AI, Volatility, and the Future of Supply Chain Decision-Making

PODCAST TRANSCRIPT

Greg Giesecke

So I think a lot of people will talk about AI being fun and cool and interesting. The main thing that supply chain practitioners are thinking about with AI right now is frequency.

Brian Betkowski

Hey, welcome back to Jabian’s Strategy That Works podcast. My name’s Brian Betkowski. I’m here with my friend and colleague, Ed Haines.

Ed Haines

Hello.

Brian Betkowski

And then we got a great little topic today, although it’s not a topic that I feel like I have expertise on. It’s a topic that I definitely feel like I experience, which is supply chain and all of the crazy disruption that we’ve been in. I mean, disruption’s always been a thing, but since COVID, if you think about the amount of disruption in supply chain from, of course, COVID, then everything from tariffs to the Suez Canal closing to just all sorts of just craziness that has happened in the world. The topic today is volatility kind of the new norm or is that an exception anymore? And that’s what we’re here to explore today.

Ed Haines

Excellent. An excellent topic. Yes, I have a lot of experience in this as well in terms of getting packages delivered on time.

Brian Betkowski

Yeah.

Ed Haines

And I think what’s interesting about it is the volatility has always been there, but our reliance on the supply chain is much, much bigger as well. And so the way we feel that is obviously very different. But I have some stats.

Brian Betkowski

All right. Give them to us.

Ed Haines

I’ve written down here. Thought we’d start off with. So interestingly then, obviously, as we just said, the amount of volatility and the amount of demand has increased dramatically since post COVID or during COVID times in terms of how much business is using that supply chain. And we see that in our daily lives, but that demand has obviously gone up. Interesting stat here around freight rates from Asia, so typically China, but Asia to the US used to be around $2,000 per container. They peaked to $20,000.

Brian Betkowski

It’s amazing.

Ed Haines

Which is for those doing math at home, 10 times as much, which is crazy. I mean, and think about what that does to that individual price on an item. And when we talk about peak, how does that vary in terms of, do you have to buy at the peak or do you buy a little lower and so on. So I know that’s a topic for today a little bit for us. There’s lesser stat, but more of a comment. The ratio of sales, the amount of inventory that most companies keep in order to have on hand to sell has actually gone up because of that volatility. So that’s one of those variables and part of the equation. And at the same time, AI has given us this capability that we can do much more analytics than we ever have. And yet the volatility’s there. And I know we just talked about that, but that’s kind of the key to this, is the stats kind of show that volumes up, volatility’s up, and then the ability to calculate and do analysis is up as well. So how do we balance that?

Brian Betkowski

Yeah. Yeah. So today we have a great guest, and this isn’t going to be around a particular tool or a particular process. We’re basically going to discuss, is volatility the new norm and are we using in our businesses, are we using assumptions that are just the old world, just not the new world? And in the new world, are there just a whole new list of assumptions or can we break some of our old assumptions and adopt some new ways of thinking? And we’re fortunate to have a great guest with us today. And so Greg Giesecke, who runs our Denver office and who also leads our supply chain practice, is going to join us today and is definitely an expert in this topic. So welcome, Greg, to the show.

Greg Giesecke

Thanks everyone. Excited to be here.

Brian Betkowski

All right. Well, yeah, I know in planning for this, you have some really interesting ways of just articulating not only the volatility issue, but what other things, what other doors does that open when you start to have this topic? So yeah, ease us into it. Let’s jump into it.

Greg Giesecke

Yeah. Yeah. I mean, you guys started to speak about both sides of really what gets us excited about the supply chain optimization opportunities that we have ahead of us right now. And there’s two sides of this coin. The one side is volatility. It didn’t just increase. I think a lot of times when we’ve saw something like COVID over the past 50, 100 years when you do a long-term study, they refer to it as a black swan event. Right. It’s a once in a lifetime occurrence. But the challenge that that has is it sometimes can have an aspect of explaining it away. It’s a black swan event. It happened once. Don’t worry about it. It won’t happen again. But what happened during COVID was a series of implications that drove sustained volatility. And that’s really the new way of thinking about things, is we’re not trying to think about the risk of something happening.

