00:00:00 Introduction to the debate
00:05:26 Joannes’s opening remarks
00:12:03 Milos’s opening remarks
00:18:56 Joannes’s rebuttal
00:24:17 Milos’s rebuttal
00:29:26 Joannes’s concluding remarks
00:31:27 Milos’s concluding remarks
00:33:25 Follow up questions
01:02:02 Open discussion between Milos and Joannes

About the guest

Milos Vrzic is a dynamic professional with over two decades of expertise in finance and supply chain management, spanning the watchmaking and pharmaceutical industries.

Throughout his career, Milos has been a catalyst for transformative growth and operational excellence in global organizations. His proficiency in implementing S&OP has been pivotal for companies ranging from startups to multinationals, successfully navigating complex business units and diverse product ranges.

In his most recent role at Galderma SA, Milos serves as the Head of Supply Chain for the EMEAC region. Here, he has led the implementation of S&OP processes, significantly enhancing operational efficiency and aligning practices with the company’s ambitious growth targets.

Prior to his tenure at Galderma, Milos spent a decade at Blancpain SA, a prestigious watchmaker and subsidiary of the Swatch Group. There, he implemented robust S&OP processes that tripled business volume, improved sales forecast accuracy by 20%, and implemented Master Scheduling. His efforts in S&OP significantly strengthened the link between supply chain and sales, enhancing demand reactivity and reducing the internal bullwhip effect.

Milos holds a Bachelor of Arts in Business Administration with a specialization in Marketing from Webster University Geneva and various professional certifications, including CPIM by APICS. He is currently pursuing his Lean Six Sigma Green Belt™ certification. Fluent in English, French, and Serbo-Croatian, Milos brings a multicultural perspective to his role, fostering collaboration and sustainable growth within the global supply chain sector.

When he’s not managing supply chains, Milos can be found driving his two boys to their extracurricular activities or perfecting his swing at the golf driving range.


In a debate hosted by Conor Doherty of Lokad, Milos Vrzic, former Head of Supply Chain at Galderma, and Joannes Vermorel, CEO of Lokad, discussed the value of S&OP for companies. Vermorel critiqued S&OP as simplistic and outdated, while Vrzic emphasized its role in tactical planning. The debate highlighted the complexities of S&OP and the need for a nuanced understanding of its role in business.

Extended Summary

In a spirited debate hosted by Conor Doherty, Head of Communication at Lokad and host of LokadTV, two industry experts grappled with the question: “Is S&OP a net good for companies?” The participants, Milos Vrzic, former Head of Supply Chain at Galderma, and Joannes Vermorel, CEO and founder of Lokad, brought their unique perspectives and experiences to the table, offering a rich and nuanced exploration of the topic.

Vrzic, with his background in finance and supply chain across various industries, brought a contrasting perspective to the debate, shaped by his financial background and his interest in Lokad’s contrarian opinions on various topics. Vermorel, on the other hand, brought his expertise in mathematics and computer science to bear on the issue, drawing on his experience in predictive optimization for supply chains in his role at Lokad.

The debate was structured to ensure fairness and symmetry, with Doherty outlining the format and defining the key terms: S&OP and net positive. This set the stage for a robust exchange of ideas, with each participant presenting their views, followed by follow-up questions and a free exchange between the participants.

Vermorel took a critical stance against S&OP, arguing that it is not a net positive for companies. He contended that the framework of S&OP is too simplistic and outdated, failing to consider the complexities of modern businesses and supply chains. He also criticized the reliance on people for information flow and problem resolution, arguing that modern computers should handle these tasks.

Vrzic, while agreeing with some of Vermorel’s criticisms of S&OP, emphasized the need to understand S&OP’s role in the corporate landscape. He explained that S&OP fits into the tactical arena of planning, focusing on the ‘how’ rather than the ‘why’ of strategy. He argued that S&OP covers a time horizon of a few months to 18-20 months and focuses on volume rather than product mix.

In conclusion, the debate offered a rich exploration of the role and value of S&OP in modern businesses, with each participant bringing their unique perspectives and experiences to bear on the issue. The debate underscored the complexities and nuances of the topic, highlighting the need for a nuanced understanding of S&OP’s role in the corporate landscape.

Full Transcript


Conor Doherty: Welcome to a very special episode of Lokad TV. Today, I have the pleasure of hosting a debate between Milos Vrzic, the former head of Supply Chain at Galderma, and Lokad founder Joannes Vermorel. Now, the topic of today’s debate is as simple as it is controversial: Is S&OP (Sales and Operations Planning) a net positive for companies? Arguing for this is Milos, arguing against is Joannes. But before we go any further, Milos, thank you very much for joining us and I invite you, please could you introduce yourself to the audience?

Milos Vrzic: Yes, thank you very much. So, I’m Milos Vrzic. I’ve been in two different disciplines, finance and supply chain, and in two different industries, watchmaking and pharma. So, I’ve seen pretty much the full spectrum of the end-to-end supply chain, everything from sourcing, planning, making to delivering. I’ve had that experience, but because I’m a little bit of a newcomer in supply chain because of my financial background, I have a very different perspective on what works or what doesn’t work. And what’s really interesting about this debate is that I’ve been following Lokad for quite a while now. It’s very entertaining to watch. I like the fact that there’s a contrarian opinion about many topics, some of them like ABC or stock, etc., or DDMRP to mention but a few. And my perspective is very different from the one that Joannes comes with. He comes from engineering, from Grand Ecole in France, where he has very much the perspective that is different from mine. And my perspective is, I was in the thick of it. I got into supply chain almost by accident and until I got my CPIM certification, I didn’t know what I was talking about. So that really led me to have a contrasting opinion. But overall, my V diagram and that of Joannes is quite overlapping and that’s what’s so interesting about this debate, our different backgrounds and our opinion on S&OP.

Conor Doherty: Again, good to have you and thank you very much. And Joannes, I mean, people are familiar with you, but for the sake of formal symmetry, could you please introduce yourself to the audience?

Joannes Vermorel: So, I was doing a PhD, dropped out of the PhD which was about computational biology, to enter the world of supply chain as a trained mathematician and computer scientist. And well, 15 years later, I am leading Lokad and what we do nowadays is predictive optimization. And so, we, on behalf of our clients, we pretty much on a daily basis optimize their daily operations. So, I came to supply chain relatively late in life as well, but still, I do preserve, as Milos completely got, that I preserved indeed this computer scientist vibe, I think, that permeates the way I approach the domain.

Conor Doherty: Well, gentlemen, thank you and I’m sure it’s going to be a very informative and entertaining debate. But before we get into that, there is a little bit of housekeeping. So, indulge me as I talk to the audience directly. So, some housekeeping: first, there will be opening remarks of a maximum 7 minutes duration. As per Milos’s request, Joannes will speak first, then Milos will take his opening remarks.

Following this, there will be a five-minute rebuttal from each speaker in the same order, and then there will be a two-minute concluding remark. Following that, I will do my best to, in an unbiased fashion, ask some follow-up questions to press both speakers on things that have been raised. And then, if we’re all still on good speaking terms following that, you guys can go at it and clarify any points that don’t fit into the constraints of a formal debate.

Second piece of housekeeping: to prevent this from devolving into the sort of annoying online debate that people typically find, where people just shout at each other and no one can agree on any terms, prior to this debate, I spoke with both of you and agreed upon two key terms to sort of frame or anchor the discussion. First of all, S&OP, and I will read the definition that was agreed upon and hopefully will display it as well. S&OP is a strategic monthly process that tries to balance supply and demand through interdepartmental collaboration. Its goal is a unified long-term forecast and execution plan. And when we say net positive, speakers have agreed that this means generates more value than it costs, quite simply.

