Building AI That Augments People, Not Replace Them, in Logistics

Andrew Verboncouer: Optum's got a great culture, right? You care about people, and that's pretty evident in what you do. Yeah. You care about optimizing the numbers, the time, the utilization, all the other things. But at the end of the day, it's that balance of both,

Shaman Ahuja: right? So money is really just a, a measure of your success and a measure of the value that you've created.

But it cannot be an objective.

Andrew Verboncouer: Hey everyone. Welcome back to this side up where we dig deep into the minds of founders, builders, and operators, reshaping the future of logistics and supply chain. Today I'm joined by Shaman Ahuja, deputy, CEO at Optum, a company using optimization algorithms and AI to help transportation companies do a lot more with a lot less from airlines to LTL to full truckload.

In this episode, we'll get into Shaman's transition from Wall Street to freight tech. We'll talk about augmenting not replacing humans, and we'll also talk about organizational design, really building products with care. We'll touch on lessons from the airline industry, multi objective optimization, and what it really takes to deliver enterprise grade software that works incredibly well across complex operations.

Let's get into it.

Shaman Ahuja: Optum, for those of you who don't know, is a technology company and we specialize in different modes of transportation. And our main goal is to do more with less. How can we optimize the operations of a rail company, a airline company, a trucking company, um, to better optimize or better utilize their assets through intelligent routing, through intelligent scheduling, and just overall better planning.

And our goal is not to replace humans. It is really to augment humans. How can we take your. I would say entry level planner and get them to operate as, as highly, if not better, than your best planner. And my role, um, my title is Deputy, CEO. Um, so Optum is very much a family business. Optum was founded by my dad, who was a professor, uh, for 20 years teaching theory that ended up becoming the roots, um, of the, I would say, all of the algorithms in our business.

And um, I am also in charge of leading our truckload division. So as I mentioned, we have different optimization modes, specializing different modes of transportation. So I lead our truckload division.

Andrew Verboncouer: We've got to know each other, uh, know each other over the last, I dunno, couple years, two, three years maybe.

I think initially at Insight and, um, always run into each other. There's not a, it's not a, it's a big industry, but small industry events, right. You know, you get to know a lot of people and a lot of faces. So it's been good to, to get to see you over the last couple years and learn more about Optima and what you guys are working on.

So I'm excited to, to really chat and dig deeper into your background and really like. Um, I think with any, with any industry, like part of being a consultant across multiple industries, obviously inter mode, we service logistics, supply chain, but, uh, headway. Our parent company, we do a lot in FinTech, we do a lot in healthcare.

And you get to see these patterns across these industries. So I'm really curious from your background, obviously you, you had a shorter stint, um, than you, than you do at Optum, but you were at Goldman Sachs and you went to Carnegie Mellon. Like maybe talk about your background and interest in finance first, and then we'll talk a little bit more about what Optum's core businesses today.

But, um, where did you start and what were you hoping to do career-wise before you got into, into kind of the family business?

Shaman Ahuja: Yeah. Well, thank you for having me on. Um, and what, yeah, what got me into the space? So growing up, um, I was always obsessed with money, so I always wanted to make money. Um, I would always, I was always curious to see like what, like what roles, what professions.

Um, like people were doing and more importantly, like how much, how much that made. So like literally we'd be at the dining table and I'd go around asking like my dad's, like coworkers or any guests that we were happening or having for dinner, how much they made for like, what was their like annual salary, which I later found out.

Yeah. And everybody's like, here you

Andrew Verboncouer: go.

Shaman Ahuja: Um, most of the time they never really answered because, and I later found out that that is not an appropriate question to ask. Yeah, but I was, I don't know, six, seven years old. So like you can get away with, uh, silly questions like that. Um, but like in addition to just obviously wanting to make money, I also wanted to make an impact.

Um, and I've, I asked my dad, I think I was in high school, um, what area should I go in if I wanted to make an impact, um, and where the world is heading? And he mentioned that. The world is heading to basically utilizing computers to do just about everything. Um, so if I wanted to make an impact, computer science was a great field to get into because that's where all industries were heading.

Um, also little did I know that he had a tech company, um, so like secondary like objective was obviously to kind of better like set me up for success. I would say make me like more impactful, not only for the world, but also the family business that, um, he had just started. Now when, so I did some research, found out that CMU was a great school.

It was all on the East coast. I wouldn't be too far away from the family. Um, I ended up getting in, which was an accomplishment on its own. And then while I would say in my freshman and sophomore year. Uh, like I never lost my passion for, I would say money, but a better way of putting it is, is finance. Uh, I wanted to understand how money moved through these tech companies, um, and then how money enabled these companies to grow and prosper and get to that next level.

