Good evening. Thanks for joining us at the Morgan Stanley TMT Conference. My name is Guru Gupta. I am the Global Head of Transportation and Logistics, Investment Banking Practice for Morgan Stanley. Today we have joined here by the entire crew from C.H. Robinson. We have Dave Bozeman, who's the CEO, followed by Damon Lee, who's the CFO, Arun Rajan, who is the Chief Strategy and Innovation Officer. First of all, thank you for joining us.
Yeah. Thanks for having us.
C.H. Robinson operates in the physical world, not the digital world. You're very well known in the industrial circles in transportation. For the audience here, maybe start with just describing what C.H. Robinson does and where do you fit within the logistics and supply chain ecosystem?
Yeah. Thanks, Guru, and thanks for having us. Glad to be here. At C.H. Robinson, we are one of the largest global logistics platform, and we're actually the largest in North America. The best way to think about it is we sit in the center of a really complex, fragmented market in which you have capacity on one side, that would be our carriers, and on the other side, you have shippers or customers. We connect those shippers with that capacity, and our job is to make simplicity out of the complex. It's a very complex market to do that. We do 37 million shipments annually. We have 75,000 customers, relationships with over 450,000 carriers.
All of that complexity, we feel we are doing a really nice job now, with our technology, with our operating model, and with our logisticians. We call it Lean AI, and we feel like we're the number one driver of Lean AI, supply chains. We'll get into it a lot more, but at the end of the day, a really, really complex way of we move the goods that really power the world.
That's great. It sounds like, Dave, the easy way to think about this is like a two-sided marketplace.
Right
...where you operate. Just help us help the audience here understand, as you think about your strategy and long-term thesis, you mentioned Lean AI. How specifically you are using that in your business model, and why is it such an interesting story within the broader transportation and logistics industry?
Yeah. We'll all jump in here. It's an interesting story for a couple of reasons. We're a 120-year company. And we've moved freight around for a long time. Now, I would say that we're somewhat of a startup and a disruptor in doing that. How we do that and why we're so excited about it is when you apply a lean operating model and lean are principles that have been around a long time, part of my background that I've brought, some of what Damon has brought.
When you connect that with industry-leading technology, and we'll get into that, Generative AI technology, agentic technology, and the experience of our people, those three things we think have driven a competitive moat that has delivered what we think are demonstrable results, but more importantly, very, very hard to replicate in our industry. I'll have you guys like jump in on it.
Yeah. I think certainly we would give equal, but we get asked a lot how much of our performance the last two years is attributed to technology versus the lean operating model. The true answer is we don't know, right? I mean, those two are so combined, so symbiotic, they both work together. What we will say is they both have been material impacts and demonstrable to our, to our results. If you think about what does lean bring, lean on its own brings, operating discipline, it brings rigor and problem-solving. It just brings a pace in which you run the organization. You just think of the structure and business model in which you run an organization, that is what we use our lean operating model for.
Obviously, the technology has brought, you know, tremendous performance benefits to Robinson. I think what gets the number one bill in is always productivity. I think people wrongly assume productivity is the only benefit generated from AI. In Robinson case, that's only one benefit. Now, certainly, you know, AI has been a great benefactor. We've been a bit of a big factor of AI from AI. 40% productivity since the end of 2022 across the enterprise. I think what's more exciting for us or as exciting is AI has unlocked revenue opportunities for us. It has been a key contributor to us outgrowing the end markets for over 10 quarters now. AI has helped us use our industry-leading data set to drive revenue management capabilities.
Think of that as price optimization and cost of hire optimization, so how we procure freight. It's allowed us to use our data set to drive demonstrable benefits in revenue management. You couple those two items with, again, 40% productivity generated across the enterprise since the end of 2022, there's rarely an area of our P&L that we can't point to demonstrable benefits from AI. Those benefits have been supercharged when you combine them with the continuous improvement mindset from a lean operating model. We feel like the combination of lean principles, a lean operating model with cutting-edge technology backed by AI and now Agentic AI. We believe that's a recipe for success.
Yeah.
That, you know, we believe we're one of the few companies in the world that are demonstrating that combined capability today.
