Watch Now


C.H. Robinson’s new CEO doubling down on lean management, large language models

Bozeman previously built out Amazon’s middle mile, held executive roles at Ford, Harley-Davidson and Caterpillar

C.H. Robinson's profit for the first nine months of 2023 dropped 85% compared to the same period last year. Can its new CEO turn the company around? (Source: Shutterstock)

When C.H. Robinson (NASDAQ: CHRW) announced its next CEO was Ford executive Dave Bozeman, insiders likely had questions galore. The biggest, perhaps, was: How is he going to fix this company?  

On Monday, during a very rainy day in New York City, I met with Bozeman to answer just that. He spent 45 minutes answering my questions on how he’s going to reinvent the massive freight brokerage C.H. Robinson.

Robinson is in need of a turnaround plan. The freight brokerage failed to fully capitalize on the massive upcycle in freight in 2020 and 2021. Instead, Robinson appeared to just run up its costs. Robinson’s previous CEO, Bob Biesterfeld, had admitted that Robinson hired too much during the freight bubble of the early 2020s. When demand suddenly and precipitously declined in the spring of 2022, Robinson was caught with too many employees and too few loads to move. Analysts have also pointed out that, under Biesterfeld, Robinson invested massively in technology, with results that weren’t commensurate

On Jan. 3, Robinson announced that Biesterfeld had resigned as CEO and from the board — ending a stretch at the company that lasted more than two decades.


The rest of 2023 has proved to be brutal. In the first nine months of 2023, compared to the previous year, income from operations dropped from $421 million to $63 million, an 85% decline.

Robinson’s board deliberated for months, it seems, on who would be the next CEO, as FreightWaves’ Mark Solomon reported. Insiders told Solomon that former UPS COO Jim Barber was a likely pick

So many were likely surprised when Robinson plucked Bozeman instead. He’s an outsider to Robinson and the traditional freight brokerage world alike. However, he has served in executive roles at Ford (NYSE: F), Caterpillar (NYSE: CAT) and Harley-Davidson (NYSE: HOG). Perhaps most notable is his five years at Amazon (NASDAQ: AMZN), where he built the retailer’s middle-mile delivery network. 

Robinson’s stock dropped 6.4% on June 5, when the freight brokerage announced Bozeman as its next CEO. As Susquehanna’s Bascome Majors pointed out, that may be because “a clear ‘slash and burn’ hire would have driven a short-term pop, and Mr. Bozeman’s resume doesn’t fit that narrative (on paper at least).”


While insiders may have questioned Bozeman’s lack of Robinson background, it seems that having something of a “beginner’s mind” to deep-seated issues in brokerage could be more conducive to actually getting those problems solved. As he said at one point in the interview, Robinson’s culture has “gone from admiring problems to solving problems.”

I wanted to know more about Bozeman’s long-term plan to rebuild Robinson. The gist is that the company isn’t shying away from technology investment yet — but it will seek to lower head counts. Still, Bozeman doesn’t think that deep knowledge can be entirely replaced by technology. That would be good news for Wall Street and Robinson’s 15,500 employees alike. 

This interview transcript has been edited for clarity and condensed.

Bozeman is new to Robinson, but he’s not new to transportation – or freight brokerage! 

FREIGHTWAVES: Obviously you’ve had a long career at Amazon, Caterpillar and Harley-Davidson, but you’re new to the brokerage side. What’s the competitive advantage around being in the shipper space for so long and then now coming into the brokerage side?

BOZEMAN: There’s a few things here. Sharpening my knife at Harley-Davidson, it was getting into good Midwestern engineering, manufacturing and honing your skill set around people development, people leadership and then execution. Those were my early days. That played on both the product and operation side, so you had that kind of balance.

Going to Caterpillar, it would’ve been pulling in that global mindset, but also driving sharp financial bottom-line type of execution. I launched the lean execution operational excellence. This would have been starting to hone the skills of the lean talents within Caterpillar. I spent three and a half years in the executive office of Cat. I had the logistics part within Caterpillar as well, so dealing with all of the big steam shippers. I was responsible for that logistics arm at a big level, but as you say from a shipper perspective.

Then going to Amazon was interesting in taking on that mission of building out its middle mile. It was kind of a hybrid of building this asset-light versus asset-medium type of world. The asset-light, where the brokerage part came in, was building out Relay and standing up Amazon Freight, Amazon shipping and the technology that goes with that. Those are really the basic building blocks of connecting freight from origin to destination. That’s what the technology play was on Relay.

The more asset-heavy part is we procured almost 67,000 trailers within Amazon. That’s what you see going on the road today, both power and trailers. Having that understanding of the asset side plus building out the technology on what you would call the asset-light side really started to form my experience around that space. 