What do we do to react to it? We really need to think about the volatility levels that we operate under and how those are going to constantly be in place. And the other side of that coin that makes this so exciting is really the capabilities that are available to us. So when I kind of got into the supply chain practice, there was a new solution that was available on the market that handled perishability really well. And the amount of computational power that that perishability software took required that you had to add a ton of servers to a physical infrastructure and run an overnight batch that was really fragile. It was something that worked about five out of seven nights a week. And when it didn’t happen, you didn’t have an update on procurement and purchasing decisions for the next day. That level of computational power can be done every minute.

That level of analysis that’s done right now is not a weekly, a monthly, it consumes a lot of resources. It’s something that happens all of the time. So I think there’s two aspects that we kind of wanted to chat about today. One is, what do we mean by volatility? What is the different ways that you’re experiencing that? And then how do we keep a pulse on it? And then that keeping a pulse on it drives what I’ll call the optimization function. One of the benefits that we have kind of within the supply chain practice, the supply chain field of thinking is that there’s a lot of math around it. There’s a lot of mathematical study. There’s been a lot of work that’s done to try to model what happens in the real world, to model that through mathematical equations. Again, those are very time intensive and they’re hard to do, but that aspect of what’s fixed, what’s volatile, and then run the optimization function, that’s basically what we try to do within the supply chain field.

Brian Betkowski

So is it that the underlying math has changed or is there like some other. Are the other forces at play here? Give us some insights into those differences.

Greg Giesecke

Yeah. Yeah. So there’s two things to think about. So the theoretical math has not really changed, right? It’s evolved over time, over the past handful of years, and it’s getting better. We’re thinking about that. But lets kind of break down what is volatility. So the first concept that we think about within supply chain is start with the customer. The 1980s aspect of supply chain was we’re buying stuff, so we start with suppliers, but we’ve really turned that around and said the arrow begins with the customer. Our main goal is to understand what do they need, where do they need it, and when are they going to need that demand signal? So that’s what we refer to as demand. You can think about that from your perspective as a business leader of the SKUs that you sell and the purchase rate that they’re bought on, but that’s really the second step.

The first step is what is your customer need in the market and how are they going to buy that from you? So during COVID, we really saw customer demand shifts immediately. That was something where the pace of change happened much faster than we’ve ever seen in really historical times. The consumption demand, the in demand of what I need did change, and the channels that I bought it from changed dramatically. So an example of that, you think about food. Right. We eat about the same amount of food a day, but if you’re eating at home versus eating at the deli at your office versus taking a client out to dinner, that’s different channels that you’re buying food from. So we still have fundamental underlying definitions of how many Americans are there that are within my market and what do they need, but how I apply that to then my service and support of them shifts dramatically.

One of the really interesting things, at that time, my neighbor worked at a manufacturing plant for Del Monte. They made spinach and the consumed amount of spinach was about the same, but they needed to put it in a lot of small bags and send it to the grocery store instead of put it in great big bags and send it to institutional customers. Restaurants were closed, grocery stores were empty. That was one of those tiny examples that kind of made me realize there’s a demand on a business, but it’s driven by a consumption, and that consumption channel, that consumption path changed dramatically. Along with that same volatility shift, we think about mix. So the financial reconciliation that we do all the time as business owners says, “What am I going to sell? What’s my implication on the supply chain and how’s that going to kind of move downstream?” But even where you’ve got the same customers coming to you, that mix shifted dramatically.

And I think the big takeaway with demand planning is the historical. The way we think about this is I look at my history, I do some analysis and I project the future. COVID taught us that is not enough. So that idea of taking historical data and trying to predict the future, that’s not enough. We’ve always known that other inputs are helpful, right? I want to know if the price has changed. I want to know if I’m doing an ad campaign through marketing, but we’ve kind of put that on the back burner as supply chain leaders and said the most important thing is understanding trends and how they’re going to go in the future.
But I think a huge shift in that demand volatility, again, it’s not that the math has changed, or our theories were wrong, it’s really that we need to reinforce all of these points to take external demand signals, take input from marketing, get an input from the customer on what they see happening with the end consumer. Demand volatility isn’t just more noisy. We’re really seeing a tide change on how we think about that demand side.

Brian Betkowski

So go ahead.

