Ep. 8 - Supply Chain Design: From Mainframes to AI Agents with Laurie Tuschen

What if your supply chain could answer hard questions in minutes instead of days? We sit down with Laurie Tuschen, Head of Customer Strategy at Optilogic, to explore how modern design stacks blend optimization, simulation, and AI to deliver faster, smarter decisions without adding new buildings or bloated spreadsheets. From the early days of mainframes to sensitivity-at-scale tools, Laurie shows how teams can test uncertainty directly and pinpoint when a decision breaks, where capacity buffers matter, and how mode shifts change the cost-to-service curve.

We walk through the practical power of a digital twin: a living model that captures facilities, flows, policies, and costs so leaders can run continuous what-ifs. When tariffs swing or lanes get disrupted, the twin helps reroute through existing assets, rebalance inventory, and protect service levels. Instead of periodic, months-long studies, cloud-native collaboration turns design into a weekly habit. Stakeholders enter data through intuitive apps, see clear scenario results, and build alignment around quantified risk, not gut feel.

AI is the accelerant, not the autopilot. Large language models make the twin queryable in plain English, while AI agents fill data gaps, profile quality, and automate scenario assembly. Humans stay firmly in the loop to judge tradeoffs and feasibility, but they spend their time on insight instead of grunt work. The payoff is real: flipping the old 80/20 so more energy goes to evaluating options, answering more project requests, and making resilient choices under cost volatility, labor constraints, and global shipping shocks.

If your team is ready to move beyond spreadsheets and embrace continuous design, this conversation maps the path; faster modeling, richer scenarios, and decisions that hold up when the world shifts. Listen, share with a colleague who owns network strategy, and leave a review with your toughest what-if we should tackle next.

Transcript

Host Setup And Guest Intro

SPEAKER_00 0:07

I want to clarify that this podcast is distinct from my responsibilities as a professor in the Sam M. Walton College of Business. Nonetheless, it aligns with my aspiration to provide practical insights to professionals and business by showcasing companies and people that can enhance your ability to manage, lead, and strategize and market effectively and the retail value chain. And now without further ado, let's get into the exciting episode. I have with me today Lori Touchan, who is the head of customer strategy for OptiLogic. And she has an incredible background in supply chain management, supply chain optimization, and product management of supply chain optimization solutions. So, Lori, thank you for joining me today. I appreciate it.

SPEAKER_01 0:55

Thanks. Thanks for having me.

SPEAKER_00 0:57

Lori, would you tell us a little bit about your background and what you learned, you know, in being involved in product management?

SPEAKER_01 1:05

Sure. Yeah, I um have been working in the supply chain space for about 25 years. Um I've been both on kind of on both sides where I've worked both in the software side, working with different software companies, and also um in the industry. So it's been kind of a nice mix of being uh, you know, putting together what these great supply chain software products are, but then also using them as a customer as well. Um, I started out my career at a company called Aspentech, and I did a lot of different things there on the supply chain uh side of things, mostly in planning, scheduling, um even forecasting and demand management. From there, I ended up joining a chemical company and being a part of the supply chain center of excellence there. So we were able to, you know, really look at how we implement tools like these supply chain tools that I work on today and roll them out across the business, find value, all of those great things. From there, um I joined Lamasoft, Coupa, and now over at Optologic and have been working more in that supply chain design and planning space. Um, it's been a great, uh, you know, a really great road that I've been able to travel to go through all these things. Um, worked with tons of great companies along the way. It's always fun to work with a lot of these um really interesting supply chain problems that come up in business challenges. Um, so yeah, it's it's it's been a fun place to be. Supply chain is, you know, always changing, always interesting, and lots of new things going along.

Then Vs Now: Optimization’s Evolution

SPEAKER_00 2:45

It is changing a lot, but what seems like it's changed more recently and more quickly. Um, you know, when I first got exposed to and involved in supply chain optimization back in the early 90s, you know, the whole problem was really around getting trying to get the transportation rates. And um so in some cases it might be which DCs serve which stores, it might be which uh factories serve which DCs, et cetera, et cetera. Um which DCs serve which markets. Um and and all of that was driven by you you'd you take a ton of time to figure out what uh rates to use, freight rates, which wasn't easy. Um you you you get that all loaded in and then you uh you have forecasts of of volumes that you're gonna have over the lines, and you you use uh you know we used um uh an optimization approach, um uh like a linear optimization, but uh mixed integer uh linear optimization. And it would come up with assignments and you know, you'd get a cost. Um and and all that took a long time. Like to run of a a decent size problem, you'd have to go on a mainframe and it might take the whole weekend.

