Arkaro Insights

Niels van Hove on Human-AI Collaboration in S&OE, S&OP & IBP

• Mark Blackwell • Episode 46

Share episode

Niels van Hove joins Mark Blackwell to explore how AI is transforming S&OP and IBP from bureaucratic box-ticking to decision-centric planning. Discover why planners waste 50% of their time on low-value tasks, how automation and augmentation differ, and Niels's bold vision for AI-powered "decision avatars" that could revolutionise executive decision-making.

____


Is your S&OP process more painful than productive — a "monthly trip to the dentist" as our guest puts it? 

Niels van Hove, globally recognised thought leader in integrated business planning and human-AI collaboration, joins Mark Blackwell to discuss how AI is reshaping supply chain planning. Niels has defined the current era as the "third wave" of supply chain planning technology, and he argues that the competitive advantage lies not in technology alone, but in creating a culture of human-AI collaboration. 

In this conversation, we explore: 

- Why planners spend 50% of their time on data crunching and low-value work — and how AI can free them to focus on what matters 

- The critical distinction between automation (handling repetitive tasks) and augmentation (advising on decisions) - Explainability versus trust — Niels's contrarian view on whether we really need to understand AI to use it effectively 

- Decision avatars: a vision for AI-powered executive teams that can debate scenarios and recommend options at unprecedented speed 

- Practical advice for CEOs looking to modernise deteriorating S&OP processes 

Key insight: "Anything that can be automated will be automated. The competitive difference will be in human-AI collaboration." 

Niels challenges listeners to take personal accountability for learning with AI rather than sitting back and demanding explainability. As he puts it: "AI is here to stay and will only get better. You can't sit back and say, show me how it works, without showing any interest yourself." 

About our guest: Niels van Hove is an expert in IBP, S&OP, and the emerging field of human-AI collaboration in supply chain planning. He advocates for decision-centric IBP — working backwards from decisions rather than focusing solely on forecast accuracy. His work emphasises that future planners must adapt to collaborate with AI or self-select out of the role. 

Resources mentioned: 

 "The New Machine" by Nada Sanders

Roger Moser's work on decision intelligence

Dick Ling's foundational work on S&OP

Send your thoughts to Arkaro







Connect with Arkaro:

🔗 Follow us on LinkedIn:
Arkaro Company Page: https://www.linkedin.com/company/arkaro
Mark Blackwell: https://www.linkedin.com/in/markrblackwell/
Newsletter - Arkaro Insights: https://www.linkedin.com/newsletters/arkaro-insights-6924308904973631488/

🌐 Visit our website: www.arkaro.com


📺 Subscribe to our YouTube channel: www.youtube.com/@arkaro

Audio Podcast: https://arkaroinsights.buzzsprout.com


📧 For business enquiries: mark@arkaro.com

Niels van Hove:

Planners spend 50% of their time, educated planners, coming from university with years of experience in data crunching and stuff which doesn't add value. That's a waste of human capital. It's a disaster in our modern world. So we have all this brain power, all this waste of human capital. Do we still accept that? That we're gonna keep working like that? I can't. I've been a planner. I can't believe that we keep going like that. And if I talk to supply chain directors that lead a planning team and they tell me they're all burnt out, I can't believe we'll continue like that. So, what will change?

Mark Blackwell:

Hi, this is Mark Blackwell. Welcome to the Arkaro Insights Podcast. This is the show where we help B2B executives with tools and techniques to thrive in a complex adaptive world. So taking us one step further on our journey of serendipity, we have today Niels Van Hoov. Niels is a globally recognized thought leader in integrated business planning, sales and operations planning, and increasingly human and AI collaboration. He focuses on transforming IBP, and we'll learn more about that, from information sharing to a decision-making discipline or decision-centric IBP. And he's really has energy in moving the process forward from being more focused on forecast accuracy to really getting decisions to drive business success. And he's known for defining the current era as the third wave of supply chain planning technology, emphasizing autonomous and augmented planning and cognit cognitive automation. But he does argue that success not lies in technology alone, but in creating a human and AI culture. And stresses that future planners must adapt to collaborate with AI or self-select out of the role. Welcome, Niels. How are you doing?

Niels van Hove:

Very well, Mark. Thanks for having me. That was a great introduction. I couldn't do it better myself. That's very kind of you.

