Where does AI investment pay off in supply chain?

Sorry for the few minutes delay for everyone. Thank you all for attending. We're very excited to welcome you to this webinar on AI in freight, AI investment and supply chain, and really today, the goal is for us to bring down to earth this very popular topic of deploying AI in freight and logistics. To do that, we have a conversation between myself and two incredible panelists that I'm honored have agreed to do this with me.

First, let me introduce Miles Mattson. Miles is a partner at BCG in their transportation and logistics practice. He helps really large and scaled freight businesses with AI and digital transformation, and he also helps private equity clients that own those companies with very similar efforts on AI and transformation. We think he has a very unique lens into what all the scaled players are trying to do with AI these days, and how they're thinking about it, and those players come to Miles and to BCG for help to think about this stuff.

Yeah, very much look forward to the conversation today. Thanks for having me.

Thank you, Miles. We're also very lucky to have Peter Hasenkamp. Peter is a very seasoned supply chain executive with experience at top-tier automotive companies like Tesla, Lucid, and Ford, and he's been in the operator seat — the person on the shipper side that's making the decisions of what are the KPIs, what are the pain points, what are they thinking about when they think about technology. Technology's a tool for them, it's not the end-all, be-all, and we wanted Peter to bring that voice of the operator, of the shipper, to tell us what he's seen, what continues to be the same pain points that have been around the industry from all of his many, many years in the industry, at many companies. Some more technology-driven, some less. He's really seen it all.

Well, thanks, Deepak. Nice to be here again.

For those of you who don't know me, I'm Deepak, I'm the founder and CEO of Nuvocargo. Nuvocargo is an AI and logistics company that helps mid-market shippers — think manufacturers, distributors, retailers in North America — reduce their North America truckload freight spend, improve their service levels, simplify their tech stack, and just keep on the cutting edge of AI. We've been spending a lot of time deploying AI across our operations with shippers. We've analyzed over 100 companies in the space, so I'll try to chime in whenever it makes sense, but I'm going to be moderating the discussion.

We're going to try to break this up into three sections, then we'll have Q&A at the end. First: the status quo — what continues to be true in freight as it relates to deploying technology and AI, what are the pain points. Second: the future — where is it going, what does that mean for PE firms, large-scale freight companies, startups. Third: build vs. buy — how should you think about buying third-party vendors versus building something in-house? Then we open for Q&A.

For the first section on the status quo and where things are today, Peter, I wanted to start with you, because I think that there's so much hype in AI that sometimes we forget how things are done, why they're done a certain way. If I got this right, you spent over 25 years across Ford, Tesla, Lucid, and so many other companies in the space — very different companies, very different stages, very different philosophies around technology as well. Can you just tell us, across these 25 years, what are the persistent technology and operational challenges that followed you from one company to the next, and maybe that you still think are relevant today for everyone that's dealing with trade and logistics?

Yeah, for sure, thanks for that. At Ford, you had very well-established systems — a company that had been operating for 100 years, largely with established partners, suppliers, and fairly stable supply chains when I was working there in the late 90s and early 2000s. Then Tesla and Lucid were very much setting everything up from scratch — establishing brand new supply chains, setting up warehouses, setting up freight relationships globally, and trying to do all that in a very compressed amount of time.

I think a lot of the challenges remain the same across those decades in terms of information flow, end-to-end visibility of your freight and of issues that come up. The landscape's changing certainly with these new tool sets — you're seeing a much accelerated information flow in some cases. A company like Ford is going to be grappling with a lot of those legacy systems: how can they implement new technology and AI effectively with them, can they link all those data sets, can they get those systems working smoothly together?

A startup company — whether it's a startup or even an established legacy company going into a new region — still has all those same challenges: managing the space, managing the communications, and building connections between different companies and different systems across the supply chain, from the corporate procurement functions down to the warehouse and also out to the various freight providers. We're still in an experimentation mode where people are still struggling with those implementation timelines. If you need to set up a new warehouse management system or new freight brokers, those configurations at the detailed information level take quite a bit of time to set up and architect.

Just to play that back to you — lengthy implementation timelines, visibility, managing all the different parties in the system, whether it's carriers and brokers, maybe the software vendors. And large companies: it is hard to get everyone coordinated and trained on new systems. These things just take time. Can you share one example of something you felt noticeably improved as it relates to technology in freight and logistics in your 25 years? Like one or two things that were very different from your last year versus your first year?

Moving to Miles — you're advising private equity firms and these large LSPs, the majority of large logistics service providers. From all your years working with these companies and where we are today, what are the consistent pain points your clients are trying to solve with technology? What is still broken today? Before we talk about where the future's going — what has that been for all these years, and what does it look like today?

Building on a few of the things that Peter had mentioned — having the right data and a consistent single source of truth in the supply chain continues to be a challenge. You often have a disjointed set of systems which can create unnecessary manual work. Another piece just in the full workflow and lifecycle of a shipment is managing exceptions — and just the fact that exceptions exist in the first place means that something wasn't predicted effectively, there was a gap there, and then actually going through and solving whatever problem came up, whether it's a delayed delivery or a stock out.