Uh, so I ended up, I knew that eventually I would end up joining the family business, but. I wanted to do something on my own right outta school. Um, so I joined the, I would say one of the biggest banks on Wall Street, lived in Manhattan. Um, had the time of my life, and, but it exposed me to like a whole nother world, which was like basically a polar opposite, uh, like tech startup, uh, which is like big bank, like thousands of people just in one building.

My floor alone had 500 traders on it. So it was, I would say it was a very unique experience and quite different from the startup. And Optum at the time probably only had like 50 employees, so, yeah. Um, that I, I would say it's different like finance, working at a bank, or at least like a company that treats other companies more as numbers is very different than working inside of a tech company.

Um, where your main product is not is dollars. Your main product is value. Um, and you, you, you look at things in a very different lens. Um, yeah. Finance tends to view things more as like balance sheets, projections. It's really just, they kind of just pigeonhole a company. Uh. I mean, with all due respect, but they pigeonhole a company and do a set of numbers and KPIs.

When a companies have a lot, I would say, deeper story and like the culture and the leadership, the values don't really come across on an Excel sheet. So I would say I actually ended up learn learning a lot more about business and finance having been on on the other side whenever I joined Optum. And actually had to do a fundraise and put together a pitch deck and craft that story that would be attractive to the other side that I originally started on.

Andrew Verboncouer: Yeah. Yeah, that's right. I mean, there, there's money matters in business, right? You have to pay the bills, you have to make money like you're in, in business to make money. But there's a, there's an old Henry Ford quote when I first started, uh. Getting into, into business, and it was a business that makes nothing but money is a poor business, right?

Like you missed the human element of all of it. And from everything I can see and from meeting you, Chris and other folks from your team, like Optum's got a great culture, right? You care about people and that's pretty evident in what you do. Yeah. You care about optimizing the numbers, the time, the utilization, all the other things.

But at the end of the day, it's that balance of both. It's not just the number

Shaman Ahuja: right? Right. So money is really just a, a measure of your success and a measure of the value that you've created, but it cannot be an objective. Yeah. Um, it's, it's just a yardstick.

Andrew Verboncouer: Yeah, for sure. So let's, I guess, let's take a look at that, right?

Like. You mentioned finance, you know, very heavy numbers, logistics, supply chain, transportation, very heavy in numbers. But like what, what do you see as the similarities between, you know, what you were doing in finance and analyzing that to really analyzing how loads move, how you utilize a fleet, how do you utilize drivers?

Like how are they similar and maybe how are they a little bit different?

Shaman Ahuja: Yeah, I mean, every company has the same like underlying goal, right? So we wanna grow our business. And we wanna be more profitable. The question is, what are your ways of doing it? Yeah. And in finance, at least the area that we worked, which was, um, it was equity, it was basically institutional sales.

So companies like Fidelity or large institutional investors, they would invest in and hold large equity positions like buy a million shares of Apple. Um, as part of some mutual fund. Yeah. Or they would have a large customer and they would say, Hey, trade out of this position, but don't move the market. So for them, the, the tools that they had was trading algorithms.

So how can we intelligently trade based off of where the market is so that I can lock in a certain price. So estimate what price, like I should charge the customer. And then there's have those intelligent trading algorithms to. Exit out of that position, but do it intelligently so that I am basically maximizing the margin for myself.

Those are the tools that I have. And like our division was responsible for coming up with more efficient trading algorithms to increase their profitability and then ultimately satisfy our customers. So in, in all cases, you're gonna have a client, whether you're a back office team, a front office team, whether you're a bank.

Or you're a tech company, there's always somebody that you're gonna be like offering your services to, and you're always gonna be wanna, you're always gonna wanna make sure that they're satisfied and that you're growing and offering your services to more and more customers and also have more and more product offerings.

Um, so I would say like the skills that I've been able to like transfer over directly from finance to Optum. Our Goldman to Optum has really just been that customer focus, right? So my customer at Goldman was sales traders, and I knew that if the sales traders were happy and I really kept them front of mind whenever I was building applications that I would be successful.

The same thing has transferred over to Optum, where my first project was with Southwest Airlines. I got, I'm like, my main users were the planners of the network planning and scheduling department. So if they were happy and I understood their business problem and their challenges, and I built an application that allowed them to overcome those problems, then.

I would be successful and yeah, you solve one problem for one customer, they're gonna give you more and more problems, right? Companies don't want to be in the business of working with hundreds of vendors. Like for sure, if you have a vendor that meets your needs, you're gonna stick with them. You're gonna double down, triple down and give them more and more business.

So what, like, if like, like if it's in that same space, and if like. And one of the benefits of technology is that technology is very versatile and can be applied to lots and lots in different sectors. So like, and our model specifically are even like versatile to apply beyond just a single mode of transportation.