Maybe I can just sort of wrap that up and say, well, why does this all matter, right? Ultimately, we're building a scalable business model, right? Meaning we're decoupled from every marginal piece of volume that we process through our system has to have very little human or technology cost, right? Think of sort of the Amazon-like model, right? That's what we're building. A mousetrap that is so efficient that ultimately customers have to ask, well, our shipper customers have to ask, "Well, if C.H. Robinson can run a platform more efficiently to move freight and also do it at higher service levels, then why don't I just outsource my entire logistics department to C.H. Robinson?" Customers already do that. That flywheel will just start to accelerate as we continue to build what we're building.
Yep.
As you build this sustainable model, since last two days we've been hearing about trillions of dollars of investment in AI infrastructure, how are you investing in that?
Yeah. I start. You know, ultimately we sit at the application layer, right? LLMs are effectively a commodity to us. The way we use the LLMs is, you know, we're effectively, we go to Azure AI Foundry, we use whatever model makes sense for the problem we're looking to solve, right? Ultimately, we're this application layer. What we care about is token costs, we pick the right model based on price performance ratio. Ultimately, everything we're doing with AI, you know, I think Damon mentioned this earlier, whether it's AI or prior versions of technology, we're always working backwards from our strategy and our financials, right? Ultimately, we're saying, "Well, we're gonna create this scalable business model." It means we have to have operating leverage.
It means we have to have gross margins that are really strong, right? Within that context, what we say is, "Well, any technology investment we make in AI drives one of those metrics." An example might be, on gross margins, we have to say, well, in this industry, price discovery and cost discovery is not as sophisticated as it is in other industries. However, we have the data to say, "Okay, on the price side of the gross margin equation, we're able to discover prices that we can charge for the services we deliver," and equally on the procurement side of it, because it's a two-sided marketplace, we can procure or we can discover costs that are the most appropriate based on our algorithms, right? Because we have this granular data that allows us to do that on both sides of the equation.
Ultimately, every AI investment kind of ladders back to our strategy.
Guru, that's why actually we're so excited to be here because, you know, you can go from an industrial conference and really talk about at an industrial conference kind of where we are, as you know. Come to one of the preeminent technology conferences and you say, "Why would an industrial, 120 year industrial be here at a technology conference?" Because of what you're hearing here. At the end of the day, we are the beneficiaries. There's a lot of conversations to say, what's happening upstream with AI? Who benefits from that? We are benefiting from that, period. It goes to the bottom line. We like to point to the bottom line and say, "There's no asterisks. Everything adds up.
The benefits are there." We just couch that by saying it's our technology with the way we've transformed the company and our operating model, and our logisticians who we think are the best in the industry. That technology, that's why we're here, because this is the example of how you can get benefits, right from the bottom line.
Yeah. I'll just put a bow on what both my colleagues have just said here, which is, if you think about the AI ecosystem, right? We are in the sweet spot of that ecosystem, right? We're in the application layer. We are the company that is benefiting from the hundreds of billions or trillions of dollars that are gonna be spent from the hyperscalers on driving better problem-solving and capability at a lower cost, right? At the end of the day, we reap the benefit of all of that investment, right? When the question comes, where is the ROI for all the hyperscaler benefits, the answer is, it's at C.H. Robinson, right? We're benefiting on revenue growth, we're benefiting on gross margin expansion, we're benefiting on productivity benefits.
The other thing is, when you talk about investment, Guru, our investment has been contained within our spending. If you look at our spending, since we've started investing in AI, our total spending has not gone up, right? We have been able to, you know, contain the entire investment in our AI infrastructure within our current levels of spending. We're, we're deploying AI in a very efficient way. As Arun mentioned earlier, you know, once we've built an agent and we build our own tech, hopefully we get to get a chance to talk about that in future questions. We're a builder of our technology. We don't buy our technology. Once we build an agent, the marginal cost of that agent is close to zero, right?
It's essentially just the cost of tokens, which are coming down every single year at an exponential rate, right? Versus somebody that's trying to buy AI solutions off the shelf or use a third-party vendor where you're paying by the drink. Not only have we had demonstrable benefits from our adoption of AI, the way in which we build our technology, the which way we deploy it, we believe we have industry-leading cost advantage as well.