And hopefully, if you need to add any more color to that, you can do so, but then it comes out of your speaking time. You both have timers in front of you. I will also keep track of the time. When you have about 30 seconds left, I’ll give you a gentle nudge. When your time is up, I will cut you off for symmetry and fairness. And with that, unless there are any questions from anyone, I think let us progress into the arena of battle. And I ask you both, is S&OP a net positive for companies? Joannes, please, your opening remarks.

Joannes’s opening remarks

Joannes Vermorel: So, is S&OP a net positive for companies? The short answer is a resounding no. The longer answer requires us to step back and assess what is at stake. There are at least three angles that need to be considered. First, the integral framework of S&OP itself. The framework is deceptive, it is defective, it is, by design, too low dimensional to be of genuine value for an actual company.

Second, S&OP assumes that the information flows from and to people, and this assumption ceased to be relevant decades ago. And third and last, S&OP relies on people doing the thinking of the resolution, which has also been an outdated perspective for a decade. So, stepping back, with the rise of the giant corporation in the 19th century, the challenge of the division of labor became more acute. Behind the so-called economies of scale, it’s usually a better division of labor that is at stake. However, as a downside of this approach, the giant corporation coordinating many people within a single organization becomes difficult. The left hand doesn’t know anymore what the right hand is doing. And S&OP emerged as one of the macro solutions to this coordination problem.

Yet, and this will be my first point of critique, S&OP comes with a two-dimensional take on the case: Demand versus Supply. The demand is being assessed through a statistical means, that is, forecasting, and the supply is adjusted by turning those forecasts into business commitments. I reject the idea that any business can successfully operate at scale on such a simplistic view of the market. I’ve met corporate bureaucrats who appear to think it is possible, but whenever I talk to entrepreneurs, CEOs, investors, I get much more nuanced views of all the concerns that are blending into the vision of the market and the vision they have of the company within the market.

On this ground, I also reject the S&OP variant of the supply chain triangle as advocated by Bramm Desmet, balancing cost, cash, and service. Going from a two-dimensional view to a three-dimensional one doesn’t materially change anything. The take remains simplistic, ignoring most of the specifics of the company. In essence, S&OP is a simplistic tool that appeals to the bureaucratic mindset of many corporate middle managers.

Secondly, S&OP implicitly but extensively assumes that the information must flow through people, hence its emphasis on numerous meetings that characterize this practice. Information through people as a necessity used to be the case up until the 1970s, but not anymore. Supply chains were digitalized decades ago and nowadays, the coordination problem is not only expressed but mediated through the applicative landscape of companies. Denying this is denying the operating reality of present-day supply chains.

Thus, whatever solution is envisioned for the coordination problem, it must treat the applicative landscape as a first-class citizen. And let’s immediately dispel the misconception that having some sort of S&OP software matters in the slightest in this regard with regard to the applicative landscape. My criticism is not about the medium used to convey the information, whether it is pigeon, spreadsheets, or video conferencing tools. My critique is that information flowing from employees to other employees is the problem.

This is an antiquated view that has no place in an age of digitalized supply chains. People can supervise the flow of information, but they should not be expected to be involved with the flow itself. And thirdly and lastly, S&OP assumes, quite explicitly, that the resolution of the coordination problem must be performed by people. I would like to fully disabuse the public of this notion. The resolution, almost all of it, should be done by the thinking machines of our age, that is to say, modern computers. To the objection that we don’t have artificial intelligence yet, I answer that we have had access to some artificial intelligence for the last 70 years.

Let’s ask a man in 1924, a century ago, about computing interest rates, for example. This man would say, without any doubt, that such a calculation, in order to be done correctly and swiftly, requires a man of great intelligence. The fact that mankind has been moving the goalpost on AI over the last 70 years doesn’t diminish in the slightest the achievements that have already been obtained. It doesn’t matter if some problems remain beyond artificial intelligence. The fact is that what Lokad has been doing for a decade is the living proof that the daily coordination problem is, by far and large, a problem that can be entirely mechanized, as far as supply chains are concerned.

And thus, in conclusion, is S&OP a net positive? No, it isn’t. It is not even close to being one. And let’s not be fooled by the argument that S&OP is better than nothing. By this standard, workhorses would still be considered as a net positive for supply chains because they are better than breaking our backs carrying goods. But businesses cannot afford antiquated baselines. The opportunity costs are immense and the coordination problem is critical and must be addressed. This is no question about that. But it must be addressed by leveraging the best paradigms and the best technologies that our age can offer. And S&OP is not even close to that.

Conor Doherty: Within limits as well. Joannes, thank you very much. I shall reserve further comment. Milos, I invite you to make your opening remarks and Joannes, recall that you will have an opportunity to rebut, so do make notes I guess. When you are ready, Milos.

Milos’s opening remarks

Milos Vrzic: Okay, so thanks for that, Joannes. I completely share a lot of the criticism that you face versus S&OP. My rebuttal is going to be a little bit professorial, so please forgive me for that, but I really need to situate S&OP in its correct landscape. So let me start first off by saying what is S&OP for? Where does it actually fit in any corporation? We have at least two massive layers of planning. One layer, the top layer, is the annual plan, the three-year plan, which looks at strategy. And one of the discussions that you talked about, effectively, strategy isn’t discussed in S&OP, something else is. Strategy is discussed definitely in the three-year plan.

And then at the very bottom level, we have another planning topic or process called Master scheduling. Now, Master scheduling is different from the Strategic plan, obviously, because in the master scheduling aspect, you really need to look at the bits and bobs. How exactly are we going to be making what product for which market, etc. And so, between that huge plan that covers a large spectrum of time, which is the three-year plan, and master scheduling, sits a gray zone, a massive gray zone. That gray zone is known as the Tactical Arena. What is the Tactical Arena? Well, we can look at Eliyahu Goldratt, who basically pointed out that tactics answers the question ‘how’. How we’re going to get something done, while strategy answers the question ‘why’ we’re going to do it.

So the very core question of what S&OP does, it’s a planning process for tactics only. So what does that mean? That implies two very distinct things. First of all, we’re not talking about a spectrum of time from now until the next few months. We’re talking from the next few months, in other words, outside the spectrum of the master scheduling, all the way up to an 18 to 20 month horizon. That’s the time horizon that we’re looking at. Point number one. Point number two, contrary to master scheduling, in the S&OP, we’re looking at volume. We’re not looking at the mix of products, whilst in master scheduling, we’re looking at the mix of products. So just to illustrate this point, I’ll use an example of a potato chip company. While we can have in the 70s maybe 50 brands of potato, today we have 10,000 brands of potato chips in the same company.

In the Tactical arena, in other words, in the S&OP process, we’re not going to be looking at how many bags of potato chips we’re going to be making. We’re going to be talking about in volumes of potatoes, so in terms of tons. So that hasn’t changed from the 70s, nor has it changed to date. It’s a volume business, not a mix business. And I think this is a very important consideration. So in order to understand also what, why do we need S&OP, why do we need a planning at that stage, we have to take a look at what kind of decisions are we going to be taking in the Tactical Spectrum. So the kind of decisions, and this is the table that I sent you, Conor, this is the point where you want to present it, basically illustrates the lead time and the sort of decisions that you’re going to have to make. The decisions would be something like acquisition of a new company, new product development, market launch, purchasing equipment, hiring people, etc. This can’t be done overnight and this is the kind of decision that we take.