Ed Haines

Well, so in that example of the demand changing then, Greg, I mean, it’s a classic situation of the data. You would have data typically to kind of understand what your forecast would be. And if everything stayed the same, you could generally kind of predict it. And then suddenly you have these spikes or these black swan events that then changes obviously your forecast. Does what’s happened over, the volatility in itself over the last few years, has that data helped with future forecasting, if that makes sense? Or is that something that we just really don’t know, volatility is volatility after all? Are companies using that old data or are they kind of doing the classic sense of, well, that’s a 100-year events, so therefore it’s an anomaly and so we’re going to take it out of there? I mean, how are people using the data we’ve collected over the last seven years?

Greg Giesecke

Yeah. Yeah. And you can think about the data as two simple data points. So if you think about a company that sells one widget, let’s break it down to say they just distribute one item to one customer. There’s two things you try to define about it. One is what’s the average, right? What’s the center of the bell curve? What’s the mean? And then what’s the spread? So how wide is that bell curve? And the two things that we saw was both the demand was moving constantly, so we can’t assume mean is based on some level of history and that volatility signal isn’t just a signal of how much inventory I need to hold to meet the width of the bell curve. It’s a signal of a lot of other things that are happening within my business. So the fact that the volatility is up or down is a signal to other components.

And that’s where this concept of I’m going to plan demand and then execute a bunch of math that turned into S&OP, sales and operations planning, where we did a balance and a reconciliation and that has matured. I’d say 2018, we were talking about integrated business planning. Today we have the business case for it and we can really realize what happens when I predict demand, I constrain that based on operational supply, I reconcile the finance and controls, and then I look at what do those signals tell me for the rest of my business. And that’s another big kind of impact that we’re seeing, is that you probably had a business model pre 2020. You probably had a business model that was mayhem during the three, four years after 2020, and now you’re thinking about this new business model.

And those business model shifts have classically been years on time horizons. And now what we’re saying is those can pivot in a month, and that’ll kind of get us into our AI and capability side, but thinking about how a company’s business model, what’s their role in the economy, what do they rely on, those are driven by some of those volatility signals that we see that says, my customer’s demand for a handful of products, those all correlated together tells me something that product management, and strategy, and other components of the business need to take into account. And they’re not huge monolithic components. The other-

Brian Betkowski

Oh, I was going to say, so this might be where you’re going, but what I was thinking is like, and it sounds funny to say because COVID was only six years ago, but it just feels like eons ago because it was like, well, that was a whole event and then the economy went crazy with all the stimulus and all that stuff. And it was like the age of the employee and it was of the war for talent and then op, that turned around, went the other direction. And then all of a sudden AI was introduced. And so I feel like there’s been five whole generations since COVID. So which one are we on now? In your world, if I’m a supply chain executive listening to this, I’ve been through all that, and now what should we be thinking now? Which one are we on?

Greg Giesecke

Yeah. I mean, you can think of that, took it to code level five, it was red. That said customer demand shifts are off the charts. They’re extreme. Pre 2019, it was relatively stable. We kind of had stability of demand, represented through a lot of things within the marketplace. Right now, we are still in extreme volatility. So we’re not on this once in a lifetime phase, but we are in this mindset of what do I need? As a consumer, what is my demand going to drive? And then the level of shifts that we see within channels, you talked about the age of the employee. Seems like any conversation we have goes to, with executive leadership, it’ll come up that says, “What’s the work from home policy? How do we think about that?” And that drives customer demand. Thinking about traffic, logistics, where do you buy your food from? How much do you need of given products?

As those things shift, we’re making much bigger consumption demand swings in the past five years, not even as a result of COVID, just a result of all of the shifts in the marketplace that are happening post-recovery. We’re seeing more dramatic shifts in demand than we’ve seen in the past. So that level of predictability, that level of demand is off the charts. And the demand volatility is one of the three major sources that we think of. That’s one of the major challenges of volatility. There’s a concept that says, let’s try to hold everything constant and think about just one, but now let’s talk about supply volatility, which obviously has an impact on demand and vice versa. But there was essentially global congestion. You talked about the $20,000 shipping container. That was essentially a result of the millions of shipping containers that were sitting in the ocean waiting to get through the Suez Canal.