Mark Blackwell:

Well, how did you get yourself to that position that we described? I mean, what's the background?

Niels van Hove:

I suppose it's just a journey. Uh I mean, I started my career 25, 26 years ago. I always had a curiosity, so interest to learn new things. Back in 2000, it was online marketplaces. Uh that didn't exist back then. That was new. I started to research it and I got invited by uh McKinsey to brief them on it just because I was an expert. Because nobody knew it. So over my lifetime, I have yeah, I've done, you know, I pick up new trends and whatever have my has my interest, I add to that and I try to paint a picture of understanding. And yeah, with a background in supply chain and integrated business planning, yeah, came, of course, the uh the era of AI, and I started to think about it a bit more, write about it, publishing, and then uh, well, it results in uh just yeah, uh your introduction, I suppose.

Mark Blackwell:

So great. So not all of our listeners are going to be supply chain experts, and indeed not all of them will know about SNOP and IPP and these terms that we're branding around quite freely. And anyway, my experience is even if people know the terms, they often have different understandings of what they mean by sales and operation planning, the purpose, the time frame. So maybe you can give us a little history of how all of these processes evolved.

Niels van Hove:

We want a history of SWP and I think. Look, for it was a supply chain management back in the Ace where people try to connect the planning in terms of, oh, this is my sales demand, this is what I have to manufacture, so this should be my material requirements planning. But in short, in very short, what IBP integrated business planning is trying to do is integrate different silos in a business to create one common view. So a siloed view is usually not very effective, so we create an integrated view horizontally for a future business plan, but also vertically. So we have horizontal integration between departments and maybe even customers and suppliers of how my business looks like, but also vertically: hey, what's my strategy? What should therefore be my long-term plan, my midterm plan, and my short-term plan if I want to execute that strategy? So now I'm aligned horizontally, I'm aligned vertically, which is sort of this utopia which can't be reached, actually. That's what I'm always saying. But it's so interesting to get there that uh that would have been part, big part of my purpose of having an interest in integrated business planning. And so setting your strategy and be able to execute that vertically and horizontally across your business, regardless of what business or vertical you're in.

Mark Blackwell:

You know, the way, you know, sales and operation planning really was originally the focus was balancing supply and demand. But I love the integration and certainly what we experienced in DuPont was making it evolve and mature to IBP, which is the way that you run your business. Yeah. And really having a view, which is a 24-month view, not just a three-month view, which sadly all too many processes have. Having said that, you know, one of the one thing has to be said in the history is that as I say, is the emergence of sales and operations execution. Because we can touch on this today. The world is moving faster. And there was a recognition that a monthly planning cycle is insufficient, and hence the more daily and the weekly rhythm of sales and operating execution evolved from that. So that's so that's we've got some terms which I think we can all be aligned with. What's been the history of machine learning and AI in these processes?

Niels van Hove:

Okay, these processes, long-term planning, SNOP, IBP, shorter-term planning and execution, SOE, they're supported by technology often. And that technology can even be a spreadsheet. I don't think machine learning in most companies entered that the technology support in those processes beyond five years. I mean, machine learning is a part of AI. There's many elements of AI. And machine learning is often applied for classification. You know, it's some picture a cat or a dog, and classification you can also apply to root causes. Okay, what causes this problem in the supply chain? But it's mostly used, and most of the technology providers would have added a machine learning algorithm to their prediction capability to create a baseline forecast. So that's, I think, where machine learning or AI entered these processes and most of the supply chain in planning. I would say probably in manufacturing processes, picture recognition, etc. Um, that's uh that would uh maybe go back a little bit. But if we talk about planning specifically, I don't think beyond five years ago, there were many places where you would find a machine learning algorithm doing a forecast uh a few maybe.

Mark Blackwell:

The reality is where are we today? What's the good, the bad, and the ugly of the use of these tools right now?