So what are the characteristics of tactical decisions? The first characteristic is they have extremely long lead times, which means you need to take a decision relatively quickly if you want to see the results of that being implemented. The second point is that it’s a capital-intensive exercise and whenever something has a long lead time as and is capital intensive, that means that it’s going to have a large impact on the P&L. And this is a capital point because it takes me very neatly to the third element, which is who is responsible for the S&OP. Now, contrary to popular belief, it isn’t supply chain, it’s not sales, definitely not finance, it’s actually the CEO. Because the CEO is the person in charge of the P&L. So only the CEO can take these capital-intensive and long lead time decisions, such as the ones that I’ve illustrated in the graph. So this is, I think, a very key component of understanding what S&OP is and how does it cooperate in the Tactical Arena.

Now, here’s where you have to step back and not be an engineer and imagine that these decisions need to be taken by a CEO who’s not a supply chain professional. And so we have to enable the person to be able to pull the trigger and make these decisions. That’s not easy. So let’s do a little thought experiment. Imagine you walk into the office of the CEO and you’re like, “This is a check for 10 million, please. I need to buy a new warehouse.” What would be the probability of you succeeding in that? The answer is zero. It’s never going to happen. So we need some system that shows the Tactical landscape to the CEO regularly, relatively frequently, and I’ll get back to why it has to be frequent, so that the CEO or general manager can actually pull the trigger. So what does that mean? That means that we have to get into a mode of pattern recognition.

Pattern recognition means having monthly S&OPs, so the person continuously sees the exact same deck with the exact same information, except there’s going to be outliers. And in that one outlier, guess what, there’s the warehouse that we need and this is why we need it. And therefore, they’re going to be able to pull the trigger a lot faster. And by being able to pull the trigger a lot faster, that gives them a huge competitive advantage. And by the way, this is the reason why fast-moving consumer goods (FMCG) companies are the ones who are so keen on having a good S&OP. It’s because they can’t afford to lose any time in the Tactical space. The last point I would like to make is everybody talks about alignment in S&OP, but I don’t think that that is the right definition. And the definition that you given on S&OP does mention something called balancing supply and demand.

And again, here what needs to happen is there comes a point where the general manager is going to either trust the sales numbers. Let’s say in a certain scenario, sales are not doing that great for a certain product type. Your proverbial sales guy says, “No, we’re going to catch up in the next three quarters of the year.” Well, the general manager has to take that into account and say, “Yes, I’m going to bet that sales are going to be there” or, “No, they’re not going to be there.” So my closing remark is, you would be in a lot more trouble if we didn’t have an S&OP because none of these tactical decisions would be able to be taken in time and that would cost the company a lot of money.

Conor Doherty: Well, thank you, Milos. I will take that as a seeding the final seconds. At this point, Joannes, I invite you to respond to the comments that Milos just made. Five minutes, thank you.

Responses to the opening remarks of each other

Joannes Vermorel: On this presentation, I think the biggest problem is the sort of paradigm in which, you know, the sort of reasoning operates. Let’s start by, do we even need to have humans in the picture? And I know it’s a little bit extreme. When e-commerce started, you know, when e-commerce started, the reason why Walmart didn’t become Amazon was simply because for them, it was unthinkable to be able to sell things without the human experience. It was like unthinkable. So the idea that you could actually have the entire retail industry where there is the entire front-facing aspect of the business is just fully mechanized, was just unthinkable. So my take is, okay, let’s start by making a leap into the future, 20 years ago, where AI will have been progressed and running a supply chain can be done entirely with machines. I know it’s a little bit early, but let’s start from that.

So what sort of thinking is going on? What sort of engineering techniques? And when we start looking at a few things that have been presented, we can sort out by looking at that, what is truly the essential ingredient of what will take to actually engineer the supply chain of the future and what is just tradition that we specifically tailored for humans. And here, we have first, the slice and dice of the domain and there are a lot of things that really do not make sense.

For example, slicing and dicing with regard to the horizon, short term, mid term, long term, super long term, whatever. Here, when you have machines, the first thing is that there is no horizons. It’s always from now to infinity, all the time. So you can revisit every single horizon at any point of time, every minute if you want. Thus, it has this sort of depth where the only reason why we have the operational, the tactical, and the strategic, it’s just because humans can’t do everything all the time, so we need to slice and dice and say we are going to revisit some stuff more frequently and some stuff less frequently, but it’s fundamentally a human constraint.

And then what about that’s not the only slice and dicing that is happening. Why should we have a process that decides about the mix and another process that decides about the volume and another process that decides about the pricing and another process that decides about quality, etc. Those are literally slice and dice silos, thinking silos, paradigmatic silos that are only for the division of labor when done by a human. Again, when you have a machine, those separations do not make sense in the slightest. You can have something that will do everything at once: assortment, mix, volume, pricing, geographical spread, etc. I mean, obviously there is an engineering problem in doing that, but fundamentally near, I would say all the slice and dice is pretty much something that is built for by humans for humans. And so, that’s what gets me to the sort of thinking where, okay, we have people and you say we have to be the CEO responsible. Obviously, the COO is ultimately responsible for everything, so that always gets back to this guy or this person at some point. But my take is that when you look at the granularity of decisions that are needed to run a supply chain, it is just vastly in excess of what a person can do.

Again, I would say 40 years ago when my parents started at Procter & Gamble, a company like Procter & Gamble would operate with 200 products and that would be the same 200 products that would be sold in every single country so that people in Cincinnati, I believe headquarters, could just supervise everything with just a few, not even spreadsheets at the time. So, you would have like it was manageable, but supply chains have exploded in complexity. Even a company that used to have like 200 products like Procter & Gamble, they are probably, I don’t have specific numbers, but I would be pretty sure that worldwide they have like probably over 50,000 products nowadays. So, it is inhumanly large and with twice as many countries and probably four times as many storage locations. So the bottom line is that the complexity is immense, has grown, and I think it has grown due to the applicative landscape that I was mentioning. And so we have to really embrace not the sort of slice and dice that were the tradition, but really the fundamentals of the problems and the what are, you know, the engineering limitations.

And when I say pros or cons of S&OP, is that S&OP unforces, you know, puts the company in the place where it operates with humans instead of thinking what comes next.

Conor Doherty: Joannes, perfectly on time. I didn’t know that tone was about to sound sorry, but with that, Milos, your five-minute rebuttal please.

Milos Vrzic: Yes, actually I would agree with you. There’s going to be a time and space where there’s not going to be a CEO to a company. It’s going to be, Chat GPT 1,500 that is going to be running the company. The day that happens, the very day that happens, I agree with you, you no longer need S&OP. The day before that happens, you’re still going to need S&OP. Why are you going to need it? It’s a purely behavioral issue. So, you’re talking about how the complexity of the business means that there’s no way that you’re absolutely right, there’s no way there’s anybody has the bandwidth to take all the data and shove it into a meeting of one and a half hours so that the CEO can decide. So, basically what that requires us to do is to work in terms of families for the CEO to be able to make the decision for that actual thing to be able to take place.

So this is, I think, very misunderstood in a lot of S&OPs. This is, you’re not the only one, there’s many other people who don’t see this but that decision making cannot be done on SKU basis and cannot be replaced by a machine. A machine can make a recommendation, but the question is who’s going to be pulling the trigger? It’s a little bit like the drones that are used in warfare. Sure, there’s a drone but somebody needs to pull the trigger at some point because we will have to be accountable for the result.

