believe that for elite consulting groups, AI will be largely inconsequential. They’ll adopt those tools just like everyone else, making PowerPoints, memos, and emails faster. But, it’s similar to when they adopted email 20 or 30 years ago. It’s just a tool that will become part of daily life for consultants, as it will for any office worker.
However, for supply chains and sales, the changes will be much more pronounced. We’re seeing this happening with our clients. For example, we have a small client that has completely automated their request process. They receive free-form requests for quotes via email for specialized equipment. They had multiple people monitoring this inbox to reformat inbound requests into properly codified requests for quotations, which would then be converted into PDFs through ERP systems. Now, this process has been completely automated using LLMs, significantly reducing the need for many employees.
Conor Doherty: When you say fully mechanized, how many moving parts are involved in that?
Joannes Vermorel: It’s just an LLM that takes emails in, potentially with attachments, such as spreadsheets or PDFs, and creates a standardized JSON output that matches ERP expectations. It’s automatically injected into the ERP system, generating an email templated with a PDF attached that represents the quote. While it’s just one function, it has a huge dollar impact.
They went from about nine full-time employees on this job to zero. It was pure back office work, and similar automation is happening across large companies, especially in supply chain operations, which tend to have a big bureaucratic core. In the next five years, many clerical back office jobs will be completely automated.
Conor Doherty: There’s some precedent to that. Recently, Shopify stated they expect all their employees to become AI literate. For my parents’ generation, things changed too, like managers at Procter & Gamble losing secretaries who used to type their letters.
Joannes Vermorel: Yes, those jobs disappeared with microcomputers. Now, personal secretaries are mostly for very high-level positions. It was a massive extinction event for certain jobs, and what’s happening now is bigger as it’s affecting all clerical tasks at once, rather than just a few positions.
Conor Doherty: If I summarize, the advice used to be “learn to code.” Are you saying now it should be “learn a manual skill” like plumbing or fixing engines?
Joannes Vermorel: It depends. LLMs are incredible teachers, making it easier to learn coding. However, if your job is repetitive, like a back-office white-collar role, it might be wise to consider a plan B as automation will affect those jobs. While front-office roles involving client interactions might remain unaffected, those operating mostly behind a computer will likely be automated. Some companies may delay automation, but ultimately, it could lead to their collapse if they don’t adapt.
Conor Doherty: I don’t have further questions, but I’d like to revisit this topic in two years and see how your predictions play out. Thank you very much for your time; it’s always a pleasure.