Ford was a short stop — getting back into more of the product and global customer service, in which you really start talking about the next level of electrification and things of that nature.

Bring all of that together and you come into Robinson. That culmination of experience really has you familiar with the asset shipper side, but also that asset-light side and understanding. 

Some things don’t change and it’s really about that execution. As I tell the team, it’s inputs and outputs. Understanding those predictable inputs will drive you into predictable outputs. That’s what a lot of my experience has been in driving into that. I think it’s proven really quite convenient coming into Robinson with that.

One big area of Robinson’s comeback is technology, but the story has changed from ‘automated freight brokerage’ to large language models

FREIGHTWAVES: I read in the initial announcement of your appointment to the position that you have a background in “reinventing complex operating models.” What needs to be reimagined or looked at again at C.H. Robinson?

BOZEMAN: It’s a tough market right now, and I would almost say it’s a complex market. Coming off of the pandemic, you have different buying patterns from end users to different shipping patterns. 

With the advent of large language models, how do you reinvent and think about that? We’re actually starting to do that. 

In our world, unstructured data is a big deal. We have 500,000 unstructured data points that come at us daily. You look at that and say, “Hey, what are you going to do about that?” We’re not just talking about large language models, we’re doing them, and I’m a big proponent of driving that. I’m all for technology and really driving things. 

Using this generative AI and large language models, we’re able to now start tapping into those 500,000 unstructured data points. 

Robinson is aiming to quote freight in less than a minute, even if you send an email 

BOZEMAN: We have 100,000 customers. Some of those customers are not going to have the technology needed to plug right into us and have that instant connection. So they send us an email — unstructured data point. Typically, you would have humans having to go and parse that information out, make it correct, so it can plug into our system and be actionable. 

Say we can quote by using this large language model. It’s able to see that email, parse that information out, jump into Navisphere, jump back out and do a quote in less than a minute, which is where our bar is. 

I think you can see in the industry that a total digital solution is not the business model that works here. Freight is hard and it’s hard because of disruptions. [You have to have] people, who have experiences and deep relationships and technology, which is a force multiplier. It allows our people to take away that unstructured model activity so they can just be working on customer-facing strategic solutions. That’s a way of reimagining different things. 

Robinson also wants to deepen its customer relationships and offer more services 

BOZEMAN: I think another example I would give you is just kind of reimagining how we show up as an organization, as a company. Our competitors don’t have all of the elements we have to solve sophisticated logistical solutions for customers.

Global forwarding, customs consolidation and warehousing, surface transportation, taking all of those things and reimagining how we stitch those together for the customer to generate more value, it takes a different conversation than what we’re doing. 

Now, I am pleased that 50% of our revenues as you know comes from our customers who use both Global Forwarding and NAST [North American Surface Transportation] services. That’s great, but I also say that there’s some we’re probably leaving on the table.

As I go out — and I’ve spent this last six months talking to investors and customers and employees — customers will say, “Dave, we love Robinson. We love that you guys work with us and the solutions.” But they’re desiring even more solutions and for things that they don’t even know. We can provide that and I think it’s a competitive advantage that our competitors may have a slice of what we do.

Bozeman is going through all of Robinson’s portfolios … but we had to ask a few follow-ups to nail down the big fixes he’s looking at

FREIGHTWAVES: Are there any sort of departments where you’re looking and thinking, “OK, we need to restructure this from top to bottom”? Or there’s something economically or physically that doesn’t seem to be clicking here?

BOZEMAN: I’m pretty pragmatic when I go through and diagnose across the portfolios and across people, across products and processes. I would say that there’s products and processes that I looked up and said, “We need to redo this.” You have to say why, you don’t just redo it to just redo it. It’s about solving specific problems. 

I have gone in and there are some things that we look at and say, “Hey, those processes are either not solving the problem that they were intended to solve, we’re not fast enough in softening or we’re not inventive enough.” 

We’ve looked at a number of those things and I’m pretty happy that the culture is kind of shifting now. It’s a culture that has kind of gone from admiring problems to solving problems. It goes back to my tech days of identifying problems fast, innovating fast and executing fast — because time is money in our business. 

As you look at the various portfolios, you just go through top to bottom and it’s really just about solving, identifying problems, putting them on the table. Every company has this issue and can get better at it. It’s identifying the root cause of your problems no matter how ugly they are, but putting them on the table and then solving them with the proper debate and the proper innovations. We’re moving to that culture to do that. It’s something that I require. That’s my bar.

Dave Bozeman. (Courtesy of C.H. Robinson)

FREIGHTWAVES: Is there any sort of example portfolio that comes to mind that you’d be able to share where these sorts of problems have been diagnosed and solved or are in the process of being solved?