So, and the other point I would like to make is that’s just the first aspect. The second aspect is that every single center, every single department is not just an input and an output result. You have human beings and human beings, guess what, they don’t always tell the truth. So they’re sitting there at the end of their first quarter, they’re looking at their sales figures and they’re crossing their fingers and saying, yes, I hope we can overcome the lack of sales that we did in the first quarter in the next three quarters so we can build a number for the rest of the year. And that’s where somebody else, the person responsible for the company, this is this balancing of supply and demand, has to step in and be like, “You know what, I don’t believe your story. I think we’re going to run into huge inventory issues. This forecast needs to be corrected to what it really should be.” And these human interactions are still the case for companies today. They’re still going to be the case for the companies tomorrow.

Where I join you is that if we can have a system that can modulate all this information and synthesize it for us and throw off recommendations, that would be a huge step forward. And any supply chain manager of his or her ilk would agree to that perspective. But the bottom line is you’re still dealing with human beings at the end of the day. And those human beings cannot pull a trigger if they are not convinced themselves. It’s actually, at the end of the day, it’s pattern recognition and emotional management. That’s what S&OP is really all about.

And we cannot start preventing it. So the day that we have somebody who is going to be replaced by an LLM, then of course we’re done. We don’t need decisions, they can be taken in real time as you’ve illustrated in other debates and this whole S&OP process goes out the window. But until that happens and until people are still making decisions, that’s what needs to take place. And also, in one of your other debates, and maybe we won’t have time to talk about it today, you were talking about who’s responsible for inventory or the fact that people are not the stock is not under your control.

Well, some of it is under your control and some of it is definitely under the control of the GM or the CEO who has to decide yes, I believe in what they’re telling me and this inventory raise is going to be lowered because our forecast is correct or no, I don’t believe it’s going to happen and we have to correct our forecast immediately. Or the opposite, we’re going into an uncharted territory, we are launching in a new market. I really have strong faith that we can do better in this market so I’m going to be willing to take the risk of having a higher capital expenditure, more investment in inventory because that’s where I’d like to go. That’s the fine-tuning, the balancing that is still done by humans. And if it’s still done by humans, it needs to be done in S&OP.

I know that’s not the great answer. I know I wish, just like you, that we could press a button and get a recommendation, but unfortunately that’s the state of affairs today, and maybe one day it will no longer be required, but that day will be when we don’t require CEOs and we don’t have heads of business units and sales and no supply chain or finance, that would be the date.

Conor Doherty: Well, thank you, Milos. I shall prevent the ringer from sounding. Okay, well gentlemen, as the opening remarks and the rebuttal have been made at this point, I invite you both to, starting with Joannes, to make your concluding remarks and then we’ll transition to some follow-up questions. So, Joannes, two minutes.

Concluding remarks

Joannes Vermorel: Excellent. So, the problem of saying we can, we should postpone the getting rid of S&OP until a definite amount of time. The problem is that just like e-commerce, it is already too late. You see, having a fully automated supply chain execution has already been happening at Lokad for a decade. It has already been happening at Amazon for a decade. So this is not the future, this is already the past, not even like the super recent past. And that’s the thing with e-commerce is that why didn’t Walmart pay any attention in 2004, 2005 because they were looking at the numbers say, you know what, e-commerce is still 1% of the market, why should I pay attention? And the answer is because when there is a new technology that works at scale and when you start thinking about it, you say, well this thing is just going to crush the rest, just a matter of time.

Late and you, you need to catch up and so, my problem is that you say when you start thinking about until the day where we can do this and that and that, it’s excuses for management to postpone the transition because the transition will not happen naturally organically. The transition in markets just happen through Darwinism. Some companies do it, some don’t, and what is left are the modernized version of the market. Markets are not educators, markets are filters. And so my take on S&OP is that the lack of skills of the management, of the teams, of most software vendors, not Lokad by the way, are mere excuses to preserve fragile egos and markets don’t care about egos. S&OP is not a competitive option anymore and to some extent probably never has been, but it will be either voluntarily terminated by companies or involuntarily terminated by their rivals.

Conor Doherty: Thank you, Joannes. And Milos, your concluding remarks please. Two minutes.

Milos Vrzic: Sure. So the way I would put it is, maybe one day it is possible to have an S&OP process and I’m sure that you’re right, I’m sure that you’re 100% correct in saying we can have an entire S&OP process. It can go through the four different stages and it can give us recommendations and we can even automate triggering. The question is, is the market mature enough to actually take that decision? In other words, to have something that is highly impacting your P&L be done by AI. The question is, are people ready for it? And the answer is no.

It’s a little bit like, would you be willing to board an airplane where there are no pilots but only AI piloting it for you? What do you think the answer to that question is? The answer is probably not. You’d be worried that there isn’t a human being because of the uncertainty factor, because of the comfort factor, etc. And unfortunately, companies are not any different than airplanes being flown by AI pilots. They need somebody to be on board to be accountable. It has to be a physical human being and as soon as it is going to be a physical human being, that physical human being is going to have to take decisions and those decisions cannot be taken lightly.

You can’t, you know, like I said in my opening remark, go into somebody’s office and be like, here’s 10 million, please sign the check now. You’re going to say, well that 10 million can already go to the best bidder of a warehouse and it can be ordered without any intervention of human beings. But then we’re basically saying that the corporations as an entity have ceased to exist, and I don’t think that we are there yet. We might be there one day. I’m not saying otherwise. That could have happened and probably it has already begun in some sense. But until that actually happens, we’re going to need this thing no matter what.

Follow up questions

Conor Doherty: Thank you very much. And with that, we have essentially concluded the more formal, very tightly controlled rules control the fun section of the debate. Thank you both for what is clearly a tremendous amount of preparation and you’re both credits to academia and supply chain.

Now at this point, I do want to ask some questions. These pages were blank at the start, they are now filled with notes, and I will do my best to distribute them accordingly. Just to clarify, because there were some things that were said that I think audience viewers might want to have a little bit of clarification on and I do bear in mind that this is still part of the debate. So I would ask for relatively terse responses so we can distribute time accordingly. But first, I will go to Milos, considering you are in favor of the proposition. And this, this may be a quick question, but the nature of the debate was a net positive for companies and that was the very explicit agreed upon wording, generates more value than it costs. And yet, I didn’t hear at any point an explicit metric by which you’re measuring the efficacy or the profitability or what is the metric essentially by which you predicate your support of this practice please?

Milos Vrzic: Sure. So the metric is simply put in your P&L. So what I would really like to illustrate sometimes is show the example of companies that don’t have an S&OP, that have never had an S&OP and then what happens to their P&L once they’ve implemented an S&OP. There will be a difference of quite substantial characterization and you’ll see this in their service level will go up. You’ll see this because their inventory level will all finally be under control because sometimes their inventory level, you know, sometimes because you’re lacking an S&OP process, what supply chain managers sometimes say, oh we don’t have a process so I better increase my buffer because I know those guys in the top shelf of the office are not going to make the right decision. So you’re going to see it there. You’re going to see it in networking capital. So there are definite benefits to it.

Now those benefits are not necessarily very visible. I’ll grant you that. You know, I think in one of your previous remarks, Joannes, you said nobody ever outcompeted another competitor just because they’ve implemented S&OP. I would tend to agree with you on that. But what I would also say is it’s also because it’s not very visible. I mean there’s no Olympics of the S&OP. There’s no event when we can actually compare and see which one is more competitive. But what we do see is what the end result is at the end of the day. And that can be better performance, better service level, mastery of the elements and most importantly, you have your CEO who comes in and thanks you and like, thank you. I know where the business is going. I have a handle on the steering wheel of the vehicle that is my company and I’m able to make decisions because we have these repetitive meetings and I know which way we’re going. It’s really just a big exercise just to get the CEO to make decisions and be the judge and advocate for balancing supply.