Niels van Hove:

Yeah, so where are we today? So I think um it really got a a big boost in uh so in general, uh in the in the C-suite, in the executive and on board level, uh, when uh ChatGPT was launched. All of a sudden it was this big adoption on a type of AI. Let me say that Gen AI is AI, but AI is not gen AI. And many people had made that mistake for a couple of years. AI is gen AI. No, no, no. Gen AI is a type of generative AI. But it woke up the executive level, and they am I aligned to say this on your podcast, had an oh shit moment. Really? We need to do something with AI because whoa, what is what's possible? All of a sudden there was this sort of this use case, what wasn't a forecast somewhere, you know, somebody in planning did. This was visible, this was all around the world, 100 million users in a couple of days. Um, so all of a sudden it got board-level attention. That was a distraction, I think, for a year, maybe two, up to last year, 24, that oh, AI is Gen AI. We need to do something with Gen AI. No, no, no, no, no. Gen AI is a type of AI, and you have to define what do you need for your business processes or rethinking your business. So it was good for the attention for AI, but it was also a distraction. And now we see that, okay, there's been so many use cases of Gen AI. There is more realization that it's just a part of AI. And Gen AI by itself, yeah, it's nice, but it's not definitely not a complete picture to rethink your business. You know, having a Gen AI bot in your customer, you know, service, yeah, great, but it doesn't transform your business really.

Mark Blackwell:

No, I I loved your framing of it from the moment that the CEOs and the C-suite recognize something needs to be done. Yep. And we've talked about that on previous podcasts of the fear of missing out, the FOMO moment and the and the rush to bring in the shiny new thing, which is one of the reasons why 95% of companies, or according to MIT reports, that started AI projects haven't seen any benefits from them.

Niels van Hove:

Yeah, look, it is that, but it's also, again, where was AI sitting? You know, people were using it at home. You were talking to Siri, you know, National Language Processing. It's AI. It was already on your phone, it was already in your email inbox and in your credit card system for fraud detection and spend detection. People have been using it for years. But in business, it was still like, oh, the data science team, there's a little project here, and I think Jet GPT launch was an oh shit moment, or you know, the world has moved on. We need to look into this thing. Um, which is good. That was a wake-up moment.

Mark Blackwell:

So again, that's the reality is one of the good, the bad, and ugly. There are some companies that are making lots of steps at the moment, but honestly, I was looking at some statistics. Yeah, something like 70 to 80% of demand plans are still made largely with the influence of Excel. And then another survey said 46% or something of uh demand plans are fundamentally made by Excel. Yeah. If that's where we are, what is it? What's what's the prize and you know, is the and what's the opportunity that we can get by going for it?

Niels van Hove:

Look, first of all, Excel is great. We all love Excel. You know, I think it gives just the double click on that for a moment, it gives you freedom outside of the online system. Nobody sees what you do. Uh it provides you creativity, you can make any scenario you want. You feel autonomous, so all these intrinsic human needs are fulfilled. If I'm on a system that everybody can see an integrated system, then I can't do that. So, yeah, bless Excel. I think Excel will, or an Excel-like, will always be there because it fulfills a deep uh human need. Once you start integrating Excel at a large scale, yeah, you can run an Excel supply chain for a couple of products, one DC and one manufacturing plant. Once you do it globally across uh 150 countries, 50 factories, 100 DCs, and uh 10,000 products work that well anymore. Yeah. Yeah, going back. I think what we see here is that look, I started implementing APS advanced planning systems to support planning uh 25 years ago. Okay, and and I walk now into businesses that still run that system from 25 years ago. Well, maybe a slightly updated version, they haven't progressed. So you see this range of of uh and Excel will always stay there, and probably 80% will always use some form of Excel. Um, but yeah, we'll we'll we'll see a progression, I think, at one point of time that we move forward, but what has happened with APS, it's been very slowly. I think here we're in an inflection point that it's not a linear change with the technology we see now. It's an exponential change if you look at Gen AI and what it can create, and then the agentic AI. It's a different world, it's not the same world as we were in the last 25-30 years. And yes, you can stay in it all worlds, but um yeah, how long can you be competitive? I really wonder that. And you know, again, I love Excel.

Mark Blackwell:

Yeah. So tell me, you know, what is it that these AI tools can do? Let's focus first on the on the demand plan. You know, intuitively, one of the things I remember often talked about, but not really done, was scenario planning of demand and maybe thinking of demand as a range of probabilities rather than one number. Yeah. And but because of the sheer slog of just getting one number out, discussions, you know, we never really got to that point of thinking about a range of possibilities and planning for those potential futures. What's possible now?