SPEAKER_01 4:17

Yeah.

Simulation Joins The Toolkit

SPEAKER_00 4:17

But things have changed a lot, and I remember, you know, I uh you know, in the in the maybe late 2000s, it got to the point where you could do it on a spreadsheet if if you if you knew how to do it and using a solver type of thing. But now it seems like everything changes so quickly. What what is the state of the art now in terms of thinking about supply chain optimization problems?

AI Agents And Data Gaps

SPEAKER_01 4:42

Yeah, no, that's a a good question, and yeah, definitely a good lead up to there. I do think in the last three, four years, there have been a lot of changes where, you know, in the past, like you said, we all took advantage, any software company that I worked with in the supply chain space, we all took advantage of those optimization solvers, linear program, mixed linear, mixed integer programming. And that was kind of where it was, right? And and they do a good job getting you to a solution. But now we can start to take advantage of other technologies and bring them together with that, you know, mixed integer programming that we've had in the past. So um it, it's I, you know, I find it really interesting now. One, we can bring in things like simulation to look at, you know, not just what should those answers be, but what happens when things change, right? So we know that there's variability. Uh, you're never going to have a perfect forecast. You want to be able to say, well, what if my forecast goes up, down by 3%, 5%, whatever that is. So the ability to go in and start to simulate some of those types of changes, um, I think is really powerful. But then in addition to that, um, we are starting to bring AI into all of this as well. And that is, you know, a real game changer. Um, as we start to layer things like, you know, AI agents that can help with your data caps. You mentioned, for instance, hey, I don't have a freight rate for something I've never done before. So I want to look at potentially putting a DC in a location where we don't, you know, we don't have DCs there today. What would be the rates to get there? Hey, let's create an agent that can help you find benchmark rates or go out to a source, get those rates, bring them in, and much more quickly be able to model out a what-if scenario like that and answer your question of, you know, okay, what would my network look like if something like that changes? Um, so bringing the tools together and layering them on top of that powerful optimization engine that we have, bringing in that simulation and the the AI tools that now are available today are gonna make this process of answering questions a lot faster, right, than it's ever been before.

SPEAKER_00 7:01

Well, that's that's really good to hear. Um because uh and again, I I haven't been, I haven't actually done a supply chain optimization in years, but I remember one problem was, right, you have you have the expert set up this problem in an optimization tool, and um you're pulling all this data in from many p places, and it's a mess. You get it in there, you run it, you take the results, you put it in a system where people can look at it and actually execute against it, um or or a spreadsheet even. Um and and then they look at it and say, well, this isn't quite what we want. And getting all the people to because you know, you have to get a lot of different parts of the organization involved in this process. It can take a long time. Is it easier now to do that kind of what-if analysis?

Cloud Collaboration And Speed

SPEAKER_01 8:05

Definitely is. Uh, you know, I would say one of the big changes that made all of that kind of work easier, right? The ability to start to collaborate with different users around your organization, get their input where you need their input, and then be able to run additional, let's say, scenarios and things like that. That really, you know, it was a big change when cloud software came around. And we could start to, instead of trying to pass around these large, you know, mixed integer program models that we had, which were huge and big files, and you know, it had to be on a machine that could run a big problem like that. Um, once it became cloud software, it was really easy to take a problem like that, store it somewhere on the cloud, give the right people access to it, share with them the pieces of it where you need and want their input and you know, to make that process very, very collaborative and much easier. So that's been something, you know, just the advent of cloud software has made all of that collaboration much easier. In addition to that, I think now, you know, people are so used to using apps, right? We all use apps on our phone. So the ability to go, you know, to go in and to tell someone, hey, get into this app that I've got on the cloud, you know, whether it's on the AppLogic platform or somewhere else, and there's going to be a screen there where you can put your inputs in and, you know, maybe you can see results, things like that. People are so used to working in environments like that that um, again, that's a great way for us to be able to, as software providers, um, you know, put that collaboration capability into the hands of our users so that they can much more again, more quickly get to those answers that they they're looking for.