Conor Doherty: Alright, well thank you, Milos. And for parity, Joannes, do you want us a response to that?

Joannes Vermorel: Yes, I think my take is that the argumentation would be kind of contingencies that are very difficult to argue pro and against just because you can show companies that would illustrate that and some that illustrate the opposite. So my take is more like a by design sort of thinking. My approach is when I propose mechanization of the supply chain execution and planification which by the way doesn’t have to involve any S&OP in any kind of form. It is not about mechanizing S&OP, it’s just about mechanizing the decisions. How it happens can be completely different. My take is that you transform this practice into a capitalistic, value creating asset. That’s it. This is a machine, a thinking machine of some kind.

It’s not that it is without humans. It’s not about eliminating humans entirely. There are humans to tweak the machines and improve the machines. But fundamentally, you have turned the practice into a value generating asset. If people stop working on it, it will keep operating on its own. It keeps generating. At some point this machine is going to be obsolete and deemed non-competitive against the market. But this is what you have. So you transform, fundamentally, Opex practice into a Capex investment. So you see, and that’s why it is like a, it done right, it can be a money printing machine. Not because there are some contingent accidents, improving service level or this or that. It’s just because by design, you have something where every Euro or dollar that is injected into this line of investment is a capitalistic investment that generates interest rates of its own. That’s it, no more. And so that’s a very by design thing.

Conor Doherty: Thank you very much. Milos, I do want to follow up on another point and again, you may need to correct me. I don’t put words in your mouth. But there were a great many concessions, points of agreement where you said, yes, I actually agree with Joannes on this and maybe it’s not perfect but given the current situation. And I just want to be clear, is it correct to say that you didn’t so much present S&OP as a net positive but rather as a necessary evil given the state of software and people’s trust. So for example, you gave the plane example. Like maybe an AI pilot would actually be far superior to a human, a flesh and blood human. But people won’t trust that. Therefore, it’s what we have. It is a necessary evil. Is that a fair approximation?

Milos Vrzic: It’s almost there. What I would add to it is that if you don’t have this process, nobody would trust an AI. That’s for sure. No, no CEO would sign off his company and say, you know what, thank you guys. Now that you’ve implemented this software, I don’t have to handle S&OP anymore. So that would never happen, in my opinion, number one. Number two, because it’s there, what ends up happening is that you get a net positive as a result of it because these decisions that are highly impacting on your P&L will be able to be In my opinion, you’re never going to be able to pull the trigger. It’s not a question of what needs to be done, it’s who’s actually going to be pulling the trigger. So yes, it’s a necessary evil, like all administrations by the way.

I wouldn’t say that it’s anything other than that, but that necessary evil has a massive return on investment. If I can digress for one second, I had, in one of my experiences at Blancpain, I worked with a director of operations. When I started working there, we were not even on speaking terms. Basically, supply chain was wreaking havoc in his operations, nothing could handle it. After I’ve implemented Master scheduling and not only on one level but two levels of the BOM plus S&OP, not only did he realize himself what would be the impact, how much easier for him it is to do his operations, but he was saying he was my biggest advocate. He would go over to the CEO and advocate the necessity for an S&OP.

This is not something written in my CV, but it’s one of my crowning performances in my career. That just gives you an idea of how much of a net positive it can be. The end result is also our market share. Remember, we had one of our clients here in Switzerland, which is a huge store in Interlaken that sold millions worth of hundreds of millions worth of watches. They liked working with us because, guess what, contrary to all the other companies who, by the way, don’t have an S&OP, we were always on time, had delivered, and they were happy to work with us.

Even when others weren’t delivering them correctly, they had no idea that it was S&OP of course because this is, you know, maybe a well-kept secret within the supply chain, but that’s what delivered it. Now, the entire pyramid, I will concede this to Joannes, the entire S&OP process can be automated, AI granted. There’s a couple of things we have to discuss because we never discussed supply reviews, but those are quite tricky to get right. But at the end of the day, at the tip of the iceberg or the tip of the pyramid, who’s pulling the trigger?

Conor Doherty: Well, thank you, Milos. Before I throw it back to Joannes for comment, I do want to synthesize that response with a question that I did want to pose to you. It’s something on top of that, something that Milos has commented multiple times, the idea of emotional management and the involvement, the feeling of emotional involvement in decision-making. But the implicit challenge there is that, let’s say, Lokad’s quantitative supply chain perspective can’t really cater to the need for the practitioner to feel emotionally involved. So even if the tech is there, the person feels excluded. Now, feel free to address that as well in your answer because I think there’s overlap.

Joannes Vermorel: Yes, I think so. First, let’s revisit the case of the aircraft pilots because it’s quite actually a very good example. For the last two decades, what people, the general audience, don’t know is that aircraft from Boeing and Airbus, when you look at the most difficult maneuvers, they are now done entirely automatically. So whenever there is a complicated maneuver, the human pilot is not even authorized to touch the plane.

So basically, the pilot is only authorized to touch the plane when it’s so simple that a human can do it. If it’s complicated, it’s a machine that does it. And by the way, there was, I forget the name of the program, but something like two or three years ago, the US Navy just implemented that also for their fighters to land on aircraft carriers. So, same principle, people would just not do it, and they managed to divide it by something like five, the number of unsuccessful attempts at landing an aircraft on the aircraft carrier.

So again, the most difficult maneuvers are the ones that get automated first. The bottom line is that, yes, there is responsibility. It’s not about having Skynet, it’s not about having that, again, artificial intelligence, the way I use it, it’s not about having a superhuman intelligence yet to just do everything. This is not what I’m talking about. What I’m talking about is just shifting the responsibility from people who are directly responsible for taking the decision right now, toward people who have the time to engineer a numerical recipe to generate the decision for them.

So fundamentally, it is just like saying that you can have aircraft security because the pilot is super good, or you can say, well, people at Airbus and people at Boeing are just going to engineer safety devices so that most of the security is on them, and then those automation will run automatically. It doesn’t mean that we don’t have people at the top of the pyramid, as you were saying, it’s just that those people are in charge of the engineering rather than the direct control.

And the CEO will challenge those people on what have they engineered. So that would be, you know, the sort of things. And that’s where also I think, you know, in terms of adoption, the emotion is very important. But people can have their pride in the quality of the engineering, even if the thing is delegated. You can take your pride in the fact that planes don’t crash because automation is good, even if you cease to take pride in you being a crazy pilot capable of doing stunts and saving the day.

Milos Vrzic: The transition is perfect because as you were speaking, I wanted to give the illustration of, I think it was a movie called Sully or Captain Sully, the one who landed his Boeing in the Hudson. My question is, would you rather be a passenger in a plane where the captain lands it in the Hudson or an AI lands it in the Hudson? So that very much goes to the core of the issue that I’m illustrating. And the other point that I would like to make as well is, there is no, you know, until we have another way of dealing with human beings, there’s always going to be that human aspect.

I always liked the example of people meeting at conferences where they’re talking about the anarchists meet at such and such day in such and such venue, and they have to meet because that’s the only way to get it done. So this is a little bit the same sort of issue where, where that’s, this is what human beings are used to doing. We’re primal animals, we’ve been around for at least 100,000 years, and we do not evolve anywhere nearly as fast as our technology does, unfortunately or fortunately, I don’t know, but that is the state of affairs for now.