On the engineering side, we're not investing more in engineering because our engineers are more productive and, you know, we might have 500 engineers, but they punch like thousands of engineers based on.
Exactly, yeah.
what they can do.
Yep.
Damon, you made a great point. What's your approach towards hyperscalers and LLM models? How do you pick which one to go with?
Yeah. I'll give an unsophisticated answer, and then I'll let Arun give the sophisticated answer. You know, we're architected where we can use any LLM, right? We're gonna go into the details here, but essentially, you know, we're not, we're not held to any given LLM, right? Whichever LLM gives us the best performance versus cost ratio, we can use that capability within Robinson. Our architecture is flexible enough to allow any LLM to be utilized. I don't know if you wanna talk-
Yeah. All of our applications are abstracted from the underlying LLMs, which means we can switch any application or any agent can call any LLM. We have the observability and sort of the test harnesses built in. Where, you know, a given team, if they think that their costs are getting out of control, they'll simply switch to a different LLM, or they'll switch to an older version of an LLM, basically pick the best LLM for the price performance that we expect, out of that investment.
Yeah. It's important to note that for the compilations that we're doing for the, for the application of AI that we're using, in most cases, we do not have to have the latest LLM model to get the optimal benefits, right? We can get the cost advantage of using an LLM that's a generation or two old, still does what we need it to do, but yet we get the cost benefit to using a second or third generation.
I always joke around, Arun's doing way cool stuff. The compute is pretty sophisticated, but it's not like we're doing genealogy. He's able to use, you know, some of the models that may not have to be the latest ones.
Well, I think maybe, like, there's one other relevant point, though, right?
Yep.
If you don't get your context engineering right, those costs could become runaway.
Right.
Also you're gonna be subject to hallucinations. The context engineering is set up. We engineer the context such that a given agent does, you know, if it knows what task it's doing and the guardrails and the context that it's provided is pretty clear. Because of which, A, it doesn't hallucinate, B, you got contained co-token costs.
Arun, you wanna explain, what's been our usage, increase as, compared to cost?
Over the past year, You know, obviously, when we first started out, we hadn't figured out how to tune things and optimize. Over the past year, our token usage is up 85x and our cost is up 1.5x.
Yeah.
You got orders of magnitude of leverage, as a result.
That's a phenomenal efficiency of being able to generate. Help me understand. You're using all LLMs, you definitely have phenomenal cost advantage. Why is it that your competitors are not able to do it? What are the modes around C.H. Robinson? You operate in a highly competitive industry.
Yeah. It's a couple things, Guru, on that. I think this room will hear more and more about this, I think, in the coming year as well. I think Satya talks about it a little bit as well. It's about driving continuous improvement. We call it Lean AI. you know, having that continuous improvement element to get the benefits out of AI is gonna be super important and something we obviously are doing within our business. There's other things as well. The dataset that we have is the largest in the industry, that is super hard to replicate. You can't buy that data, that's a moat to have to deal with.
How we go about being an internal builder versus buyer, that just all drives to replication, and it's really hard to replicate a number of the things that we are doing at Robinson. Our team kinda calculated that it would take about for you to partner with about 15 to 20 different companies to kinda replicate what we're doing, and you have to have orchestrators and on top of that. What does that do? You could do it, but it's gonna be a cost pressure. For the one thing that I value a lot is speed and velocity. You know, we get to do speed and velocity experiments all the time because we build every day, and we can do those micro experiments. We can learn, we can innovate, we can drive.
That's super hard to replicate when you're just going out and buying. We didn't invent this. We didn't invent AI. We didn't invent lean, but we certainly are executing to it.
I'll add to what Dave said on Lean AI. You know, the AI, the technology, the base technology and infrastructure, we've built that out, right? Dave talked about data, but the other type of data is the context that sits in our people's heads, in SOPs and everything else, right? If you think about it, we've got a transportation management system, which is our system of record. That's our base layer. Then you've got humans interacting with our transportation management system. Those humans are doing things on the UI. You know, they might be reading emails, they might be going to a third party, like one of our partner's websites, and they're either, you know, they're off system and on system, just like in any other company, right? People interact with their system of record in some way.