That was one ship that had about 100 ships sitting in queue waiting to go through that process. That caused disruption for 18 months. That was a wild process that broke the markets. Those ships that were on the water for three or four months constrained the core supply of getting my product from continent A to continent B. And there’s a lot of businesses that don’t have $20,000 worth of margin in a shipping container. Nike has a classic example where they were paying $40,000 for spot rates on shipping containers because on their containers is incredibly valuable and high margin product. If you move nails from a steel mill in China to a distribution facility in North America, you don’t have that margin. You didn’t have that option to be able to buy that, so your product stopped. Right now, we think about lead time volatility as the opposite of customer volatility.

So what’s my standard deviation of demand and what’s my standard deviation of supply? We think about that as you’re going to get your orders in full. There’s some risk of that, but you’re going to get it slower or faster than you expect. So during this time of probably 22 to 23, we saw a demand recover, supply channels had not prepared and recovered for that, and lead times doubled. So the lead time variance went to exponential impacts there. The reliability was just a dramatic decrease that we saw. And supply chain, you can think of it as being fast or slow, but the real hard part is when it’s unpredictable.

So we don’t know how prolonged product is going to be on the water. We don’t know how much it’s going to cost for a freight forwarder to give us a container. The safety stock math depends on knowing some inputs and some variables within a range. And when you don’t have that, it really just broke a lot of the mathematical calculations really thinking.

Brian Betkowski

And you’re kind of depressing me going through all this stuff. It’s like, all right, all right. Well, obviously any supply chain executive lived through all this. And now let’s talk about the other side. So what? Now that obviously is the new world, volatility is the new world, like what are you seeing people do about it?

Greg Giesecke

Yeah. Yeah. And this is the exciting part that says that volatility, and it’s funny, when you talk to a supply chain person, the level of volatility isn’t a challenge. Sometimes that’s what makes the business case, and that’s what leaps us forward in operations. So sometimes you do have a little bit of a negative mind on these things that kind of drive you to positive outcomes, but the real pure capability that we have is the improvement and analysis, both the decrease in cost and the increase in confidence that we have from that. So I think a lot of people will talk about AI being fun and cool and interesting. The main thing that supply chain practitioners are thinking about with AI right now is frequency. I have the ability to rerun fundamental inventory optimization and sourcing analysis every minute. We have the capabilities to be able to run that constantly.

There’s a couple of clients we can think about where they’re logically constrained by how many FTEs we have doing a given function. So I can’t have 150 demand planners planning my 10,000 SKUs that I buy. I’ve got two or three people doing that. Now we’re starting to introduce these capabilities that says, what if their job wasn’t just managing your Pareto rule 20% most important SKUs? It was taking that same logic that we do for your most important top 10 high volume, high value SKUs and doing that process automatically with a low level of cost, but a high level of confidence for every item. And that’s kind of the capability we’re starting to see right now.

Ed Haines

If you’re doing those forecasts with such frequency, how does the rest of the business. I mean, I kind of get the idea of, well, if you’re doing more SKUs, therefore you weren’t doing it before and now you can actually do something about that. But how does the rest of the business manage that frequency demand, right? Because yeah, you can do that in one space, but then suddenly now maybe you’re ordering things one day. I mean, it’s an extreme example, but you’re ordering things one day and you go, “Oh, next day, actually I didn’t mean that. I mean this.” How does the rest of the business downstream of that handle an increased capability in that analysis?

Greg Giesecke

Yeah. And that’s kind of getting us back to this point of integrated business planning. So it’s not to say that we’re going to redefine and recut and adjust POs every minute, right? That’s obviously not something that’s good business practice that we would execute through, but you have the ability to play with those scenarios. So we take one client example here, we were thinking through, their lead times are very volatile. Every three months, once a quarter, they take a look at a year’s worth of POs and say, “What’s changed? Do my lead times that are represented within SAP, are they accurate and are my lead time volatility, is that accurate?” There’s a concept that says, let’s do that every day. And then make the thoughtful decision that says, what change are we actually going to make? But being able to do that scenario planning and see, we’re continuously going to be looking at the inputs we have and we’re going to continuously be assessing what decisions should I be making?

And then you may still often say, “We’re only going to update the system monthly because it coordinates with the monthly buy and that drives our capital allocation.” There’s still a lot of these interdependencies, and that’s one of the challenges I think we’re going to face, is that the supply chain world should be the leading edge function. It has classically been a laggard function, but what I mean by that is because supply chain is largely based on an optimization function, it’s largely mathematically driven, we should be in a position where we can have the ability to do daily buys instead of a monthly process. We have the capability to do it. We may not because salespeople don’t have time to spend a consensus meeting every day, but to kind of have that alleviation of what’s our constraint being the supply chain team, I think that’s going to be an exciting thing that we’re all thinking about is when do we change? It’s not our constraint. It’s going to be how we work with other functions across the business.