Yeah. Um, because of just how we're able to extract the nuances of the business into some generic model that can be optimized. And then we take that solution and then map it back to the nuances, um, of each particular business. So yeah, I would say maintaining that customer focus, I know it's kind of, it's obvious, but it's also, I would say, very hard to do and to do right to consistently deliver value.

Um, to say that, Hey, we will solve this problem in one month, and to actually solve it in one month is like, is a very difficult task, especially to do consistently. So you can solve it, but if you take a year, there's like loss of time, value, and then reputation and then like it hurts the reputation of the, the project sponsor that you had on the customer side.

So you always want to be making sure that the person who vouched for you on the customer side, like you, your main job is to make them look well by delivering on the promises that they sold to their leadership.

Andrew Verboncouer: For sure. Yeah. I mean, it's always, like you said, hey, it's obvious looking at it like, Hey, care about customers, learn about problems.

I think there's, I mean, especially in this space doesn't sound like ad Optum, but like other big people in transportation, logistics, supply chain, like as they're building teams, you're still seeing so much, so many silos, right? Where you think of like the time it takes to go from a customer insight or like a problem that's very understood from the nuances they have to.

Design a solution that would work to checking with them to then ultimately delivering that to the, in some cases you're talking years. Mm-hmm. If not one year, multiple years. Right. And when you, and this is kind of like, I dunno, companies that we've talked to, maybe companies we've worked with, other companies that like their core business is not a tech company and you need to transfer.

Your knowledge into, um, and all of your power into really transforming, obviously the AI comes into play, um, other things, but truly understanding your customers is really the root of everything. Like everything else is either accelerating or decelerating, you delivering on that. So if it takes you that long to go from insight into, you know, impact on a roadmap like.

You gotta play some catch up big time. Yeah. Um, which I would imagine between Goldman and, and, um, what you're doing now across all the modes, like there's has to be, I mean, and this is me blind to what's happening at Goldman and finance and all the different scenarios. There has to be more variables as you think about.

Transportation in this space, and like each mode has its own quirks and each supplier, each carrier, each shipper, each whatever. Like they have their own unique systems. Like there has to be more variables, I would imagine, in transportation compared to, you know, larger head funds, making transactions or making, you know, moves on a position, those sorts of things.

But I guess what's your, what's your take, you know, from working inside of both?

Shaman Ahuja: Each mode has their own challenges. And it's there its own complexities. And one of the benefits of have working at a company like Optum is that I've been exposed to so many different ones that it's, there's really never a dull moment.

Yeah. I never get bored because there's always another like difficult problem like around the corner, uh, to solve after we've solved the current one. Um, the, I spend most of my career at Optum on the airline side, so I'm relatively new to logistics. Um, I spent about eight, nine years focusing on airlines and then for the past three, four years on, on trucking.

And I feel like every day I'm learning something new. Yeah. Each time I visit some customer, I'm exposed to a different aspect of the problem that we will need to ultimately solve if there they're going to get value like from our solutions. Yeah. And there's, there's also the balancing act of like using that Henry Ford, everybody's heard of that Faster horses.

Andrew Verboncouer: Yep.

Shaman Ahuja: Now like for us implementing like our optimization solutions and augmenting the planner operations is, is difficult because the planners are like, well, this is how I do it. When I hit optimize, it doesn't do it that way, and I needed to do it that way. Well, if you just want system that just tells you exactly how you've done it, then you're gonna.

It's gonna be very difficult to increase your profitability because you're gonna be arriving at that same solution. Yeah. You'll just be arriving at it faster. So you can reduce your planning costs, but you're not necessarily gonna move the needle and Right. Come up with new ways and more efficient ways of like building your network.

So how can you balance that? Hey, I can do it the same way. And yes, if you tweak these things, then you can get that slightly faster, worse, or the same horse. Or like if you configure it this way or if you like release the reins, then you can get a solution that could be like game changing for your operations.

Um, and come up with basically new ways of running your business, considering new variables that planners just can't think of. Yeah. On their own. Because like there's only so many like numbers we can crunch in our head.

Andrew Verboncouer: I mean, and like, like to pull on the horse analogy a little bit, like really putting blinders on.

Like, hey, I know if I make my plans within these certain scenarios, like most of the time it's gonna be correct, right? But with something like Optum that you can remove those blinders and really get that like, peripheral vision into the other factors, like you said, that you can't, um, you just can't take into account.

If you're one person doing it, you know, with a couple spreadsheets or something, um, you just can't account for those, those amount of variables. Um, yeah, as you think about, and there

Shaman Ahuja: really is no true optimal either, like, so that's another thing that people like, don't necessarily understand or value that ultimately it's like, it just comes down to like, what is a priority for you and your business?