This, the interaction is based on SOPs and other tribal knowledge. The data that Dave's talking about, there's some data that's captured in our transactions, but there's other data that's in their heads, which we effectively harvest and put into what we call the context layer, right? Our context layer then has the collective intelligence of our people and the different things that they do for different reasons, which renders sort of the UI irrelevant. Now you've got this context layer, and you build AI agents on top of that context layer to go do the work on their behalf. That, that doesn't happen. The change management around that is, like, non-trivial, right? What I just described means
The jobs of people who did operations now is to effectively manage context, not to run operations, right? Manage context so the AI agent can do the work. This is where the operating model is a huge deal, right? In terms of like how we drive this change to the organization, that's when sort of Lean AI comes together.
Yeah. I'll just do one final double click on Lean AI because, you know, again, it's probably a new term for you guys. It's one we've coined is it's important because I think the companies that are gonna be successful in getting value out of AI. Today, why so many surveys where you survey CEOs and CFOs, and they say, "Look, we're investing in AI. We haven't really seen any tangible benefit from that example or that investment yet." I would argue it's because they don't have a delivery mechanism, right? What Lean does is Lean allows you to categorize all of your opportunities within the company.
It allows you to understand what is my cost-benefit ratio for automating those opportunities, and it gives you a delivery mechanism, which is what problem am I trying to solve with data, with redundancy, right? Every time we look to invest $1 of AI investment at C.H. Robinson, we are running it through the discipline and the rigor of the lean operating model, right? We are always going back to what problem are we trying to solve? How are we gonna solve it, and is this the biggest problem we're trying to solve for the organization, right? That allows us to make sure our investment dollars related to AI are going to the biggest opportunities within the company.
Last comment on this is, I think where many companies err on AI investment and developing agents is they develop an agent for a task, right? They don't develop an agent for a workflow. Well, at the end of the day, you only get productivity if you optimize the workflow. Simple example of that would be if you've got a workflow that has five steps and you only point an agent to automate or optimize one step, all you've done is created a bottleneck in that workflow, right? You have to optimize the entire workflow to get productivity. When we deploy AI and invest in AI, we are optimizing workflows, which is why we can point to the P&L and say, "We've generated 40% productivity since the end of 2022," and it holds water, right? It shows up in our earnings, right?
You can see it in the quality of our earnings. Whereas many companies today that are invested in AI, they really struggle trying to find the benefit. I think that scenario I just walked you through is one of the reasons why and why we believe companies that partner lean with AI will be the winners at the end.
Where Damon's going is, you know, at the end of the day, we were very deliberate, not haphazard, but very deliberate about going after the order-to-cash process, because in our industry and within our company, that's a, that's a process that has a lot of friction, and it's a workflow. It boded well for our technology to really go in and automate a lot of those steps in the order-to-cash process, which benefit or drove a lot of that 40% productivity.
We've talked a lot about the efficiency and productivity gains, and freight markets have been in recession for four years.
Yes.
As we start inflecting, which is what people are now beginning to see early signs, how has AI helped you on the top line in growing your revenue, and what excites you for the next couple of years?
Well, what excites me is when we get this question, we get super excited because one might look at. I would argue that we've been in a four-year freight recession. If competition was going to do something, you should be doing it now.
Yeah.
I mean, if things are gonna inflect, is that when you feel like you're going to catch up? For Robinson, it's not gonna be, this kinda linear improvement. I think the industry will see a linear improvement as we start to inflect. I think we're gonna have an exponential curve, and we're gonna have an exponential curve because, the way we've engineered the company and re-engineered the company with this technology, it's only gonna get fed more. As the market takes off, we always give an example, we do 600,000 quotes, and we've automated 600,000 quotes. You could add a zero to that and make it 6 million, and it doesn't matter.
The system that agent's going to process 6 million, and we don't have, you know, the human hours to have to bring back to do that because we fundamentally changed that. We're excited about an inflection. When it happens, you're gonna see more of an exponential approach for C.H. Robinson because of the way our agents are built and the way we've done our technology and the way we operate the company that gives us that visibility. That's ultimately a competitive advantage.