Brian Betkowski

You used the AI term. Obviously everyone’s talking about AI these days and when we talk about it with clients, we often realize that sometimes there’s a semantics thing going on here because since a couple years ago when this topic got hot, what people really usually mean by AI is a generative AI. But obviously AI as a science has been around for a long time, back in the ’50s and even before a little bit, but machine learning as in the use of math to solve indiscreet problems has been around for a while. And I feel like that applies a lot in your world and has applied for a long time. What about the new age of AI do you think is interesting? The indiscrete or the generative AI components, where does that play in the world that you’re talking about?

Greg Giesecke

Yeah, so there’s a couple of concepts. So in the supply chain world, I think a lot of times we make the joke that says we’re racing organizations into the ’90s. We’re not talking about bleeding edge solutions here. We’re talking about math, automation being based on real data, but there’s a linchpin somewhere in that process that has to be a really dependent decision. It’s an important process and it’s not math, right? It is connect with the salesperson, connect with finance, understand what we can do and make a decision. So what we’re seeing, that’s why I think the interesting thing right now isn’t AI, it’s frequency that’s driven by automation and AI. But if you imagine, I trust my data 80%, I’m going to put that into a Gonkulator and then lead to a series of scenario-based decisions. There’s a couple of steps in that that are really hard.

One is, what do I do about the 20% of my data points that I don’t trust? That’s a decision. I have to either ignore them, I’ve got to fix them, I’ve got to state some events that cover them up. But that process of taking some data, doing some math, making a decision, that’s repeated throughout the supply chain process of planning, sourcing, execution. What we’re seeing today is that we’re able to really focus on some of those linchpin items and be able to do that in a way that’s relatively cheap and very scalable based on how they manage their really high important items manually. So if you imagine a supply chain team that manages 1000 items, there’s probably 50 to 100 that they spend almost all of their time on. At the rest of their time, they fight fires with the others, but that other side is kind of math and it’s relatively immature math.

And what we have the opportunity to do is take the logic that we focus our most important items, the most critical things to how our business operates and take that same level of attention and logic and do that for our entire business. And we can do that really quickly by watching what we do, defining that into some logical definition, and then applying that throughout. And it’s usually resulting in not a PO executed order. It’s resulting in some reports that create some governance steps for us to look at. But thinking about things like dynamic SKU policies, consistent forecast diagnosing, understanding, I’ve got a lot of error. It happened recently. What do I do? Instead of seeing that and doing root cause analysis, we can see that automatically do some root cause analysis, create some recommendations, and then the human being’s able to have some input on the decisions.

But that concept of what is the process that we should be doing and then what’s our exceptions? The exceptions are where the supply chain team typically lives. We spend most of our time thinking about stuff that doesn’t fit within a nice happy path process.

Ed Haines

Yeah. I mean, and I sort of think about this in a dimensional way, right, because one way you’re talking about is the breadth of the SKUs that you have and the fact that you were able to expand that. On another hand, you’re looking backwards, not backwards, but down into the supply chain to see how you can better be informed about what you want to buy. And then you’ve mentioned this at the very beginning, but looking forwards then it’s like, or up the chain, it’s like, what can I sell? Do you see, especially with the volatility, not just around the supply chain itself, but around maybe commodities, pricing around particular raw materials, like your company’s ability to make sure they’re selling the most profitable things because those prices change a lot and because you have the analysis on the supply chain, have you seen that kind of fall into the demand-based pricing situations that you get? Obviously we’ve had that with airlines and hotels for a long time, but is this being extended now more to other companies that it was typically out of reach for? Has it been accepted?

Greg Giesecke

It has. And I think that’s the nature of, from one hand, I want to hold everything constant and just look at demand volatility, the volatility of my production unit or the volatility of my inputs pricing. That’s not how it works. Those are all extremely interdependent items that we need to think about. If supply goes up or if prices go up on my supply side because I have a tariff impact that doubled my source of cost from that location, my customer demand is ultimately going to change. I’ve got a series of decisions that go through that, but my customer demand is going to kind of drive with these two points. So that’s kind of back to Brian’s point of how is the math changing? Theoretically, it’s not, but having three variables of volatility that drive three variables of internal volatility that drive three variables of sourcing volatility, all of those multiply by each other, and now the math is so complex and complicated.