Robust Decisions With Simulation

SPEAKER_00 9:55

You know, getting back to the simulation piece, I want to talk about that a minute. Um, but but when you look at just optimization, when you're doing just optimization, it's like you say, it's it's not very realistic because just pure optimization things change. And really as a leader in supply chain, you want to know how rebo robust are these decisions we're making. Um if I think maybe some rates from one location to another are going to change a lot or might change, how much do they have to change to change the solution? And um uh by using simulation um but for those that don't know who are watching, when we say simulation, it's a way to put in uh uncertainty about uh freight rates, um capacities, et cetera, et cetera, and um and demand. And so you can simulate different things and see, well, if that way if I think, yeah, demand over here is very uncertain, I need to take that into account. Using simulation, you can do that. It used to be very difficult to use simulation with um optimization. So it sounds like you all have solved that problem to a large degree.

Sensitivity At Scale Explained

SPEAKER_01 11:29

Yeah, I think you know, it's a very standard part of our um modeling and you know solving process for these supply chain models that um, you know, that can be built on the optologic platform. So to take a model and say, hey, I'm gonna do my first pass as a network optimization type of run. And then on that, I want to now start to simulate some of those different conditions, uh, you know, different scenarios that may happen. It's built right into the software, quick and easy to do that. Um, you know, we've got we've got a really like some great features, one of them we call sensitivity at scale, which allows you to kind of ratchet things up and down very easily, run those type of simulations to see how does that change that. And then from there you can understand where that tipping point is that you were talking about, which is, you know, when does my solution suddenly become not robust anymore? And maybe I need to, you know, add that extra capacity, um, look at moving to a different transportation mode if the cost inflection point gets there. So um, yeah, it's a great uh great feature that again, I think it's something that, you know, in the past people used to do things like this, but it was you had to do it by brute force. And so taking that one scenario that now you want to run the different permutations on could take you days to do, where now it takes, you know, a few minutes, even an hour. You know, you can get that done very quickly.

Mode Mix And Data Automation

SPEAKER_00 13:07

So here's another question again. I I haven't personally been involved in a network design project for years, but you know, I remember one project we were working on was it was all truckload, right? It was truckload inbound, it was truckload outbound. That's easier than say, well, we could use truckload, or we could use truckload with stop-offs, or we could use LTL, right? There's all these different combinations of things you can use. And that was really hard to do even 15 years ago. Is it easier to look at those options now than it used to be?

Modeling Inventory And Profit

SPEAKER_01 13:49

I think it is. Um, so even when you think about the different, you know, types of scenarios that you were just describing, probably the the toughest part was gathering the right data to support those kinds of scenarios. So, right, if you don't ship by LTL, how do you know what that costs? Right. If you you don't, so now I think given a lot of the tools that we have that can help you quickly go find that data, put the data into the format that you need it to be able to now model on that. So maybe, you know, maybe you're aggregating data, maybe you're looking at only a region in the country, something like that. You need to be able to quickly filter, clean, process that data. Um, all of that is something that, you know, within tools like like Optologic, we have a new tool, uh new application called Data Star. You can set up these very automated workflows to go out, find that data, prep the data for your modeling, build your scenarios. All of that can be done in a very automated fashion. So, you know, it's something that isn't, you don't have to recreate it each time. You go in, you know, you kind of set it up, you get it, get it going. And the next time you may be evaluating a different mode, okay, now we need to go out and find road rates for that mode, but we're gonna run some scenarios on that as well, right? So um a lot of the things that in the past you had to do through through real work can be automated now, and uh therefore you can get them done much more quickly.

SPEAKER_00 15:21

The other one that was really hard to deal with was yeah, if if transportation costs represent, you know, 80% of the costs, then you optimize according to that. But sometimes, you know, you're dealing with really expensive items and um it's not just transportation costs, it's the inventory costs as well. Can you model that in as well?

SPEAKER_01 15:47

Yeah, yeah. You know, these supply chain models, um, it really they are meant to create. We like to talk sometimes about a true digital twin of your supply chain. And so that idea of, hey, you know, I think my transportation costs are outweighing everything else. And, you know, why do I even need to model the other things? Because that's really where I'm making the decision. What we find when our customers are building these supply chain models, they have kind of a set of ideas that they think are true. And as they start to build out that model and look at how things change within the model, they realize wait a minute, our inventory costs are a bigger part than we thought, or could be driving them uh to do certain things that maybe they should do something different, right? Decisions should be made differently. So the supply chain models that we have, you know, you have the ability as a user to bring in any and all of those different costs and things that could impact things, even if you're not sure that they're uh, you know, influential in a certain decision. So we have transportation costs, you can bring in inventory costs, you can model your production, um, you can bring in profit as well. So you can bring in your revenue around demand, things like that. Um, lots of great ways to really get that true digital twin of your supply chain.