Yeah. I think, Guru, what this industry's been trained to do is, you know, very dependent on the cycle, right? When you go into a freight recession, you lay off a ton of people, right? The market improves, you hire a ton of people, right? That is not the Robinson model going forward, right? The processes that we've automated, they're fundamentally changed, right? When you fundamentally change the process that is now over-indexed to technology, has very human light touches versus heavy touches previously, when volume returns to the system, there's no need to add headcount back, right? The process doesn't require headcount the way it required it before, right? For us, when that volume returns to the system, the operating leverage that we're gonna see is gonna be substantial, right? In fact, we've made the comment.
We believe our operating leverage at Robinson will rival the asset players in our industry because we know the incremental cost we're gonna have to add back related to the incremental volume will be very immaterial. Therefore, the operating margins will be great, the productivity will be great, or the cost avoidance will great based on the nature of that curve. You ask how AI has benefited our revenue. Certainly that's how it would benefit it at an up cycle. How has it benefited our revenue today? We give one example that I think is pretty impactful. You know, we have one orchestrating agent that manages our quote cycle, right? Customers send in quotes, we respond to those quotes.
Historically, when that was a human-led process, right, we only got to 60%-65% of the transactional quotes that came in. Okay? Think about that. A third of the transactional quotes that came to us, we either didn't respond or we responded too late to have access to that freight. Today, with our agent, we get to 100% of those requests for freight quotes, right? In addition to that, at no fault of their own, the human before took 17-20 minutes to respond to those requests. When they did respond, you know, even though we have this large data set, they only had time to grab five, 10 pieces of data, analyze it, send a request back to the customer, and the nature of that submission back was fairly unsophisticated.
Today, our agent, again, getting to 100% of those requests, can reduce that cycle time from 20 minutes to 31 seconds, and our win rate's actually gone up because the sophistication level in which we respond to the customer has gone up exponentially. The agent grabs 1,000 data points, sends a very sophisticated response back to the, to the customer. Our win rate goes up, our margins expand, right? That's just revenue and gross margins. As I mentioned at the beginning, you know, that process has been key in aiding to the 40% productivity that we've delivered since the end of 2022. That one orchestrating agent for our quote cycle benefited revenue growth, benefited margin expansion, benefited productivity, and in the eyes of the customer service went up, right?
Now they're no longer frustrated that Robinson only responds to 65% of the quotes on the transactional side. They're no longer frustrated that we sent an unsophisticated quote back to them versus a sophisticated quote. That one agent benefits the customer and benefits Robinson on three levels.
There's several. Go ahead.
Guru, I think if you wanna take that one example, you ask what makes it exciting? Take that one example and kind of blow it out to the whole industry and every single process in the industry. What's exciting is the platform that we've been building and which is now accelerating with AI, is that it is the best mousetrap to operate a logistics business, right? Meaning it's like the lowest cost to serve at the highest levels, which then what makes it exciting is, well, every customer should be able to say, "Well, is it more cost efficient for me to run my own logistics and transportation department?
Right.
Or should I use C.H. Robinson?" It's the same way when Dave and I were at Amazon, the same thing, right? Do I run my own data center or do I let AWS manage my data center? Do I run my own fulfillment business or do I let Amazon run my fulfillment business? This industry hasn't had a platform that has sort of changed the game in terms of cost profile and service profile.
That's how we look at it. It's a little bit different, right? We don't look at it as just being a global forwarder or a freight broker. We're a solutions provider.
Mm-hmm.
As we use this technology to help us go up the value stack, you will see that and you'll see that at scale and at speed. If you go on Robinson, you talk about six months now. Six months is ancient, right? I mean, it says, you know, in six months we're gonna create something that we don't know today what it is. It'll be something new.
Yeah.
It's just a different way of operating, for the room. Go ahead.
As you have automated most of your processes and these agents are making decisions, how do you ensure that human stays in the loop and things are being done the right way?
Yeah.
Because there are not many companies who actually have done this type of automation so far at this massive scale.
Yeah.
I'll let Arun talk on it.
Yeah.
I just wanna clarify one thing.
One correction.
Yeah, one correction. We have not automated nearly as many processes. Like, we're just getting started. This is second inning stuff.