Niels van Hove:

Yeah, so let's take a step back, you know, demand planning or any other type of planning in the supply chain, you know, okay, 80% uh, you know, use uh Excel or whatever the number is. Planners spend 50% of their time, educated planners, coming from university with years of experience in data crunching and stuff which doesn't add value. That's a waste of human capital. It's a it's a disaster in our modern world. And we accept that. We don't, because I know they're burnt out if you talk to them. Yeah, we use Excel, but yeah, I'm also burnt out. So we have all this brain power, all this waste of human capital. Do we still accept that? That we're gonna keep working like that? I can't. I I've been a planner. I can't believe that we keep going like that. And if I talk to supply chain directors that lead a planning team and they tell me they're all burnt out, I can't believe we'll continue like that. So, what will change? And I've written about this since 2020 or 2019. First of all, we've seen in a supply chain for the last 100 years and in the global beyond supply chain, anything that can be automated will be automated. It's the evolution I've seen for the last hundred years in the supply chain. We got now Waymo uh driving 300,000 uh driverless taxis a week paid. It's happening. Okay, so a taxi can drive around driverless. We still can't manage our data and automate it and cleanse it. We still have to, as humans, do 50% of that. It just doesn't add up anymore. Really? I mean, it just doesn't add up anymore. So, what will happen? The automation, which started in the supply chain somewhere in 1914, and the T-Ford line, where manufacturing got automated. All our physical assets have been automated. Our manufacturing, our warehousing, we got uh driverless cars, all our physical assets, a lot of money there, a lot of labor there, have been automated as far as they can, and it will still continue, but we'll you know we're sort of reaching a max. The knowledge worker under which the supply or the planners, they haven't seen any of that automation. They're still working in the 90s because the focus has been, and this automation, on the physical assets in the supply chain. And we can see the proof, it's just just there over the last 20, 30 years. Now, what will happen is that we see elements of the the dull repetitive task under which data crunching and blah, blah, blah, the repetitive analytics and reporting. We can automate that because it's repetitive, it's the same, we can digitize that, we can uh automate that. Great. Now, from the 50% time that this planner wastes, his capital, his brain power wasted on non-valuated task, we now free up to do higher-level tasks, which we've seen the evolution of humankind pretty much. You use a tool to do something uh 5,000 years ago. Hey, now I can do something else with my time. So we'll free up time with that automation. But then there's a second level, and that's so automation will be the foundation. And but that's the foundation. It's not the ultimate step, is the augmentation. So now the machine is going to advise what a planner could or should do. So I could plan this emotion uh promotion differently or this marketing span differently. Hey, I've detected the gap against your budget. Here are five options you can do about it. Do you agree or not? That's what we're talking about there. So now the automation will be the foundation. That's a given. That's evolution, it's going to happen. Now we see the augmentation of the knowledge worker and the collaboration with AI, between the knowledge worker and AI, to improve performance. And that's where the because anybody can automate, the the competitive difference would be in uh will be in the uh the human AI collaboration. That was a long that was a long answer. It was a long answer to your question.

Mark Blackwell:

If I can I can easily see how people will accept automation for execution processes when it's it's optimizing under relatively low uncertainty because you've got your customers, you've got your orders, and you've got your stocks. And it's just a question of optimizing against that reality. Yeah. As we go out in time, we're losing certainty. Yeah. So this is where the human I think comes in more and more because as I was a great fan of SNOP IBP, for me, it brought visibility. So everyone had the same mental model of the problem that they were trying to solve rather than differing views and differing things which will inevitably cause them to talk over each other. Getting that mental model alignment is key, and that's because of visibility, and hence all the big you know, reliance on what have I assumed this month, like what in each case, what are my top three assumptions, or whatever process you use, and what has changed since last month, so I've got a really clear understanding. One of the challenges then obviously is the black box phenomenon. How do I explain how this forecast was created so that I can start aligning with it? Can you talk about that?

Niels van Hove:

Yeah, I can. I will first look back to a previous point. I agree with you, and I've highlighted it in the paper. IBP, the process of the underlying plan, you know, that can be automated, but the decision making can't. Decision making will be augmentation. And I I wrote the my theory on that uh in 2020 and a published paper. So I agree with you there. IBP decision will be augmented, not automated. But the underlying plan. Planning to create an a forecast, you know, that can be that can be automated.