Conor Doherty: Well, thank you, Milos. Before I move on to the next one, for parity, is there anything you want to add to that thought, Joannes?

Joannes Vermorel: Yes, I mean, again, the situation with Hudson is very interesting, and we can continue on that. You see, I’m not saying that it should be no human. What I’m saying is that if you have automation, you have people who have bandwidth to even try to attempt the landing on the Hudson. What people don’t realize is that, imagine the situation in this plane having this mechanical failure above the Hudson. Imagine that the two pilots are actually super busy already tinkering the aircraft because if they don’t tinker the aircraft all the time, the aircraft doesn’t fly.

So, and now they have something on top of that, which is a mechanical engine, so they have like zero bandwidth because they are already completely mentally exhausted tinkering the aircraft. So my take is that you should have extended automation so that when something goes wrong, at the very least, you have people that have just the energy and the mental energy. If you go from one firefighting problem to another one, you know, for years at a time, then when there will be a crisis, people will be exhausted and just can’t deal with that.

So my take is, yes, we need humans. Again, we don’t have superhuman artificial intelligence yet, not even close. So, yes, they are. But my take is that automation is actually one of the best super simple recipes to make sure that when there will be something out of the ordinary that requires this human intelligence, people are ready and available to deal with that. Because the reality is that what I see in large companies right now is that when there is a crisis, what they have is like hordes of low entry clerks who are completely lost in this new situation.

And maybe the higher-ups can have the time to think it through, but then they have to explain to their armies, and then due to the fact that they like 40 countries and whatever, it just takes six months, which is too long. So again, if you have automation, you can reduce the number of people and have those people, you know, do their intense thinking and indeed discuss and whatnot, do the human animal human things that humans do, and get to some sort of solution.

There might be something out of the ordinary that requires this human intelligence. People are ready and available to deal with that because the reality is that what I see in large companies right now is that when there is a crisis, what they have is like hordes of low entry clerks who are completely lost in this new situation. And maybe the boss, higher-ups, can have the time to think it through, but then they have to explain to their armies, and then due to the fact that they like 40 countries and whatever, it just takes 6 months, which is too long.

So again, if you have automation, you can reduce the number of people and have those people, you know, do their intense thinking and indeed discuss and whatnot, do the human things that humans do and get to some sort of solution. You see, that’s my point. It’s not about removing the people. It is about making sure that at least they have the opportunity to act when the time will call for it.

Conor Doherty: Well, I do want to thank you. I do want to push on and just make it a little bit more concrete. So Milos, again, I will go to you first because again, you are representing the pro side here. So, take a few minutes as necessary, but when you’re talking, give a perspective, a glimpse into your perspective on the life cycle of the tactical decision-making that you’re talking about. So again, we’re talking tactical, 12 to 24, possibly 36 months. What is the life cycle of that decision in the context of monthly meetings? Is that monthly retinkerings of plans to open a warehouse, constantly revising, getting everyone’s input on, “We’ve decided to spend 9 million, not 10 million,” basically hindering progress? So, all of that’s to say, please, could you sketch out a life cycle of a decision, please?

Milos Vrzic: Absolutely. So, one of the outputs of an S&OP process is not just the consensus on what the forecast should be. One of the outputs is a list of decisions that have been taken. And what needs to happen after that is, guess what, in the next cycle, we don’t have to revisit it. But if there is a reason for a supply chain manager or others to point out that the conditions that we had last month have changed, well, that’s the opportunity to do it.

I mean, imagine a company that doesn’t have an S&OP, and they’re looking at their long-term plan. They’re like, “Oh, we absolutely need a factory in Asia,” and they start building it, but nobody’s checking to see whether, you know, are sales still going to be there? Has there been a new entrance into the market? Yes, all this can be automated 100%, Joannes. Nobody’s going to say otherwise. But the bottom line is, then you need to pull the trigger. And then when you pull the trigger, I didn’t mention this, but you don’t only have one option. You will have one, two, or three options.

So, for example, you might say to yourself, “Option one, I need to get new machines. I need to acquire them.” Or, “Option two, you know what, the growth, I’m not sure it’s going to be there in 36 months. There’s an uncertainty. Rather than buy the machines, I would rather we invest in CMOS.” And those decisions, when they’re taken in the S&OP, then you can go on month after month. So, you don’t go back to them. It’s very clear that you’ve made the decision, but it’s an opportunity to revisit it because sometimes you will have to revisit it. And that’s, you know, the primary reason why it has to happen on a monthly basis. It cannot be done daily because you don’t have daily tactical decisions, and it cannot be done quarterly because that eats into your whole lead time for implementation.

Conor Doherty: Thank you. And, Joannes, I will immediately give you a chance to respond to that.

Joannes Vermorel: That’s why I defined supply chain as a mastery of optionality in the presence of viability for the flow of physical goods. So, the mastery of optionality is, you know, pondering those options and revisiting them. But again, if you have machines, you can revisit all the options all the time. And most of the time, you would say, “Well, altering a decision that has already been made is not worth it.” You see, the economic cost of undoing that or amending that is just too great.

So, people would think that if it’s a machine, it would just change its mind all the time. No, if it’s correctly engineered, it will stick with a given decision, a given course, for as long as it makes sense, but not a day more. And again, how many options should we consider? Well, as many as you can possibly humanly conceive. Again, the machine is not going to invent options, but if you have smart people, you can implement the options to be considered, and then they become part of your daily options that are considered, again, all the time, in real time.

So, that’s where I think, you know, entering the Machine Age really changes your perspective because suddenly you realize that your capacity, your bottleneck, is not what you can do but what you can think of. Literally, you know, if you can’t think of it, your bottleneck becomes the human intelligence. And I would say that’s a good bottleneck to have, as opposed to having as a bottleneck just the man-hours that you can put on the case.

Conor Doherty: Well, thank you. And I do want to, in reverse order, so I’ll actually come back to you, Joannes, and then I will go to Milos again for variety. So, just in trying to respond to two things that have just been said. When, Milos, when you talk about decisions, and to you Joannes, when we think about a quantitative supply chain, on the in terms of the quantitative scale, if we have some clients for whom we, let’s say, we generate 60,000 purchase and dispatch or allocation decisions per day. And the whole or part of the cell there is, we take care of that, and so it frees the bandwidth for you to focus on strategic decision-making, possibly tactical, like the kind that Milos is talking about. So, I mean, isn’t Milos sort of selling what we also sort of sell? We might be using different names for it, but it’s, is it not the same thing?

Joannes Vermorel: Not quite. I mean, Lokad, you see, the problem is that we also operate within the market, and the problem is that the market has pretty much framed, over the last couple of decades, the sort of decision that could even be automated. So, Lokad, as a matter of fact, very frequently, we are only automating not the decisions that could be automated, but the one where the client thinks it is possible, and which is not entirely correlated.

So, there are many decisions that could be completely automated, but people just do not even think it is possible. One of them would be, for example, it’s fairly straightforward to automate, but people don’t even think that it’s typically possible, is range planning. Which is, you’re a fast fashion company, and you have like 50 design ideas, and you want to expand those 50 design ideas to 20,000 distinct products with size and color variation. This macro inflation process to go from 50 design ideas to 20,000 products that just explore the combinations and whatever can be entirely automated.