Like, how do you like balance a one hour delay to a customer delivery to a five hour delay to getting a driver home? Like, what is the absolute difference between those two? Yeah, I would say there really isn't one. It's just, and it's contextual to the customer shipment you're dealing with and the driver that you're planning on getting home.

Um, right. And how important that time off is. Was it, uh, his wedding, um, is it his kid's baptism or is he just trying to get home to, because he. Seeing the new fantastic Ford movie,

Intermode: right? Today's episode of This Side Up is brought to you by Inter Mode a software consultancy. Doesn't it feel like everyone in logistics is trying to figure out how to design and build great software just to stay competitive, but how do you make it happen if you don't have the right talent on your current team?

Well, that's where Intermod comes in. We're the folks who help both startups and industry giants bring their tech dreams to life. Within logistics and supply chain, we don't just give you a strategy to follow. We execute with you by designing beautiful interfaces, building scalable design systems. And bringing Silicon Valley level engineers to build it right the first time.

Whether you're a new player with a bold idea or an established company looking to modernize your systems, we've got your back. We've actually worked with some names You might recognize Ford, Proctor and Gamble, Trimble breakthrough US Venture and more helping them tackle their biggest tech challenges.

Here's the thing, no matter how big or small your project is. We're all about turning your vision into reality. Wanna see what we can do for your business? Just head over to inter mode is inter mode making ship happen.

Andrew Verboncouer: You mentioned your dad really wrote the foundation of the theory for all the algorithms and things that really drive the business today.

Like how are you weighing maybe traditional algorithms and ML and that versus like some of the LLMs and the ai, like true AI LLMs and stuff like out there today, you guys? Augmenting what you're working on, you know, without revealing the secret sauce. Are you exploring new ways? I, I remember there was a new product you showed, uh, you showed me, uh, at one of the conferences of like, you know, optimizing and using AI and, and some of the other things that were, that were brand new that I think were really groundbreaking.

Like, how are you guys balancing old foundational, maybe not old tech, but foundational tech versus, you know, embracing what's coming down the pipe, um, in the future here.

Shaman Ahuja: Yeah, so I mean, AI is advancing, I would say, every single day. So it's hard for me to like predict. Yeah, I think we're past that point of accurate prediction just by virtue of the exponential growth curve of ai.

Now, for us it's, I mean, we ever, I would say at the beginning of the year we stood up a internal AI council where like, how can we take the latest advancements in ai. And bake it into our solutions. Yeah, and I would say where we're, what we're trying to do is, so we have these very complex models and user adoption, like it basically requires a lot of change management, a lot of high touch where their customer success teams to get planners to understand like not only like what the models are like suggesting as like what is the best course of action.

But why they're suggesting that, yeah, why are they assigning this load to this driver? Why are they saying this driver should take this route for these 10 deliveries? Um, why did the driver not take this route and take this other route? So a lot of like, that requires digging into the data. And so what we're trying to do is how restructure our data so that we can benefit from the value of these.

Like LLMs so that they can kind of like dig through that data and answer that question for the user so that to increase that adoption so that like the learning curve isn't months. It's days where I can say, Hey, why did I give this recommendation? Oh, it's because you have this parameter set to this setting.

If you don't like that, you can change it. Do you want to change it? Right? Like. That is very powerful. We still have our models doing all of the heavy lifting, but now we've created this bridge between our models and the end user that makes adoption easier.

: Yeah.

Shaman Ahuja: There's also, with optimization, we always say garbage in, garbage out.

Now how do you identify that garbage? Right. If you have. 50 different like input files. Each one with thousands or tens of thousands of rows, like one erroneous record could like yield a suboptimal or not desired results. How do I yeahs? Huge

Andrew Verboncouer: outlier.

Shaman Ahuja: Yeah. Yeah. And how do I identify that? Oh, the bad result that you're getting is due to that like faulty input.

So like that is areas where we're relying on machine learning and these LLMs. Um, and having those work alongside our models rather than replace our models. Yeah. The benefits of our discrete optimization is every day you're going to encounter like new problems and every company is going to have different strategic objectives and like different weights of this is important, this is not important.

So ultimately it comes down to understanding what is valuable to that company. Um, and then the new problem that they encounter, coming up with the best solution, factoring in the, like, what is relevant at that point in time based off of the state of the world, as well as what is valuable to that company based off of their strategic objectives.

So that requires a, I would say, uh, a, a discreet optimization model where it's very clear like if these are the inputs, these are the outputs. Yeah. Now. Because you're gonna encounter a lot of new things. If you just train it based off of a machine learning model, which worked off of historical data, which would've encountered different problems, then it's very likely that the model could say, Hey, you should like turn left when you really should be turning right.