Okay.
We're really happy about our performance and bottom line benefits, but this is early. You have a lot of processes to go, and that's what we feel really excited about.
Yeah.
Okay. This is a really great question, right? We've been doing sort of traditional machine learning for a long time, where we're collecting data and we look at the outcome compared to what we predicted, and the machine keeps getting better, right? Machine learning. You say, well, okay, if you're gonna roll out an AI agent, this AI agent first has to be trained on the context that the human, like, is executing today. The way it works is for any given process that we apply Agentic AI to, a human is involved in training the agent on said task or workflow, right? They remain in the loop for a period of time. There's two things we do.
We say, well, the agent, we back test the agent against all previous transactions and workflows to give the human in the loop some confidence, right? After that, they stay in the loop for whatever duration they choose to, till they're comfortable with going hands off the wheel, right? Before they go hands off the wheel, they usually are on the hook to say, well, what kind of observability and monitoring and alarming they want. You know, like that quote example, right? The quote example has a lot of observability around it because, like, it doesn't. There's bands outside of which it can't go in terms of win rates or profitability and at a granular level, at a lane level, at a customer level. You start to get alarms when something goes wrong.
It's just basic engineering and software culture combined with the lean culture, right? Ultimately, engineering the context, engineering it correctly with a human in the loop, and having the observability to actually run it at scale are basic engineering and lean disciplines which are required to get this out.
I'll just put context around where Dave was going on early innings, which is, as we mentioned earlier, right, the universe of opportunity is our quote-to-cash cycle, right? That's made up of thousands of processes, right? In our NAS business, where our technology has been over-indexed thus far, we've only automated a fraction of those processes, right? When we talk about early innings, it's because we have this universe of processes that have yet to be automated by our, by our technology, and that's the NAS business. On our forwarding business, we chose to start our technology deployment on the NAS side, right? That was the biggest opportunity, right? We are just now starting to index that tech stack over to our forwarding business, and we'll start seeing those results in the second half of 2026, right?
When we think about where are we at on this transformation journey, early innings from a lean perspective, early innings from a technology perspective, we say often, we think the next two years for Robinson will be really exciting, and the last two years have been quite good.
Mm-hmm.
We have 3 minutes left. Maybe we take questions from the audience here.
Sure, sure. Yes, ma'am.
Yeah, please.
As I was hearing you talk about the way the agents are trained by a human, I was wondering what happens to that human after the agent is trained and how you see your head count developing, you know, going forward?
Yep. Well, I'll start.
Then I'll jump in.
To the way we've been very transparent with our people about. There are, first of all, these are mostly operational jobs where turnover is really high, right. That's one. We've been very clear that the future job isn't actually to run the operation. The future job is to manage the SOPs and manage the context for the AI agent, right. It's such that then people have time to solve more difficult problems for customers, right. 'Cause ultimately, customers are navigating more and more complex supply chain problems, and they're asking our people to solve higher order problems. They don't want to pay us. The whole industry has this race around operations. That's how we see it. Which is, there are two things that our people can do.
A, you can manage the context for an AI agent to do operations work. B, you can move into higher order work to solve more difficult problems for customers.
Yeah. I, and I'm glad you asked that question, too, because as a company, this has been super important to me as far as head count and how that goes. First, we don't even look at head count. We don't have a KPI on head count. We just don't look at it that way. Being super transparent with people was really important for us as we went on this transformation. We have about an 11%-14% turnover rate in some of these jobs. That's the industry. That's just what happens on this.
It's allowed us to be able to not backfill that, but also tell them, as Arun said, this is what the future is gonna look like as we start moving to more customer-facing, more verticals, and we start investing in different ways of work as we go forward. We've been very, very upfront, and I think our teammates have taken that very well.
This is a question.
Yeah. I guess the obvious question from, is the extent to which you think you can keep these benefits over the longer term and whether you think that actually scale will actually increase or could increase the moat. Because a lot of what AI is doing is democratizing. It's making it easier for, you know, anyone to code and all the rest of it. I wonder whether you think that the normal laws of economics and these will flow, the benefits flow back to the customer might not apply, and it might actually be that the bigger players such as you guys get to keep those benefits as a long term or not, really.