There is very little actual intelligence in that, but people, but there is no established market for this sort of thing. So, my take is that pretty much all of those decisions can be automated. And also, the existing software in the market tends to frame things in very bad ways. When you have like service level and safety stocks, for example, it tends these paradigms create problems whenever you have like constraints between products that need to be ordered together or containers or shipped together or whatever. So, to keep it simple, is that there is a wide range of decisions. We are not constrained by that, but there are, again, Lokad has to operate with the expectations that were built for us before us.

Conor Doherty: Thank you. And Milos, I’ll give you a chance to respond to that. But in your response, a little bit pepper in a little bit more context. How many decisions, be it simple forecasting or strategic, how many decisions can you realistically expect a CEO and an S&OP team to feasibly tackle in a single meeting? Let’s just assume an hour, in the context of everything else they have to do that day. And then again, if you’re talking about monthly meetings on a two-year life scale, that’s 24 one-hour meetings, for example, on the high end. Realistically, how many decisions can they chew and dissect with any meaningful degree of investigative vigor, shall we say?

Milos Vrzic: That’s a very good question and the answer is two to three decisions maximum. And this is a very important point. I’m glad you raised it because a lot of the points that Joannes, you’re raising, are issues that are really very much tackled in the master scheduling aspect of planning.

In the master scheduling aspect of planning, yes, you take care of those mundane decisions or those operational mixed decisions. But then when you get into the S&OP, you’re not discussing that at all. You’re not ever discussing things that are not going to be of the caliber of attention of the CEO. What you will do is you will raise the issues that have gone because this is something also we haven’t explained quite fully. You know, S&OP is not one meeting. S&OP is at least four steps: demand planning, supply review, pre-S&OP, and S&OP.

So when it gets into the CEO’s office or when we have the meeting with the CEO, we’ve already outlined all the difficult issues that he needs to decide on. And even the decision of not taking a decision is obviously a decision. This is something we, I enjoy protocoling like, fine, don’t take the decision, but we’re going to come back to this in a month’s time probably. So the answer is two to three decisions and these are capital intensive, rather important decisions, and also balancing supply and demand. This cannot mean it’s at some point somebody has to pull the trigger. And this is something I’d like to ask Joannes.

Do you think, Joannes, that there can be a time where you would have an operating system like an AI that would take care of the entire S&OP process? By the way, I’d be the first customer, let me make that very clear. Because at the end of the day, do you think that a CEO would be like, “I need to spend 10 million for XYZ machines, let me sign the check off we go.” Do you think that that can take place realistically in a company?

Joannes Vermorel: But that is already taking place. We are on a daily basis, and I really mean on a daily basis, especially for aerospace companies. Just to give you an idea, an APU auxiliary power unit, it’s about $5 million a piece. We recommend on a daily basis to buy or sell those. So those are, and literally, this is the machine says and they proceed. So can we have like super capital intensive decisions done automatically? Yes, absolutely, it’s already done. But the thing is again that is that forces people to reconsider what the CEO is going to talk about.

You see, the CEO is not going to think about whether this operation should be done or not. This is accidental. At a certain scale, it is just accidental. No, it’s about what is your strategy. So what does quality of service mean? This is very complicated and this is something that is changing. For example, if you go from, do you consider that your, let’s say, aviation maintenance company should go into billing your clients per hour. So you’re now going to sell aircraft where you say, “You know what, the maintenance is going to be all inclusive and we are going to charge you by essentially by flight hours and flight cycles.”

And you see, this is a sort of consideration. Should we go all in into this sort of way to charge for the service or should we keep selling the parts and every operation differently? Those are the sort of questions that we go in, discuss and obviously what we want is to have the CEO really challenge the fundamentals of their markets, what they’re really thinking it through, as opposed to go into the fine print of whether, you know, the warehouse should be located here and there and they need a second one and whatever.

You see, again, to a certain extent, those things are still distractions. And again, when people say you have like two or three decisions per meeting, you know, that’s fine. I would say the problem is that this is a very low bandwidth sort of process. And so there are very few decisions that end up in those meetings by necessity. And I think that also explains why those companies are kind of slow.

Sometimes you need to ramp up much more rapidly. And yes, there will be, I mean, people might be surprised, but yes, there might be a day where we suggest that there should be like 20 warehouses being open, you know, same month. And they are cheap and they can be put in various places and it’s fine.

So yes, it is surprising. It gives a much more, I would say, the mechanical aspect gives a much more, I would say, rapid trigger happy sort of vibe. But due to the fact that it considers the cost being associated with that, it will also very frequently make fairly prudent decisions just because, you know, economic assessment would say if it’s cheap then you can do it quickly. If there is no, you can undo the decision especially.

So that’s why I would say, you see, that’s my core problem is that people, instead of focusing on the true essence of their business, they focus on elementary steps. Even if those elementary steps are capital intensive, they are not necessarily the things that matter the most.

Free exchange between Milos and Joannes

Conor Doherty: Thank you. Well, gentlemen, at this point, I’ve exhausted all of my prompt follow-up questions to try and impress you. I hope I’ve done it in a dispassionate and neutral fashion to everyone’s liking. We’ve been going for, I believe, just over an hour. So at this point, I will allow you, you are allowed if you wish, to engage each other directly with anything that you would like to clarify or follow up on. So, free exchange, gentlemen.

Milos Vrzic: Sure. What you said earlier just now about the kind of decisions that you’re going to make for an airline company or airline manufacturer where, you know, you’re going to be changing your business model, that’s not the kind of decision you could ever take in an S&OP. It’s not even the venue for it. That is typically the kind of decision that you take in a strategic decision. Usually, you would retire off to somewhere in a resort, take all your people and discuss, you know, brainstorm what the future looks like. It’s not a monthly basis. So my argument would be that that is a strategic decision, not a tactical one.

And I still think that, you know, at the end of the day, the human being has to be there. And one day that might not be the case. One day there might be a propensity to be able to sign off millions of dollars. But, you know, there’s an accountability issue and that accountability, you can’t trust a huge machine just coughing out a number and saying like this is what you have to invest. They have to have visibility and understanding.

And it also doesn’t answer the question. You know, you said this to yourself in one of the previous calls, not calls, excuse me, your videos, where you know, it’s a cancer that needs to be removed. And that can’t be replaced with anything else. And the argument is, yes, it might be a cancer, but I’ll be very concerned if you do remove the cancer because you’re going to replace it with Alzheimer’s.

People are going to forget why they took the decision. You’re going to lose accountability. “Oh, that wasn’t me, it was the system who did it.” And all these issues are going to lead to probably maybe even bigger problem than the cancer itself. So that’s my perspective. But you did bring out quite a few topics that I think are quite interesting, notably the fact that we can automate it.

And I would never tell you otherwise than automate the hell out of it from A to Z. And if we can replace this bureaucracy with something more synthetic, then great. But, you know, and then on the other hand, good luck with that.

When I look at the supply aspect, now the demand aspect can be automated quite surely. You’re absolutely right because we have nodes and we have a stochastic optimization just like you guys do. But when you look on from the supply side, that becomes really complicated.

You have a factory where you’re following the capacity, let’s say in one of the production lines, and it shows up that everything is okay. And then in your conversation with the factory head in the supply review, which is the second meeting in the S&OP process, he tells you, “By the way, our warehouse is a bit on the tight end.” You’re like, “Excuse me, where’s the table for that?” And he’s like, “Well, we haven’t really thought about it.”

And then you go into the process of like, “Okay, so you’re having a warehousing issue.” And it’s like, “Well, it’s not just the warehousing, we don’t have that many forklifts.” What? So this is the real challenge, the day-to-day challenge that people have to face. Yes, I’m sure that it is already automated in some companies. And I welcome the opportunity to automate it in any other future company. But it is a very tricky process.