BOZEMAN: For all of our areas, there’s something to be solved. There’s something in NAST that we have to solve, obviously. Global Forwarding, managed services, Robinson Fresh, there were things in there that we had to solve all the way. In our managed service business is our 4PL. We feel really good about the 4PL space that we’re in. We’re looking at that to see what’s the maximum breadth that we’re getting out of that business and what we’re doing. I love the stickiness of it and what customers desire for.

In NAST, obviously that’s the big part of the company. We’re truckload, LTL, ocean and air. In NAST, there’s a tremendous amount of things that we’re going through and really shoring up, everything from our technology stack and the speed that we use that technology stack on to processes, to technology adoption. Small changes in NAST are big changes for the company. The good part is that people feel really good about it. You could start off and it’s uncomfortable, but uncomfortable turns to excitement when we really see what can change as we’re doing it.

In Global Forwarding, we’re looking at our lanes. It is an ever-changing marketplace right now with nearshoring happening. We have to think about how we adapt and maximize our efforts around the constantly dynamic world economy. There’s a move to get more nearshoring. We have to adjust how we service that for customers and we are servicing that and I think we’re in a position to do that.

Robinson is doubling down on requiring carriers to embrace visibility — but some within the company were hesitant at first

FREIGHTWAVES: Is there any sort of example in NAST you could point out?

BOZEMAN: Visibility is huge for us. Traceability was a big one within NAST, and it was actually one that was one that the company was a little hesitant on. Why? Because in order to have our carriers, and we have a lot of carriers, to do business with us on our platform, and you want visibility, you have to put in rules and you have to be pretty straightforward on those rules. 

The natural nervousness around that is if you do that, you have people fall out because they don’t want to adhere to some of those rules.

Ultimately, I always say, “Build a product that you can be proud of and stand on that product.” We built a product that we could be proud of and then we could stand on our product. We made the call to say, “This is what’s going to be required for you to do business on our platform.” 

Well, listen, the next day the sun rose and the world continued. 

I would venture to say that people like the change and they like doing business with us on that. And what did that do? It gave a tremendous amount of execution on track and trace and visibility, which is obviously something good for customers and what they would desire as far as feature sets.

FREIGHTWAVES: The track and trace question is interesting because I’ve definitely heard that hesitancy from folks on the carrier side as well as on the broker, shipper and tech side. But at a certain point, if people want this technology, something’s got to happen.

BOZEMAN: For us at scale, it’s really, really important. If you’re a customer, we can let you know where your product is, but more importantly the disruptions that may happen. We’re now tracking about 98.2% confidence that we can tell customers if their product is going to be on time, because we could see disruptions.

Let’s say the border is locked down and there’s something going on. If we have to switch to an air mode, we move to an air mode. If we have to switch to a different port, we’ll switch to a different port. Giving that visibility to customers, they’ve told us that’s very, very important and extremely valuable.

Bozeman says a large language model would allow Robinson to worry less about attrition

FREIGHTWAVES: Having this email sent to some sort of large language model, what does that look like? What’s an example situation where maybe that email would get lost or not used and, instead, C.H. Robinson can use that and sort of reimagine some process differently?

BOZEMAN: One of the things I’ve said is that we’re going to have a 15% improvement in productivity this year. We were at 18% at the end of the third quarter. We feel good that we’re going to be at 15% at the end of the year, but we’re also saying we’re going to carry that momentum into 2024. We’re going to do another 15%.

Now, it’s not like the world is getting sunnier out there. We’re fighting through a freight recession and we’re going to do that, but we’re still going to drive another 15% productivity. That gives you a CAGR of 32% in a couple of years and we’ll continue to go from there. 

How are we able to do that? Taking away a lot of that unstructured data work. Now, if a human has to do it, and we’re really good with our people, you’re still going to have mistakes. It is time. You’re going to have people going back and figuring things out.

By having and leaning into large language models and letting it do that work, our people are then able to be freed up and focus on the customer-facing strategic problems. 

More importantly, that allows us to think differently on our attrition. We don’t have to backfill on a number of different things, we can take resources and put them into other areas that are of importance. It just allows the company to be more efficient. You’re seeing that show up in our numbers that we’re putting out there, but that’s a huge part of it.

[NOTE FROM FREIGHTWAVES: Robinson reported a head count of 15,577 as of Sept. 30. That same date last year, Robinson employed 18,045. The biggest reduction in head count came from North American Surface Transportation, or NAST.]

Bozeman breaks down how exactly the large language model works

FREIGHTWAVES: Let’s say I’m a carrier. I email a broker saying I’m delayed because I ran out of hours. Would an email like that be deployed or incorporated into this large language model? 

BOZEMAN: Flip it the other way. First of all, that does happen, but it’s probably going to happen on a chat that comes in. The person on the desk is seeing that chat and they can respond to that chat. It could come in on an email and we could process that, but let me give you a better example. 