Just because like, like it has never like turn right before or encounter this situation where it should be turning. Right. So. Yeah, I, that's good. That's where like, yeah, it's like, do you need to have, what we use LLMs for is really that kind of first level interaction, like maybe a planner that operates, like I view the LLMs as kind of like somewhat entry level, and then our Optum models like a subject matter expert, so the LLM can figure out which subject matter expert to call on what situation and get the result back.

And the LLM can act as that interface layer to make it easier to interact with all of those complicated models. And we find that to be a good approach, uh, which is working out well for us.

Andrew Verboncouer: Yeah, I mean, I, I, I think like for any, any technology, AI or or other, like the black, the black box method just doesn't work, right?

People don't trust it for whatever reason. Like if you just give them a result. Maybe some people won't question it, but most people would be like, well, why is it that I think it should be this other thing? And unless you give them a reason or an insight or some sort of logic based response, like, I've found that as well, like even just building, you know, standard workflow software for people, replacing what they were doing in five tools, how did you get to this result?

Right? Because there's always that, like, that criticism that people have is like, well, I know. How I would get there. I don't know how this got there, so I don't know if I trust it. Um, but that's good that you're, you're kind of mentioning like the pairing of both, which is what I thought just kind of outside looking in is like, you know, there's so many foundational models and experience that you guys have in historicals that obviously that has to play a big role into, you know, how you're, how you're servicing things.

Now and in into the future. Um, do you wanna maybe talk about you, you mentioned you spent the most time at Optum at, uh, in doing airlines. Like maybe, you know, can you walk through a quick example of, you know, what the biggest optimization gain was? You know, when you think about airlines, it doesn't have to be at a specific client or, or thinking like that.

Yeah. Like maybe just a, a scenario.

Shaman Ahuja: Yeah. So the, the problem that we were solving at airlines actually had to do with the. Timing and routing of planes. So there's a department within airlines called network planning and Scheduling, and they generally work about six to nine months ahead of the plane taking off.

Because like whenever you go to book a flight, let's say you're planning a vacation three months from now. Yeah. You want those times to be published and you don't want those flight timings to change because if they change now, your connection is no longer valid. And yeah. Yeah. That's fun for nobody and like we've all faced that.

Um, and it's, it's like not fun. So the goal is you want to one, predict the demand that's going to take place nine months from now. So, and based off of that prediction of demand, you need to basically like allocate your assets, right? You need to come up with what is the optimal timing of each of my flights, considering that.

I only have, let's say, 300 planes. And knowing that each plane, once it like operates one flight, it takes 30 to 45 minutes or an hour to turn that plane around, refuel it, deplaning the passengers before it can service the next flight. Yeah. And then obviously if a plane arrived in. Like city B, then the next flight has to be outta city B, going to C, D, C, or maybe back to city A and so on.

So you need to have continuity of planes, need to make sure that you're considering all of the operational requirements of the fleet, and your goal is to maximize the passengers on each one of your flights. So you're not making any money if the plane isn't flying and you're not making any money if the seat on your plane is empty.

Like notice like the similarity between that and a truck where you're not making any money if your truck is just parked and you're not making any money if your truck is moving air rather than actual shipments. Yeah. So this, this concept of matching capacity to demand or considering the operational requirements of your assets.

Um, and this, this concept of like, through intelligent planning, through smarter decision making, like an 8:00 AM flight versus 9:00 AM flight, like you can unlock more demand because now you're better positioning your assets closer to where the demand is. Yeah. Um, like all of those problems are common.

And one problem that we're solving on our LTL side right now is. Yeah. So whenever, um, trucks arrive at a dock, you'll have packages or shipments that need to move from one trailer to another trailer. Yeah. Now, if those, if you had, let's say 30% of a trailer, um, that needed to be e like, like shipments taken out and moved to another trailer, and if those two trailers were right next to each other, like door one, door two, then you've minimized the distance that is covered at that dock.

And you could increase the service delivery appointments for that batch of 30% of shipments. Yeah.

: Now,

Shaman Ahuja: consider the commonality between that and passengers on a plane that arrive at, let's say, Atlanta, and they're connecting from one gate to another. Uh, let's say 30% of those passengers were all on the same connecting flight, and those two planes arrived at two adjacent gates.

How much better would that be than if they had to go change terminals and take the train to get to the other side of the airport? So, yeah. So there's a lot of overlap between

Andrew Verboncouer: predicted, though forecasted demand of that as well, right? Nine months in the future. Exactly. Which is wild.

Shaman Ahuja: Yeah. And, and demand changes by time of day, right?

Because during the week you're gonna have your business travelers, and then on the weekends you'll have your leisure travelers. So. Like for on airline side, day of week variability is very essential. And whenever they schedule and their data models, it's all kind of like baked into that concept of yeah, what is the typical week or what is your rather, your typical Monday versus Tuesday, Wednesday, Thursday, and then through the week.

