Catalysts in Cars Getting Coffee Season 2 Episode 3
Doug Barton, director of the UWEBC, rides with supply chain AI veteran Madhav Durbha in Madison after Madhav speaks to the UWEBC Supply Chain Peer Group about applying AI to planning and decision intelligence.
Credit:
Transcript
UWEBC Podcast Introduction
Customer experience, technology, people, and supply chains are shaping how organizations operate today, but keeping up isn’t easy. At the UWEBC, we connect leaders navigating transformational change every day. Each week, we bring you the journeys and perspectives of Wisconsin leaders shaping the future of business.
From the University of Wisconsin–Madison, this is the UWEBC Podcast.
Doug Barton:
Hey, everyone. It’s Doug Barton, back with another episode of Catalysts in Cars Getting Coffee. I’m here with Madhav Durbha.
Madhav was just in front of our supply chain group. We were talking about AI and supply chain, and there’s probably no better person to kick us off than Madhav today. Do you want to say hello to our esteemed audience?
Madhav Durbha:
Hi, everyone. Good to see you. It’s fun to be driving on a nice, sunny, slightly hot day here in Madison, but it’s all good. Doug’s car is very comfortable, and it’s fun doing this recording here.
Doug Barton:
It is. It’s my small cheat code to get time with people.
I get to drop you off at the airport, at the right gate and TSA PreCheck, and get you on your way. But let’s talk a little bit about your experience today. I want to dig into some of the concepts you shared briefly, but also get your read on the room you were in.
First, you used a couple of things that I think are critical to keeping our thinking straight. You had a concept called value leakage. Do you want to share a little bit about why that’s such an important outcome or transformation to keep in mind?
Madhav Durbha:
A couple of things. I’ve been in the supply chain planning arena for over 25 years now.
What I have seen time and again is companies implement supply chain planning systems, or even decision intelligence systems, if you will, or decision support systems, as we used to call them, and go live with a lot of fanfare. What ends up happening is that a wedge starts forming between anticipated value and realized value.
Why does that happen? There are a variety of reasons. One is that supply chain decision systems or planning systems are built on assumptions about the real world and the rules and policies that govern how your supply chains operate. These are not static parameters.
Doug Barton:
Right.
Madhav Durbha:
They’re dynamic. Suddenly, there is a tariff change. That means there’s a change in your sourcing policies.
Oftentimes, when you implement an enterprise system, it’s like pouring concrete around these assumptions. But it’s all starting to change. In the world of AI, we have entered an era where these systems are getting smarter.
They can learn from real-world changes and adapt. Algorithms can learn and adapt. The other thing is the data foundation. Historically, data got trapped in different silos within an organization. Increasingly, organizations are realizing the importance of building a data fabric and data foundation.
All of this is coming together. I think we are entering a pretty exciting phase when it comes to intersecting AI with the complexity and volatility that supply chains are dealing with. There is tremendous potential to unlock a lot of value.
Doug Barton:
I completely agree. It’s an exciting time, but maybe not everyone sees it that way.
You had a great three-C framework about where someone might fall, from courageous to cautious. Take us through the three Cs and the people archetypes that emerge.
Madhav Durbha:
The three Cs are something I was developing based on what I’ve seen while working with a number of different companies across industries.
That includes both the supply chain planner community and supply chain leadership, whether you’re talking about COOs or chief supply chain officers. What I’m seeing is a pattern where people fall into three different archetypes: courageous, curious and cautious, especially when it comes to AI.
The courageous ones are leaning in. You’re talking about maybe 10 to 12% of the population. They’re actually learning AI by doing. I always say AI is a contact sport. You don’t just read about AI and learn. That’s a great way of learning, but those at the leading edge are the ones who are doing and experimenting.
Then come the curious ones. They’re saying, “I don’t understand this enough, but tell me. Teach me more. I’m curious enough to transition from curious to courageous.”
Then you have the cautious types, who are questioning and a little skeptical. They’re saying, “Where is this going? I’m not so sure. I’m going to sit back and see where this is going.”
What is interesting about this three-C framework is that I can take the exact framework and apply it in an enterprise context. You have enterprises that are courageous, enterprises that are curious and enterprises that are cautious.
Some very interesting patterns emerge when you intersect employees who fall along these three Cs with enterprises that also fall along the three Cs. The idea is to transition more curious people into courageous people and more cautious people into curious people. That’s how we progress our learning.
Doug Barton:
That’s how we move the ball forward.
It’s interesting that you can convert curiosity into more courage. People can start exactly where they are. There is always a next step they can take to engage with technology and unlock new sources of value.
As we said earlier, addressing that value leakage makes this an exciting time to take enterprises and people through that entire journey.