Mark Blackwell:

So going back to your your question, which was uh remind me again because now I am it's the explainability, you know, aligning if we have the same understanding of what are the fundamental reasons how we got this. By the way, I know John who's a demand planner, and I know he's a good guy, even if I don't understand everything, I can trust him because I know what he does, right? You've got that the visibility, the explainability, and the trust of a human type of process, versus the extreme of a black box stream of numbers.

Niels van Hove:

Yeah. So I I think we had a LinkedIn post on that a day or two ago as well. And uh, I'm I'm I'm I'm expecting a little pushback here or a little uh contrarian view. But I agree, explainability is great, ideally, we want it, but in the end, you want trust. I'm like, all right, do I trust this human? Do I trust this without trust? That has been proven as well. You know, the speed of trust, it's all out there, you know, without trust, nothing nothing moves uh between humans. My proposition is also not between human and machine. So explainability is good, but why do we need explainability to gain trust? So, how long do we need explainability after we build trust? Do we still need it? That's the question I would ask. I would I will always talk about explorability as a human. I need to be able to explore the underlying data, it should be a finger click away. I should be explored the reasoning of the AI, the logic of the AI, I have to be able to guide that logic, I have to be able to set the goal. Say, hey, I actually control this AI, it does what I want to do. So hey, you build you build trust over time. Now let me put you my forward my contrarian view in the supply chain. Have you ever heard of exponential smoothing as a forecast method?

Mark Blackwell:

Yes, yes.

Niels van Hove:

Yeah, have you ever heard of uh King's formula to set safety stocks? That's the standard formula with standard deviation for demand and lead time. Probably 80-90% of supply chain professionals use it, and it's the biggest fallacy in the world because it assumes a normal distribution, which in most cases it doesn't. So, so here's the thing: so these formulas come back from the 1950s, and they they are maths, it's a mathematical formula. We use them, we trust them. I put forward to you that more than 90% of supply chain professionals does not understand the math behind those formulas, and we use them every day. Now, AI is math, it's just very complicated maths. So now we use these things from the fix every day. Change my alpha number, oh, I have a different forecast. Oh, I trust that. I have no idea what the math is behind it, and now I have an AI, I have no idea what the math is behind it, I need explainability, I need to be explained what it does. Yeah, maybe, but for how long? And what do you bring to the table? What is your accountability to start learning with that AI? Don't just sit back and say, I need explainability, be proactive, lean in and learn with the AI. Like a couple of thousand people in the world know really what Gen AI does. Do you know? I have no idea. I know it's probabilistic, but do I really know? No, but I use it, I trust it with smart trust up to a certain point. I ask it to do things like you do, you make beautiful pictures. I don't ask it to uh invest in a company or uh marry somebody or uh buy a house. Uh so I apply smart trust to the AI. So there's a contradiction here that people sit back. I need explainability. Well, they're using a formula for the last 30 years, they have no idea mathematically how this formula works. So, how does that work?

Mark Blackwell:

Well, I suppose I have two reflections on that. One, as you said, it's had 30 years to demand to build that trust, so it's taken time to do that. Ah, that's it. I agree. Reflection. I agree. Time is it's gonna take probably longer than we think. The other thing is we'd like to shift the debate from either AI or human to an and and discussion of the human and the AI. If I'm the CEO or the marketing guy, I assume that the demand planner has expertise in generating the demand signals. So I'm giving the trust, and I assume that he's studied the tools. So I'm much more willing as long as I've got an expert who knows how to work with the machine, the homo father concept. So, you know, I think it's an and-and is his way of getting it, but I'm still a concern. I think it with the reality is it's going to take time before we start making big decisions and getting out into the longer term.

Niels van Hove:

I agree. Time is probably the differentiator, but uh I was I I think it was a reflection after two days ago when we had this LinkedIn. I'm like, hang on, look, what I what I'm pushing back on is, and it was a commentary I made on uh article from Nana Sanders, who is a professional supply chain who wrote the book, uh The You Machine. And you know, it's a great book, and it was a commentary on an article of her, which I wrote in Foresight, and yeah, she proposes uh you know, we need the uh a vision of a company that includes AI and AI collaborative vision, and I agree with that, but let's not forget the accountability of the single human to show interest in AI, to show curiosity, to learn with AI, to bring AI to the table, everything you do. You can't just sit back as a human, don't take accountability and say, give me the explainability. I don't accept that. If I was a CEO and I would in this era of AI, I would pick those people, yeah, they wouldn't stay for long. I would recruit on that and I would put my performance criteria on that. Because AI is here to stay, it will only get more. You can't sit back and say, you know, uh, you know, show me show me how show me how it works without showing any proactive interest in the AI. Yeah.