And then they repeat that cycle for, let's say that month. And then note that there's month over month variability. So seasonality plays a huge factor when you're coming to scheduling flights. Like considering passenger demand, um, as you can imagine it does in, in shipping as well.

Andrew Verboncouer: Yeah. Yeah. And I would imagine they're probably similar when you think about coastal travel versus connection in Atlanta, connection in Minneapolis, you know, to the other coast.

Like that kind of out outside to middle transportation, um, of people. Mm-hmm. And or goods is probably pretty similar, you know, as you. Transverse a country or you know, kinda beyond, you think about stuff coming from like Tokyo and we've done some global visibility stuff in Marine, like, you know, from far west, um mm-hmm.

All the way, you know, to the east, like a lot of that inbound outbound is, there's a lot of moving parts.

Shaman Ahuja: Absolutely. But one, one good thing about the airline demand is the movement of passengers. Is well balanced, uh, because when you leave, eventually you come back to where you left, right? Yeah. If I'm going on a business trip, I'm likely coming home at the tail end of the week or after a few days.

If I'm going on a trip to Japan for a couple weeks, I'm gonna come back. So like, generally like traffic on airlines tends to be well balanced versus like, freight tends to be more one way. And when freight tends to be more one way, having better routing, um, of, and like better like, like movement of assets.

Uh, like it basically movement of assets is much more like, like valuable problem to solve because like you'll always end up in assets like dumped in these sinkholes, um, that you need to reposition and to do that intelligently. Um. It can be very difficult for planners. Mm-hmm. Especially when they're only considering, um, the next driver, the next load, or just my region.

Yeah. And the truly optimal, um, or a more optimal solution requires a more macro level perspective or longer term perspective than what people are just used to scheduling for.

Andrew Verboncouer: Yeah. I mean, it made me think of while you were saying that of U-Haul, right? Like mm-hmm. Most people renting U-Haul. Some of 'em are inner city, right?

A lot of people are taking one way tracks across the country and not coming back, you know? And so you'll see these, did you guys work with U-Haul by chance?

Shaman Ahuja: Yes, we did. So we actually, we were doing a, a pilot study for the movement of their pods. So like, you know, U-Haul has these pods that they'll come and drop off.

Um, at your doorstep, you load all of your, like whatever your, your, your furniture, couches, all of that. And then somebody will come pick it up. And then deliver the pod to wherever you're moving to Now that there's a load consolidation problem where, how can I efficiently like take all of these pods that are moving through the network and sequence those up into multi pick, multi drop, considering that each pod is of different sizes and let's say like a trailer can only hold like, like.

Eight boxes of size, one or four boxes of size two and so on. And notice that each potted may be picked up in one city and delivered in another city. So how can I consider all of those movements and again, better utilize the space on my trailer, um, so that I'm not leaving like empty space when I could fill it up with a pod that is being picked up or dropped off in a similar area.

Andrew Verboncouer: Yep. Yeah, there's a lot of, a lot of variables, right. Even with the ones and twos. Yeah. Right, because it's mm-hmm. I'm assuming they're not doing as much real time forecasting of purchasers and pickup and that sort of thing, but still, you know, complex, right? Like, you can figure it out, but it, you know, there's, there's a method to the madness.

Um, you said something before, so you, you were in airline, now you're focusing on transportation. It made me think of really just how your teams are structured. Like you, when you think about your product teams that are delivering, you know, strategy, design, engineering, like, are you focused primarily on mold?

Is that how you guys are split up or? Problem. Or do you have like platform teams? Like maybe let's, let's talk a little bit about the nuts and bolts of how you think about organizing your team to go execute on the customer problems.

Shaman Ahuja: Yeah, so right now our teams, so we've had, I would say, mixed success with this concept of platform teams.

We are like, historically, we're like, let's just have these center of excellence teams that build these core modules and then. Um, all of our individual like business units will take those core modules and build on top of that. Yep. And it'll be perfect. Um, that I would say was very difficult because like of the differences amongst each of those businesses.

So we would have this core module and then it would require 50% customization for like industry one, and then, uh, 60% customization for industry two. And if you kind of try to layer all of that into the core module, it ends up being a mess. So we've had the best success by having business focused teams. So we have like four main verticals.

We have rail, um, airlines, LTL, and full truckload. And then we also have kind of like this r and d space, uh, like Optum Labs, uh, which is how we branch into other mos or transportation or other verticals. But those are the four that we focus on now. And then each one of those decomposes to like sub products.

So we could have anywhere from two to five different products servicing that mode, each tackling a different like problem or servicing a different user base, like within that industry. Um, but generally, like on LTL, it's really just. Optimizing the lifecycle of an LTL shipment, considering like the first mile delivery to the last mile, like line haul consolidation, driver scheduling.