When someone comes in and they say, “Hey, I need a load to go from North Carolina to New Mexico,” they send an email because they’re a medium-sized customer and they’re putting an email in that says, “Hey, I’d like a quote on North Carolina to New Mexico.” 

That quote has to have a number of different elements in order to make it a bookable load. You have to have certain numbers in there, things about the carrier. 

When they send an email, before someone would take that email, they may have to print that email out. They’d have to then go look and they’d pull out a highlighter and say, “Let me have to carry a number here. Oh, let me bounce that up against their history and what they did. Oh, go to the process. Let me go in and see what we have available for a quote.” 

The amount of time when you say, “OK, now let’s send you an email back on what that quote is,” versus now doing this in under a minute with a large language model is game changing at our scale.

FREIGHTWAVES: So instead, the large language model would instantly reply saying, “How many years have you been in service? What’s your insurance, this, that or the other?”

BOZEMAN: It’s going to go in even more than that. If you ever did business with us before, it’s going to find it, it’s going to put your numbers in there, it’s going to put in all the background information. It will send you back and say, “Hey, this is the quote for that origin destination.” 

You’ll obviously be able to confirm your key identifying information, and it even has the ability to kind of do that in a conversational manner.

But people are still important!

BOZEMAN: I don’t want you to walk away from it thinking that that’s everything. We have fantastic, awesome people that then use this technology to assist, and I do these things called gembas, which in the lean world means “go see” or “go see the work.” 

I’m looking to see what are the error states? What’s the life of a load? What’s the work that we’re actually doing in detail where I could ask questions?

It was a bit odd that the CEO was sitting at a terminal for two hours with someone. But to show you this example, we had a wonderful employee out in our Chicago area. He had all of his screens, and he was talking to me, “Mr. Bozeman, this is what I’m doing.” He said, “Excuse me, pardon me, I have to book this load.” And he did it very fast and he turned to me and he kept explaining what he was doing. In our conversation 30 seconds later, he said, “Pardon me, I have to remove that load.” I said, “What do you mean? Why did you have to take that off?” He said, “Because it won’t fit.” And I said, “What do you mean it won’t fit?”

He said, “No, it’s a loose load. I know this carrier. I know this load, and as I was thinking about it, it won’t fit. And when they show up, then that’s just going to cause double payment, waste, frustration for the customer.” 

At that moment, I thought about it and I said, “That’s the power of people and technology.” If I had just let an algorithm do that, it would’ve booked that load and it would not have had the experience and understanding of a unique situation or disruption, and this is why a digital-only play doesn’t work in this industry.

Bozeman reveals his comeback plan in under 2 minutes

FREIGHTWAVES: If you could summarize C.H. Robinson’s comeback plan, in one to two minutes, how would you summarize it?

BOZEMAN: We’re using the power of technology, the power of process development and product development to put us in a position to satisfy demand while keeping a very competitive cost basis. That gets us a wonderful return for investors, it gives us value for customers and solutions for customers, it gets us engagement and rewards for our employees and it gets us an awesome experience for our carriers. 

We are in a really strong pole position for when the market pivots. Bringing a lean focus in the company has been strong. It’s something that I like and I’m trained in. But having the company really grab that and drive better problem solving, better execution, faster clock speed, all only delivers better value, more wallet share out of customers, more solutions for customers and better returns for investors and our employees. That’s how I would sum that up.

What do you think of Bozeman’s turnaround strategy for Robinson? Email rpremack@freightwaves.com. Don’t forget to subscribe to MODES for more trucking insight.

3 Comments

  1. Darren Maynard

    Love the thinking of the new CEO. There are huge opportunities for A.I within supply chains. It is a segment that was ignored by technology for many years and is a highly fragmented sector with huge amounts of complexity that have been papered over by skilled desk-level employees who are experiential learners limited by the scope of their work. LLM’s can not only eliminate the tedious tasks at the desk level but over time take a broader non-siloed approach to the problems. Will be time consuming though, since most LLM’s need huge amounts of data to learn from.

  2. Holle

    It has a lot to do with hiring people that look good on paper to do logistics they need to hire people who know logistics.

    You can’t teach true logistics from a book.

Comments are closed.

Rachel Premack

Rachel Premack is the editorial director at FreightWaves. She writes the newsletter MODES. Her reporting on the logistics industry has been featured in the New York Times, the Wall Street Journal, Bloomberg, Vox, and additional digital and print media. She's also spoken about her work on PBS Newshour, ABC News, NBC News, NPR, and other major outlets. If you’d like to get in touch with Rachel, please email her at rpremack@freightwaves.com or rpremack@protonmail.com.