Mark Blackwell:

Well, I suppose the other thing for me from the demand side is put more emphasis that it is not a single set of numbers, it's a range of probabilities. Otherwise, you know, a demand forecast almost by definition will be wrong in actual state. But that doesn't mean it's no value, right? So this message needs to be communicated, otherwise, people will just blame the black box and throw it away because they're not really appreciating what it's really producing. So maybe that's another way of getting it is thinking about the range of probabilities more clearly.

Niels van Hove:

Yeah, which is already a next step for many planets, I suppose, as well. Uh becoming uh probabilistic. Becoming probistic is a step forward, it's nice. But you already see the trend is sort of you know, it's it's an important capability, but it's in in a range of things with working with AI, it's just you know, okay, now I have a range of possibilities. What should I choose? You know, I came to the point that okay, and uh and I had you know a lot of reflection on that. I was a planner for a big part of my life. I actually realized when I start writing about decision making, that planners don't make decisions, facilitate decisions. So I'm sitting there writing, and I really have an extension moment like gosh, I actually don't make yeah, as a planner, I hadn't make many decisions. I I recommend it, I think, but I actually did not make decisions. So for me, I I turn around and I start thinking back from the decision, say okay, I have a plan, so what? I have a forecast, so what? Does it input in a decision or an action? No, then it's a waste uh analytics waste, I call it these days. Any analytics which is not created decision or anything, you know, which is just insight, yeah, it's always analytics waste. If it doesn't take create a decision or an action, um it's maybe a bit harsh, but that's how we have to start thinking back from decisions and actions rather than focusing on planning, planning, planning only. So that's really a change I made the last five years, having been a planner and worked in planning for a big part of my life. Uh, planners are facilitators, less so decision makers.

Mark Blackwell:

Yeah. So we we've spoken about AI in the demand signal generation. Where else can you see AI having a big impact in the whole IBP process?

Niels van Hove:

Look, if we go back to IBP, um, which is in the end, you know, keeping your plans in line with your strategy and understand the execution against those plans, against your strategy in an integrated fashion. A big point that props up is gaps against your plan. I can have my sales plan, I can have a revenue margin gap, I have my operation plan, I can have a labor cost or production capacity gap, and so forth and so forth.

Mark Blackwell:

Now, if I get uh if I have a budget or forecast and my actuals come in or my new forecast, I mean, I think you described in one of your papers that uh the process that we go through now is a little bit like a monthly trip to the dentist, is fundamentally painful because there is so much work being done on data preparation. And if we could simplify that, we could increase the adoption rate. Because there's another statistic I came across this week. It's only something like 30, 40 odd percent of companies in France have got an integrated process where they connect demand, supply, finance, and executive decision making. So although we've been in businesses that have had it, there the reality is the adoption rate's low because there is so much work that needs to be done that has the potential to be simplified to focus on making decisions. So right now we can measure the costs of the fifty percent of the time just spent creating Excel, finance guys doing scenarios. I mean, that's a measurable cost that we can eliminate by thinking about how to use AI. And maybe that's the way to go for either the the immediate phase. But if I can ask you, relieve you of cons some constraints now, and let's imagine that we accelerated forward, what could happen in the future with AI? So you can see that model for decision making under uncertainty and maybe more agentic AI for the execution piece. Absolutely. Totally. Wrapping up, imagine yourself talking now to a CEO. They've had an SNO P process that's sort of been around, deteriorated, got all the flaws that we know about, um, and is hearing about you and AI and everything. What three bits of advice would you give to him right now?

Niels van Hove:

I've been listening to us. Thank you. Really appreciated that.

Mark Blackwell:

Well done. A good tour. And I hope our listeners appreciate it. Thank you very much. Bye bye.

Head Shot

Mark Blackwell

Host
Head Shot

Niels van Hove

Guest

Podcasts we love

Check out these other fine podcasts recommended by us, not an algorithm.

Just Great People Artwork

Just Great People

The Sixsess Consultancy