But each of these are like, like different aspects of that, like one overarching problem. Um, so we have four or five products within that sector. We'll have product managers that specialize, that cover one or more products, all reporting to a business unit head. Of that division. Yeah. And then we generally have a central engineering head and then different scrum teams that, again, service one or more of those solutions.

And we'll always have, we'll also have an architect and a smaller core team that focuses on reusable components within that business. And then you could have, like, we are now starting to bring that center of excellence and. Reusable components company-wide rather than just within the scope of that business unit.

Andrew Verboncouer: Yeah, yeah, that makes sense. So more like design system team versus, like you said you had a platform team that maybe tried to solve it end to end.

Shaman Ahuja: Mm-hmm. Ui,

Andrew Verboncouer: you know, maybe that was

Shaman Ahuja: isolated from the business. Yeah, yeah. And you realized that it's very difficult to do that. So what we can do is maybe have one person that's responsible for aligning.

The work that's happening across those different businesses. Yeah. And then have like members of each of those different businesses maybe do 20% working with that central person and then 80% working with their respective business. So you wear two hats. One is my Optum hat and then there's my like truckload or a load AI hat.

Andrew Verboncouer: Yep. Yeah, I mean I, I, I think in our experience too, we found that same sort of structure to be. Beneficial because like if you just come out with a, I mean like anybody, right? Like people in, in product teams, they are the customer of whatever platform or reusable components or designs, whatever that makes, looks like if you don't make it with them or bring them along for the ride, like they're gonna be resistant, right?

Because like I always say this like, developers wanna develop, designers wanna design strategists, wanna strategize, right? And if it was up to each of 'em, they'd spend more time doing their own craft, right? But like. The platform team and design system team, whatever, whatever mix you have is like meant to empower that team to deliver faster on the business versus hamstring creativity.

I mean, you're not creating art in these teams, right? You are focused on business software, workflow software, all those things. So, um, that's good to hear about the, the kind of learnings and, you know, the transition into that. I think, um. Yeah. I, I think that's, that's good. Um,

Shaman Ahuja: yeah, I mean something just because something is like cool and reusable doesn't mean it's practically useful for the business.

Yeah. And, and that's where we find that having that business person, like of working alongside that core team, yeah. They can, uh, they can kind of bring down that solution to Earth. So that is actually like useful and applicable to at least like their specific business.

Andrew Verboncouer: Yeah. Yeah, and I think like the focusing on that problem I'm, I'm sure helps you get to, we kind of talked earlier about like in this industry and others, a lot of people miss the care that it takes and the attention it takes to get the nuance of a problem to actually solve it, to not solve it 80% and let them figure out the 20 in a different tool or whatever to actually solve the problem.

It's like you need people that really care about this and are focused on the customer of that and are, you know, generating insights. So I think that, um, that all makes sense, right? Versus a team that's like, we're the mobile team, we're the web team. Like, you'll never get there. You'll have an app, but not a business, right?

Yes,

Shaman Ahuja: exactly.

Andrew Verboncouer: Yeah. Um, that's stuff. Um, so you guys recently, and it takes

Shaman Ahuja: years to like build that like business understanding. So Yeah. Like you think, you know, and then with every year that passes, the more you realize you actually didn't know. Right? So then like, the most, I would say like. The people, the, the most experienced people will actually say things like, oh, I am still learning every day, or, I still don't know.

Because like, there is so much, and that goes into this, um, that if you're gonna build like specialized software, especially enterprise grade, like it's, you're gonna need to like be focused on that area and work at it for a while. Like, yeah, it has taken like five years for load AI to get to where it is today.

And I would say only now we're really like scratching the surface of like our optimization capabilities for like servicing the truckload sector.

Andrew Verboncouer: Yeah. I mean, and you have to care, right? Like mm-hmm. If you just care about your craft of design or development and you don't care about the customer or their problem, like those are never gonna match up.

Right. And, you know, I think that's a big part of any culture building tech is like, you have to build a culture of care for the problem and the customers you're serving. And without that you're, you're always. Waiting on a middleman to tell you what to build. Right. And you're never gonna, you're never gonna get there.

Some cases you might, um, but it's gonna take a while. It's gonna take a, you know, a lot of resources, a lot of time, and some companies, most companies burn out all their runway before they get there. So,

Shaman Ahuja: absolutely. You guys, I think the planners and the users care and like these guys have been working 10, 20, maybe 30 years at the same company.

Yeah, are definitely in the same space. So we find that there's not, not a lot of movement that happens from airline to LTL to full truckload, generally starting one. And you kind of just like become a nerd in that area and then you just don't leave that area and that is your career. So for sure it's, it's hard to just come in as a fresh technologist saying like, oh, I'm going to disrupt and change the way that you've worked for the past 20 years through this cutting edge technology.