SPEAKER_00 17:11

Digital representation of something physical, at least that's how I think of it. In this case, it could be the model and the the optimization and the simulation of the physical process, like transportation network or distribution network. And so the difference between just simulation and a digital twin would be that you there's a coupling between the two. Uh so in other words, you you come up with a model, you optimize it, simulate it and optimize it, and then you take that information and operate according to it, but then you take that information that's actually being executed and update your model, re-execute the model, and then change the decisions and the and the um actual execution. Is that how you would think of it?

SPEAKER_01 18:06

Yeah, I think I think that's a good description. I guess when I one of the um things that I think about in a digital twin is the ability to take, to have that representation of your current, current physical supply chain in a way, have it set up in such a way that you can quickly ask questions about what happens if you change that physical supply chain. So by having this digital twin in place, your data, your processes, your policies, your costs are all set up in a way that you can very quickly, let's say, generate a supply chain model that you want to ask a specific question to. And so, but that digital supply chain is really important as that foundation and starting point, um, not just so that you can then ask and answer questions, but that so that you also have that baseline type of information so that you can understand as I change, as I look at changing my physical supply chain, what is the impact of those changes? So understanding those baseline costs as part of that digital twin. You can then say, okay, if I make a change, if I go to a different shipping mode, if I add more production capacity, how is that now going to change my costs or profits from that baseline that I started from? Um, and that way you can make, you know, more informed decisions.

Cost Volatility And Micro-Redesigns

SPEAKER_00 19:42

So, Lori, what are customers really thinking about now? I mean, you know, I think that what I mean, I know that you know people are thinking about how tariffs affect where you should bring product in. Um that has a big impact on your network, but also there's just a lot of change going on. And you have you have now we've got lots of new data centers being started. I would think data centers would be a big customer. I don't know. Is it?

SPEAKER_01 20:12

Um, you know, right now we haven't we have not had uh customers in that space yet. Um, but that is definitely going to be an interesting uh thing as we move forward here.

SPEAKER_00 20:24

So so what are customers interested in right now?

LLMs And Human-In-The-Loop Agents

SPEAKER_01 20:27

Yeah, I I mean you're um you mentioned tariffs. I do think there's generally a ton of cost volatility right now that customers are seeing. Some of that cost volatility is due to tariffs, but there's certainly, if you think about some of the shipping issues that um we've seen lately, like, you know, in the Red Sea, the Suez Canal, those prices in terms of shipping are fluctuating all the time. Uh, not only prices, but capacity is very volatile as well. Um, even labor shortages, things like that, right? We've kind of seen where there were some massive labor shortages that seems to have got a little gotten a little bit better. But prices are very hard, you know, the costs and things in a supply chain are changing all the time. And so what we see companies working with is they want to be able to make, they're not always looking at making massive changes to their footprint. They're, you know, you know, they're not always saying, hey, we're gonna build a new plant or build a new DC. But sometimes they just want to change the way that flows are happening through their network to look to take advantage of cost changes as they happen. Um, or, you know, to as as tariffs happen, they want to be prepared to have something in place that they can quickly change those flows without actually physically changing their network. They can make adjustments within their existing network. So I do. Think that's something where we've gone from maybe you know thinking really about supply chain design being something that these companies do once or twice a year. They look at only the big projects to being a much more continuous version of supply chain design, where every week, every month, they're looking at making smaller changes to their network as well to take advantage of any, you know, changes in costs or that cost volatility that they can.

SPEAKER_00 22:28

How you all are uh building solutions, but also um how your customers are thinking about solutions.