Yeah. We do believe we will continue to accrue and retain those benefits, right? As we mentioned, the journey we've been on the last two years has allowed us to have a pretty sizable cost to serve advantage versus the industry, right? Our strategy will continue to expand that cost to serve advantage versus the industry, right? There is no mechanics where we see that degrading, right? You mentioned AI being the great democratizer. We would argue, you know, we're already democratizing with AI within the industry already, and we're doing it in a very sophisticated way, meaning custom agents for custom solutioning versus generic agents for generic solutions, right? If we weren't using AI, I would argue that would be a huge risk for our business.
The fact that we're already disrupting and creating that cost to serve advantage with AI in a very custom manner, I think just continues to expand that moat. We mentioned earlier, you know, you can buy AI, you cannot buy our data, right? Our data is the largest, most granular data set that has been built over decades, right? You can buy data, but it's averages of averages, right? You can't execute the agents the way we're executing them with the context we're executing them with generic off the shelf data, right? We believe the moats that we have in place today will only continue to grow as we move to the right.
Very sustainable.
Just to kind of add to what Damon was saying, you know, I think it's worth saying we're effectively democratizing the cost and service benefits of this platform to the entire industry.
That's right.
That's one way to look at it. We've applied AI to make this platform. The second part of it is the AI is not just technology. It only works with data and context. If you look at the data and context, there are three attributes of those data and context that's relevant. One is the sheer scale, the volume. Two, the scope of that, of that data and context, meaning we see a diversity of customers, modes, services. Finally, I'd say the depth or the granularity of the data, right? We understand fine-grained information about customers, about warehouses, about lanes, about commodities, either in data that we're recording or context that's in humans' heads that we now put in our context.
Yeah. As I mentioned earlier, all that discussion was on the technology, right? We mentioned earlier, I give equal billing to our success to the operating model, right? Again, you can't buy an operating model, right? Nobody's gonna sell you an operating model, right? I mean, that is changing your company's culture at the core, right? That's hard work. Most companies, in fact, lean's been around for decades. There's a reason why most companies haven't adopted lean, right? It's not 'cause they don't think there's a benefit. They just know it's really hard to do, to drive it all the way down to the desk to sustain the benefit. We think you take that delivery mechanism that we call our lean operating model, and you combine that with the industry leading technology and data that we just referenced.
We just think that's a combination that's really difficult to compete with.
Thanks for the question.
Okay. We'll take maybe one last question. Okay.
Hi. You mentioned, when you kind of run this to the nth node, if you have scalable infrastructure at a low cost, high service, that you would become the outsource partner of choice for logistics. Can you kind of walk through where are you in that journey? Like are we there? Are we a couple years away? Like, where are you in capability in being able to offer that scaled solution?
Yep. You want me to?
Yep, go ahead.
I think it depends on the sophistication level of the customer, right? Today, we have the service transportation business, which is masked. A third of our volume for that business already comes from customers where we manage their entire transportation department, right? Meaning we already operate up the value stack for customers who are probably less sophisticated, right? Then you say, well, our, with our AI supercharged managed solution, we keep getting better and better, and you can start to see we already have a few really large customers that are on our platform as well. I'd say like, you know, every year, every month, our win rates on managed solutions go up for that reason, right? Customers start to see us as more sophisticated than what they can probably do internally.
Yeah. I would only add that I think we are the provider, just because of our share today, right? We are the largest provider of managed service solutions, right? We are the largest broker providing solutions to the marketplace today. I think what we're saying is there is no cap on how large we can get with this capability, right? We have this discussion two, three, four years from now. We believe we'll still be the industry leader, just a much more demonstrable industry leader than we are today, right? We think those economies of scale, that cost to serve advantage, that stickiness of moving up the value stack and becoming that integrated supply chain partner with our customers, you know, we really believe Robinson's gonna differentiate itself versus everyone else in the industry.
Okay. Well, with that, we are out of time. Dave, Damon, Arun, thank you so much.
Thank you.
Such a fascinating story.
Thanks, guys.
Yeah. Thank you.
Thank you.
Appreciate everyone.