Even though I like, I've only been exposed to your field for three months.

Andrew Verboncouer: Yeah.

Shaman Ahuja: Right. Like, they're not gonna listen to you be like, you don't know what you're talking about.

Andrew Verboncouer: Right. It'd be like me saying like, I'm gonna teach you how to become a Michelin star chef by giving you a really sharp knife.

Shaman Ahuja: Yes.

Right. You're like,

Andrew Verboncouer: how do I use it? What do I do? What's the method? Right. Like you, you're just not gonna get there. Which, um. Yeah, I mean, so obviously like digging deeper, I, I see why Optum's been successful. So kudos to you and your dad and the whole team. And you guys just celebrated 25 years, which is no joke in in tech.

You know, freight tech and Optum Tech, like you guys are, are staying on there. Um,

Shaman Ahuja: yeah. Especially got business. So we haven't actually, we haven't raised any venture funding to date. Um. I'm not saying we won't in the near future, uh, there's, I would say it has its ups and downs, right? You don't have that like hockey stick growth curve because you can't just throw a bunch of money into sales and marketing.

Uh, you need to be very tactical about where you allocate your dollars. Um, but it allows you to stay focused on your mission, right? So if, if we believe that this is the right way, then we're not, we don't have to, we're not pressured to do things the wrong way. Um, to appease investors. So yeah, I would say there, it has its ups and downs, but I would say like we're definitely very proud.

Um, we have like a lot of loyalty amongst our teams and a lot of loyalty amongst our customers as well, just because we keep them front of mind in everything that we do.

Andrew Verboncouer: Yeah, I mean, if you put them first and you build the right way, like there's, it's hard to lose. Right. It's hard, it's hard to beat someone who, who cares and doesn't give up.

Right, exactly. Yeah. Yeah. So I guess in closing, like what, um, is there any belief you have about AI or optimization that you think is unique to you or unique to Optum that, you know, other people coming in the space, like, like you said, a technologist coming and saying, oh, I'm gonna build this new, um, you know, this best new tech and you're gonna love it.

It's, everything's optimized with ai. We've seen. You know, all sorts of companies, but is there a viewpoint you have around like how to best utilize it or how to think about implementing AI into your, your company?

Shaman Ahuja: Yeah, so I think for us that, and I think we talked about it already, that multi objective thinking is essential.

Yeah. Um, there really isn't just one right way or one optimal way. Um, and the goal is not perfection. Um, but it's really like informed trade-offs and coming up with these informed trade-offs that align with your strategic objectives for your unique business. Um, so if you think that you can just take AI and just throw it at whatever problem that you have, um, and have it operate in a vacuum and have it add value, it just, it won't work.

So like you cannot replace the planners to date. Um, you need to have, like AI work alongside your business, and there also needs to be an investment in understanding of how AI works with the data that you feed it, because AI will only provide value if it's trained on appropriate data. And when it gives you a wrong answer, it's because you fed it wrong data or there's data that let it down a wrong path.

So having like basically building a strong data understanding amongst your team and like restructuring your data so that AI produces the right answer and having a culture and a mindset that, hey, AI will be wrong and that just because it. It gave me a wrong answer. This one time doesn't mean it can't like add value to my organization and even help me in my operations.

So like even if it's right 80, 90% of the time and wrong 10%, like it could still like end up being great for my business and help me like overcome the challenges that I'm facing in this freight recession. That seems to never end. So like our goal is ultimately like growing our business, becoming more profitable.

And I wholeheartedly believe that AI can do that, but used appropriately and with the right, I would say culture and with the right mindset.

Andrew Verboncouer: Yeah, I think you could go back in that whole. The entire section you just said, and just replace like human intelligence with that is the same thing, right? If you don't allow your team to be involved in customer conversations or to get their hands dirty in problems or to you, like you're not building their context for how they can use their skills, right?

So the, the very same thing is like good and team practices and building, you know, great software or businesses. Implementing that into your own team. And also, you know, like you said, showing that with ai like context and it's gonna be wrong. People are gonna be wrong, AI's gonna be wrong. No one's omniscient, uh, here on earth as far as I'm aware.

Um, if they are, let us know. But yeah, I really appreciate, my dad might be close. Yeah. I appreciate the conversation Chaman, thanks so much for, for sharing your insight. Thanks for joining us on this Side app. Don't forget to subscribe, rate and review our podcast to help others find us. If you have any questions or topics you'd like us to cover, feel free to reach out and if you want to be a guest, let's connect.

Until next time, keep your shipments safe, your logistics smooth, and your curiosity on the move.

‍