Flipping The 80/20 Of Data Work

More Studies, Better Decisions

SPEAKER_01 22:36

Yeah, so we have a lot of AI initiatives happening within our product footprint right now. Um, what we see when we talk with customers and prospects that we're working with, everybody is asking about AI. Most companies, I think, have some AI initiatives right now. They're not always sure exactly how they want to utilize AI, what's the best way to do it, but they do want to engage with companies like ours, like Optologic, that have AI initiatives in place because they know that this is the way of the future. There's really no getting around it. So when we think about AI and how we're incorporating it into our software, there's the classic LLM. I want to be able to ask questions about my supply chain, whether it is that digital twin, I want to make, I may want to ask questions about that. I may want to ask lots of questions about, well, what if I change my supply chain in this way? What's my total landed cost to get, you know, goods out to this particular customer, things like that. So that's kind of a you know standard, I think now that most software applications need to include some kind of LLM that will help you ask and answer questions. But in addition to that, I think where we're really going next is taking that a little bit further and getting into AI agents. Now, supply chains, and especially when you're looking at more of the supply chain design and planning space, I don't think they can, they can't be 100% automated or anything like that. There's too many things that occur that, you know, aren't based on historical changes, which is what AI is often looking at, predicting the future based on the past. So when you get into those agents, the way that we are thinking about that is how do we automate things that can be automated? So, as an example, data cleansing, filling data gaps, um, you know, maybe even model building. Um, things like that are things that we think can be fairly automated, maybe not 100%, but things that we can add automation to to make those more, you know, much more uh faster in terms of getting to that end result. But we want to really keep the human in the loop where it is important for the human to have uh, you know, have their hands in there. And so when we look at things like building out scenarios, again, you can automate the process of building the scenario, but starting to understand and utilize and analyze the results of those scenarios is something that humans absolutely need to be involved in before you just go and execute on any of these changes. But presenting those results to the user in a way that they can quickly understand, evaluate, and analyze those solutions is something that AI can help us with as well. So we can add agents to do things like data profiling, data cleaning, maybe scenario building. It can automate that process and then present back to the user, hey, here are the results. You get from that kind of data to model to analysis much more quickly. And then the user gets in to start to say, okay, these are the changes I actually want to make within my supply chain. Um, so I do think the AI agents are something that are going to change the way that most of our supply chain practitioners work.

SPEAKER_00 26:03

I think you're right. I mean, it's interesting because um, you know, I mean, ChatGPT hasn't been out that long, just a few years, and then all these other solutions have come out uh as well. But what's interesting is you know, for a long time companies, it seems like right now companies have just got to the point where they realize, oh yeah, we need to be teaching our employees about prompt engineering or champion programs or et cetera, et cetera. Well now agents make you realize, oh, that's uh the past. They really need to be focused on agents. And uh but it's exciting to hear what you're talking about. And I can see I mean, my gosh, if people can use agents for data cleansing and acquisition and things like that, it it really makes it more worthwhile to and maybe more valuable to have more experts in supply chain optimization doing these things.

SPEAKER_01 27:07

For sure. We, you know, we always in the past would say to actually build your supply chain model, get the data ready, do all of that. That was about 80% of the time. And then you could spend about 20% of the time on the value add of evaluating what-if scenarios, making decisions, sharing analysis. So what we really want to do here is flip that, right? Instead of saying 80% is spent on data wrangling, which is not all that much value add, but you have to have good data, right? Could we now spend 20% of the time on that? And so through these agents and automation and things like that, let's shrink that time down so that now these users can spend that 80% of the time on the value add activities of actually looking at what are the right changes to make to my supply chain to do better here, to serve our customers better, um, to save on costs, to make more in terms of profits.

SPEAKER_00 28:07

It's interesting because when you first think about AI and supply chain design, planning, optimization, et cetera, even execution, you start thinking, well, that's going to take jobs away, but actually it increases the value of all of this so much that it it really makes the marginal benefit of having people doing design, planning, optimization, execution go way up. So as an individual, I can do I can add a lot more value to the organization than I could have in the past.

SPEAKER_01 28:46

Yeah. Again, I what I've seen working with, you know, lots of different companies in this space is there's often a team of modelers that are they're really good at what they're doing. They understand a lot about the company data and the business challenges and all of those things. And often they have a list of projects that the business has asked them to look at and evaluate. And usually they can't do all of them. They have to turn people away and say, you know what, you're not at the top of our list in terms of that question. Right now we're working on, you know, this type of study and it's going to take us a month to finish. And we can't, we can't get to your questions until later. So I think now, you know, what will happen is these businesses will be able to evaluate more and more questions. They'll be able to make smarter supply chain decisions because they're able to do a more extensive analysis of how things should change. I can answer more of those questions. I can take on more of those projects as a modeling team that in the past you just didn't, you know, you never had time to get to.

Closing Thanks

SPEAKER_00 29:53

Well, Lori, um, your background is so impressive. And um, thank you for sharing your knowledge with us today. We really appreciate it.

SPEAKER_01 30:02

Thank you. I appreciate it. It's been funny.