Joannes Vermorel: Thank you. Let’s consider the case of e-commerce in the year 2000. You know, and let’s revisit all the arguments that were given about e-commerce because I was there and that was exactly the arguments that were given in saying that basically e-commerce was, you know, never going to fly. You can’t just imagine if you don’t have the feedback of the sales guide that tells you a product is defective, how would you know? People would buy online and who would even tell you that there is something that is just wrong? How would you even know that your catalog is missing something? Normally in a real store, people come if there is something missing they would ask a clerk so you get the feedback and etc, etc, etc.

So basically, there was like an endless series of objections that seemed absolutely obvious and now that we have e-commerce, people are like, yeah, it’s obvious that all of those objections have technical solutions. Most of them are actually fully mundane and they are easy. But you know, in the year 2000, I mean, just imagine the massive retail companies, they pretty much every single one failed at taking the turn of e-commerce just because they were having this long list of things. Oh, you can’t remove people from there, you can’t remove. But the reality is that if you look at Amazon, they have like a million employees. So it’s not because you mechanize things that you just remove people. Actually, Amazon has more white collars than probably any retailer in history.

So my take is that it’s not about removing the people. There will be people. It must be accountable, there must be transparency, and you must understand what you’re doing. Again, that’s why I usually, on this channel, use the term numerical recipe rather than AI.

Milos Vrzic: Welcome to S&OP because you just did the four ingredients exactly required.

Joannes Vermorel: But what I’m saying is that the key difference is that you just want the numerical recipe to automate the mundane. And the mundane, when I say the mundane, it’s disregarding the horizon. The mundane for today, the mundane for 4 months, and the mundane for 3 years. All of that is that repeats should be entirely automated, no matter the horizon.

Conor Doherty: What about the problems that are not repeatable? So for example, you might have a system that says all right, you’re going to have to buy, you know, in this scenario it’s going to be warehouses, in that scenario it’s going to be machines, and so on. But what if you have, I don’t know, let’s say a virus, a COVID that shows up and that completely disrupts everything. What about if you have not just the virus but what if you have a new entrant in the market that’s completely from the left field that nobody saw coming? And I’m thinking Nokia in the old days who were very comfortable selling phones when Apple, which is a software company, shows up and like, oh, we have this thing called the smartphone.

Joannes Vermorel: But let’s consider again Amazon, who did the best during the COVID and lockdowns. You know, so again, you would think, okay, Amazon has automated everything, so in theory, due to the fact that they have complete automation and whatnot, being disrupted should have killed them. That’s the theory, but the practice was no, they did super great. So you see, my take is that automation is freeing up the bandwidth for the management to actually think it through when there is some stuff that is happening. That’s my point is that your capacity, ultimately, you can’t predict unpredictable. That’s pretty much a definition. So the only thing that you can have is managers who have time on their hands and the mind free to think it through.

And my take is that if we robotize that, it is the best antidote against crisis because you will have plenty of people that happen to be available. Which comes with a small paradox by the way, which I said you can remove people, but the reality is that when you put those sort of super high intensive automation in place, you need to have more people than you think is strictly necessary. So you end up with a lot of people who don’t do anything on a daily basis. And I fully agree, precisely so that when there is something strange that comes on, they can step up and deal with that. And this is strange, it means that instead of having people that are busy all the time, you end up with much fewer white collars who are most of the time not that busy.

Milos Vrzic: It’s basically theory of constraints, you know, you need to have excessive capacity so that when you do have the bump, guess what, you’re in the constraint, you can respond to the need.

Joannes Vermorel: Yes, exactly.

Milos Vrzic: One last point, if I may. I don’t know how much more time we have. One other thing that I found interesting about S&OP is that it can evolve in all kinds of environments. One such environment is DDMRP. Now we’re not gonna, I’m not gonna open up that can, I know your feelings about it, Joannes, so we’ll leave that for another debate. But what I found really interesting, there was a moment when I followed a conference with Carol Ptak and Richard Ling, one of the co-inventors of S&OP, and they, you know, contrary to what you would think, you know, great DDMRP, it’s basically a dynamic Kanban that runs your entire business and is triggered only by the actual demand. In theory, sounds fantastic.

Why would you need S&OP in such an environment? It sounds like it’s something that you can chuck out the window. Carol Ptak was like, no, no, no, no, no. And you can’t accuse Carol of being anything but, she’s not shy to speak her mind and she is quite a contrarian herself. She was like, oh God, we really need S&OP and this is why. And she has made, and I don’t want to repeat myself, more or less the same arguments that I have right now. And where I reach you and I think that we do have common ground on this is that absolutely, can get it automated to death. That goes without saying.

Conor Doherty: Well, at this point, I believe there’s nothing much further to say except I will ask you both for just closing thoughts in general based on what you’ve heard today and in general, and then we’ll give the final word to Milos. But first, Joannes, anything to end on you want the audience to go away thinking about?

Joannes Vermorel: Yes, I mean, my takeaway is, by definition, when there is an evolution of a technological change, by definition, people are not ready, companies are not ready. It’s pretty much the definition of change. So my message is, do not wait, do not postpone until you feel ready because you’re never going to be ready. I mean, the reality is that the economic history is littered with companies that were not ready until they were just no more. So my take is that it is more straightforward than what most people think. Again, the amount of genuine, super careful thinking that goes into most of what qualifies S&OP is not as sophisticated as what people would think. It is sophisticated, yes, but not nearly. We are not nearly at the pinnacle of human intelligence. We are not about discovering quantum physics. It’s a lot more mundane and some reasonable numerical tricks can go really, really a long way.

So my message is, don’t be afraid. You have to go there and think. You have to think about the day when, 20 years from now, those things are entirely automated and think, when should your company do that? And not put yourself in the position of Walmart with respect to e-commerce, which is, there is time, there is time, and then there is no time. You have Amazon and Amazon has just taken the market and they could have done that at any point of time, but they didn’t. So my take is, it’s not even expensive. That’s the thing, is that there are plenty of things where you would need to make huge investments. E-commerce was quite capital heavy, you needed to build fulfillment centers, but here, it’s actually quite cheap. So my take is, give it a try and then if it doesn’t fly, retry in a few years. All things considered, it’s one of the least risky propositions that you can have with regard to AI nowadays.

Conor Doherty: Thank you, Joannes. And Milos, as is custom, closing thoughts to you.

Milos Vrzic: Well, I think that’s a fair statement. I’m more than willing to take up that risk at some point and see how a software can replace the entire S&OP process. That would be a welcome change because, you know, though we are developing it and it’s our core business, we don’t enjoy it that much. It’s not the best of exercises. The tug-of-war that you have at meetings, etc., is not the friendliest of environments. So I would be willing to test that sort of thing with pleasure because I think that’s, you know, why not? Why not try it out? Obviously, in parallel, though, would be my backup of S&OP where I’d be checking, making sure that everything is online. But yeah, I think that’s maybe the way of the future. Absolutely. But at the end of the day, like I said earlier, in the cockpit, there’s a pilot. He’s living and breathing. His name is Sully.

Conor Doherty: Well, gentlemen, thank you very much for all the insights. Sincerely, I’m quite optimistic that that will inspire hopefully some solid and even-handed discussion in the space itself. But at this point, I’ll say, Joannes, thank you very much for your time. Milos, sincerely, thank you very much for yours and for joining us. And thank you all for watching. Hopefully, we’ll see you next time.