In the past, we just kind of gave up on doing that at scale for an operations that needs to happen within a business with 10,000 SKUs. But now we can get to a point where we can isolate what is the most volatile, keep those as variables. Right. There’s a concept that says, I buy this widget from supplier A, I know they’re the cheapest, I know they’re the highest quality, and I’m not going to evaluate that every day. Maybe I do an annual or a quarterly process to see, where should I buy that widget from? Now we can do that much more frequently, and we know that tariffs are driving a lot of volatility on that side. And I think that’s one of the big concepts I think I want us to kind of talk about today, is the decision process is rapid and the volatility process is also more frequent.

But when we can tie those two things together and say, my decision cycle is going to be shorter than my volatility cycle. So what is this decision that needs to be enabled within some time period? And if source prices never change, international tariffs never change, then maybe I can do an annual sourcing decision. But when that changes more frequently, I now need to have a consistent multi-source definition that may say in every weekly or monthly buy decision, where am I getting that from? Those are decisions that we’re making on a much more frequent basis. And the risk that we’re seeing right now is that let’s just stay within this old process. Now everything’s an exception. So making sure that that process that I assume is my core governed global process that we execute, making sure that that is nimble enough to meet the volatility world that we’re operating within. I’d say that’s kind of our fundamental theme that we’re really seeing. And your point, what I buy drives what my demand is. It’s all interconnected.

Brian Betkowski

Yeah. This is cool. I feel like I learned a lot today. If I had a recap and you tell me if I’ve got this, I kind of taken away three things. One, that volatility is kind of here. It’s the new norm. Volatility is not episodic. It’s structural. It should be the structure of what we’re doing. I guess the second takeaway was when you were talking about frequency, it’s that we should throw away those old assumptions of that like, “Well, I do it every 30 days, so that’s how often I should do it.” There should be really almost no barriers on frequency. I guess that’s another takeaway.

And then you mentioned the top 20% thing as I think, “Well, I can only really do it with the most important things because I’m constrained by humans.” It sounds like, well, that’s kind of out the door because if I could get AI agents or I can replicate through automation and the combination of AI and automation, then I really am no longer constrained by like, “Oh, I can only do this important thing for my top products.” Those are the three things I took away. Did I miss any of those?

Greg Giesecke

No, I would just kind of summarize it as, even if you happen to run this theoretical company that was optimized five years ago, you say everything was perfect, frequency, sourcing, planning, we have invested at the optimal decision. That is different now, right? That math is wrong. Those three things that you talked about that said volatility is higher and constant, our capabilities are exploding, which drives much more frequency, and we can’t assume the same assumptions of the past decade. That all sums into the optimization function is pushing us to a new frontier. And it’s exciting. I think the good news is that we are able to deliver better capabilities, deliver better service essentially at a lower cost. That’s our challenge, that’s our mandate, and that’s kind of what the supply chain teams need to be thinking about. And I think the tools are there to deliver it.

Ed Haines

Yeah.

Brian Betkowski

Great.

Ed Haines

I think, right, from my side, what I’m sitting here thinking is math is more important than ever.

Brian Betkowski

We’re engineers. We’re biased.

Ed Haines

Yeah. And then, I mean, even though you don’t need to perhaps lean in as much on it because you have these tools to be able to do it, but it is important. And then the other thing, to your point about, God, there’s so much going on. It’s so volatile and it’s a bit depressing sometimes, but disruption leads to innovation, right? And we’ve seen that over the last few years. And my optimistic side goes, wow, with all this disruption and all this volatility and with the innovation of AI and other technologies, the future is probably pretty. For those who crack it, it’s quite bright.

Brian Betkowski

Yeah. Yeah. It’s exciting. Well, Greg, I appreciate you joining. Obviously, you’re passionate about this. You’re way knowledgeable than we are on this topic. If anybody wants to talk to you more, I’m sure they could reach out. I know you love talking about this topic. I know you’ve got a lot of events coming up around this too, so I appreciate you joining and we look forward to discussing again.