One of the things I thought you made a good point about architecturally, and I recognize this from my time in the software, data and analytics industries, is what you described as the power of “and” as it relates to build versus buy.
It seems like a critical time to invest in platform capabilities, but also a good time to make these applications your own and ensure they support your differentiated value creation. Talk about that power of “and” in the buy-versus-build domain.
Madhav Durbha:
This has been a perennial debate in enterprise technology purchasing decisions. Can we build this in-house, or do we need to buy some kind of packaged software?
Both approaches have pros and cons. With packaged software, sometimes there is much left to be desired around the edges. You have to bridge those gaps by doing a lot of custom coding.
The downside of building from scratch is that the wheel has already been invented. Do you really want to reinvent the wheel?
Doug Barton:
Right. Packaged applications can encode a lot of intelligence and help you learn from others. You’re effectively using their learning curve.
Madhav Durbha:
Exactly. There are a lot of best practices that come into play.
In your industry, the demand forecasting problem and its associated algorithms have been pressure tested. If you are taking production problems and applying optimization techniques, those have been pressure tested and proven.
You can start with that, but we no longer need to look at the world as a binary split between buying and building.
Increasingly, evolving technology platforms allow you to bring together the best of both. This is some of the work we are doing at Frelax, where I am now. A lot of our work is geared toward this blend of buying and building.
What you get from buying is best practices that are already proven and pressure tested. You can then build on top of that and bring in your own secret sauce.
In the world of AI, the timeline from an idea to bringing it to life is compressing significantly. What used to take months or years can now be done in weeks or days.
We should take advantage of that and stop viewing this as a binary split. We should see it as an “and” proposition. This is where the world is going, and it’s one thing I’m very excited about right now.
Doug Barton:
Totally. I saw some of the enabling technologies on your slides.
To remind people, APIs, or application programming interfaces, are what we used to call these connection points. Today, it’s MCP and A2A, or agent-to-agent protocols and connections.
There is this emerging technology enabler that will allow us to buy and build in order to address that value leakage.
Madhav Durbha:
You bring up an excellent point. There is one more thing.
Any decision in supply chain needs to have the right context around it.
Doug Barton:
That’s it.
Madhav Durbha:
For the longest time, the context was sitting in structured databases, locked into neat rows and columns. But we all know that a lot of the context resides in people’s heads. A lot of context is also floating around in emails.
Doug Barton:
Do you think of that as tacit knowledge?
Madhav Durbha:
It’s actually more explicit knowledge, but it is not captured inside a planning system.
Think about it this way. If I’m a sales representative, I might send an email to my manager saying, “I’m landing this big order next week. Will we be able to meet this order?”
Or a supplier might send you an email notifying you that an estimated arrival will be delayed by three days because of a congestion problem.
That knowledge does not make its way into a planning system in the traditional world because a human has to look at it, bring it back and codify it into the planning system.
Through MCP, A2A connectors and similar technologies, what is trapped in your email can now be extracted using language models.
That can then be conveyed to your planning system through connectors such as MCP so that it becomes contextualized intelligence.
A planner making a decision can be alerted to it. Not only can the planner be alerted, but an agent behind the scenes can create a scenario and present it to the human.
The human can still make the decision, but now the human is in the loop with much richer context around that decision.
Doug Barton:
Perhaps what we didn’t notice about the decision intelligence of a decade ago, or even five years ago, is that the reason we couldn’t be as effective as we wanted was that we were missing context. Now, we’re going to have access to it.
Madhav Durbha:
That’s correct.
I’m not claiming this is perfect. We are in the early stages of this evolution. But the possibilities are exciting when you think about blending language-based context extraction with the knowledge residing in your structured databases.
Doug Barton:
I have two more topics I want to get through.
Over our lunch hour, we had a chance to talk about your experience through some of your board and foundation roles. You talked about the 70/20/10 model, or the three-horizon investment model, and I would love for you to share more about it.
We know it’s our job as leaders to run our businesses, improve our businesses and maybe transform our businesses, but making all of those bets simultaneously can feel overwhelming.
What are the bright spots you see for the future? What are some of your hopes for industry, universities and industry-academic partnerships? Give us a ray of hope from your seat.
Madhav Durbha:
I think about these as three horizons. There are horizon-one, horizon-two and horizon-three bets that you need to place.
I would gear 70% of your investment toward what I call proven and pressure-tested AI. Twenty percent should be geared toward emerging forms of AI for which early proof is emerging. Then, 10% should go toward aspirational AI, which is not yet proven but presents a great learning opportunity.
It’s a way of thinking about this as a portfolio of bets.
It’s no different from how I design my investment portfolio. Seventy percent is in proven, pressure-tested public equities. Twenty percent is in private equity vehicles where enough proof has emerged that they can produce good results. Ten percent is what I consider more speculative, although I don’t know whether that is the right word. For the amount of risk I’m taking, the returns could be disproportionate.
The 70% comes naturally to many companies. It involves borrowing from others’ playbooks and deploying proven, pressure-tested models.
The remaining 20% can be explored through your own AI labs, AI centers of excellence and similar efforts focused on emerging forms of AI.
When it comes to aspirational AI, I think there are two areas where university-industry collaboration can play a critical role.
The first is aspirational AI, where you want to be closer to the leading edge. Fantastic work is happening at universities, including the University of Wisconsin–Madison. There is an opportunity for collaboration between universities, academia and industry.
The other area is represented by sessions like today’s. These are people who are leaders in their supply chain organizations and are trying to drive change. Bringing people like that together, focusing on the 70% and 20%, and sharing what they have learned from one another is another area where university-industry collaboration plays a big role.
When it comes to the 10% of aspirational AI, there is a real opportunity to engage in a deeper, almost R&D-style collaboration.
Doug Barton:
You also get a first-mover advantage by engaging in that type of collaboration.
Unlike other technologies, there are companies that have made moving second an advantage by letting other people lead the way. Sometimes they say the first ones through the door get shot.
But this world is a little different because there are supernormal returns to learning.
Madhav Durbha:
That’s correct.
Doug Barton:
There are significant benefits to learning and adjusting. It represents a different kind of strategy.
Madhav Durbha:
There are disproportionate returns, or disproportionate risk-adjusted returns, if you will.
Doug Barton:
I love that language because we do have to think about risk-adjusted expectations.
Let’s wrap up. We’re not far from the airport, by the way.
I’m excited about the point you made about connecting professionals in the supply chain field to help transfer knowledge about the 70% and 20% horizons.
What was your sense of the room? You got to present with one of my favorite member companies, Mercury Marine and Brunswick. They talked about their AI progress and roadmap, which I thought was very interesting.
The room really engaged with them about the tension between top-down goals and bottom-up efforts. Give us your read, since you travel the world and participate in similar conversations.
Madhav Durbha:
A couple of things.
Mercury Marine’s story was fantastic and very grounded. John did a terrific job articulating the journey, but he also made clear that it’s not all rosy.
The journey will have its own speed bumps, and you need to work through them. It’s no different from any other technological transformation, except that it is happening at a much faster clock speed.
Another takeaway is that there are CNBC headlines about how much money is being spent on AI and how AI is going to change the world.
Doug Barton:
I think your slide said $400 billion in capital investments in 2025.
Madhav Durbha:
Right.
There is a gap emerging between deep-tech companies and the way traditional enterprises are approaching AI. Deep-tech companies are running four-minute miles. You’re an athlete, so you can relate to this.
A lot of enterprises I talk to are still running 12-minute miles. There is a fear of missing out setting in. They’re saying, “We’re toast if we don’t do this AI stuff.”
I think a dose of pragmatism needs to be injected into the conversation and into the gap between where deep tech is and where many traditional enterprises are stuck.
Doug Barton:
I think they call that a capability overhang. Capabilities have exceeded the ability to translate them into practice today.
Madhav Durbha:
That’s correct, but that is also an opportunity.
There is a tremendous opportunity to arbitrage the gap between the two.
You can’t take someone running a 12-minute mile and immediately make them a four-minute-mile athlete.
Doug Barton:
No.
Madhav Durbha:
You just need to improve.
When a tiger is chasing you, you don’t need to be faster than the tiger. You need to be faster than the person next to you.
If I can progress from a 12-minute mile to a nine-minute mile, that is measurable progress.
It’s easy to lose sight of that and get caught up in the hype. The reality is that we need to focus on the tangible business problems we need to solve.
We need to deliver better service levels to our customers. We need to cut waste from our supply chains. We need to optimize our inventory levels.
These fundamental problems haven’t changed.
How can we apply AI as another tool, or as the means to an end, so that we can move the needle for the business? Ultimately, that’s where the prize is.
Doug Barton:
That’s well said. I’m not going to add much to that.
We’re close to the end of our time together.
Madhav Durbha:
And the end of my free ride to the airport.
Doug Barton:
Thank you for that.
It was free, but invaluable, as they say.
That’s all we have, friends. My thanks and admiration go to Madhav Durbha for once again being with us here in Madison.
Until next time, Madhav, and to all of you, On, Wisconsin.
Madhav Durbha:
I always enjoy these conversations, Doug. Thank you.
UWEBC Podcast Close
If you want to keep the conversation going, check out the UWEBC website for more events, resources, and details on what’s coming next. Thanks for listening, and from the UWEBC, we’re glad you joined us.