Visibility has improved. The speed at which you get feedback in terms of where something is has improved, absolutely. But moving from visibility to making the intervention to actually predicting in advance where those outages might exist, so you can prevent them from happening in the first place — that's still, to some extent, a pain point in uncharted territory that continuing to strengthen things like predictive modeling is going to be a real asset going forward.

And then when I think about making change and adopting these types of tools, I think we'll continue to see a change management component and challenge here. As we've seen with Anthropic and OpenAI launching consulting arms — there is this question of driving adoption as well, that even when the technology matures and becomes readily available, you're still going to need the people side to really fully bring to life the opportunity at stake.

Can you give us one level more of detail? For the audience to understand — a big logistics company or PE firm, they're hearing and reading about AI, and then they call you. What do they tell you? What do they ask for your help with? What is stopping them? I'd love to get behind the scenes of how some of these companies are trying to solve these problems with you and your company.

There are probably a few different levels. From a private equity lens, even before you get into the adoption and change management component, some of the key questions are: A, is AI going to fundamentally disrupt or disintermediate my business? And B, assuming it's not, what opportunities exist to deploy it in the core workflow to keep these assets competitive and ahead of the competition? The risk and opportunity lens are generally the starting point of those conversations. And then the question of adoption is critical — on one hand you can deploy these tools and cross your fingers that organically the word will spread, and on the other hand, to what extent do you have to structurally make sure there's an understanding of how they create value and how workflows fundamentally need to be changed so you can truly realize the value that they hold?

Freight companies — do they see more opportunity or more risk to their business with AI? If you could generally speak to that, I think that'd be interesting to the audience.

I was just going to tag off of one thing you said, Miles. These tools are also so flexible that I've seen some instances where one employee's running one way with the tool and another employee's running a different way. I'm wondering if you also see concern from some of your clients around AI proliferation — trying to curtail how the tools are used so that you don't have people spinning up their own agents going somewhat wild in a direction the company maybe can't fully control or guide.

We are still in that experimentation phase, and generally you want these tools to be tied to direct business outcomes — the number of carriers you can interact with per hour, or the number of touchpoints that are fully automated versus manual. In order to get that repeatability to drive those business outcomes, having the process discipline around how it's being used is important. On the other hand, given how new many of these tools are, there is a frontier still to be defined in terms of how they're used, and having the flexibility to experiment and figure out the right way to embed them in workflows to make them most valuable is still, to some extent, there's room to be defined. Employees should be able to experiment to figure out the best way to get value from them, and then deploy those learnings throughout the org.

It's so easy now to create dashboards that everyone has their own — six different employees all in the same group with all their own dashboards, which can be great if they're doing it for something tailored to their business. But if it gets into a competition over what is the right metric, I'd always reinforce with my teams: focus on the K in KPI. What is the key performance indicator? I don't need to see 20 different charts, I need to see the three that are actually meaningful. That's where I can see some of this proliferation happening.

Here's a question I wanted to ask before we start talking about the future. Something that is under-discussed about AI deployment is the culture of the company that's trying to deploy it. Everyone has now passed the hype phase of "the technology is amazing" and is now trying to make it work. That's why they work with Miles, that's why OpenAI and Anthropic started the deployment companies. Peter, you already alluded to things in your culture that might cause people to standardize or be willy-nilly, and Miles mentioned the importance of a culture of experimentation. Starting with Peter — in your experience working at such diverse companies, how do you think the culture of innovation sets companies up to do well or poorly going into this next era?

A company like Ford has a very structured approach. People have already figured out what those dashboards need to look like, they have processes in place, they have reports, they have structure, they have data systems set up. You're standing on the shoulders of giants — people have spent decades refining a lot of this stuff. The innovation isn't always as necessary to do the day-to-day job because you have a well-oiled machine functioning. The counter to that is you do end up stifling innovation sometimes, because people are used to operating in a very structured environment where they don't have to invent things from scratch very frequently. There's almost always a standard way of doing things.

In this environment when you've got this kind of innovation happening so quickly — and as Miles was saying, everyone's in an experimentation mode, so there isn't a best practice that's been established yet — in a big company like that, you need to be thoughtful about how you motivate people to experiment in-house without creating a lot of churn or noise.

At a company like Tesla or Lucid, experimentation and innovation is just a daily part of life because you're setting everything up from scratch, you're building the plane as you're flying it, which forces people to be innovative. Sometimes people who come from a structured background struggle with that environment because there are no best practices, there are no standard ways of doing things, and every time you ask where the bathroom is, you get an assignment to build a toilet. An innovative company can run, but they may run three steps forward and take two steps back because they ran the wrong direction, or they got tied up trying to decide which new software to experiment with or implement.

Very quickly — which of the two cultures do you think is better set up in the age of AI? The more structured one, or the extreme experimentation culture?

Ideally, you have a blend of both. Startups have a natural inclination — you have people who are comfortable with new things, so you're naturally going to lend toward adopting things faster, experimenting, especially in this current time when everything's a little nebulous still.

I really liked what you said — you have to experiment, but then you have to lay some foundations, otherwise you're just innovating for innovation's sake and systems don't scale. Miles, what are you seeing from a culture of innovation standpoint with your customers, and how much is that playing a role in how quick they are to understand and actually deploy these tools?

You said there was dispersion six months ago — so there's been some kind of acceleration or widespread acceptance happening in an increasing way in a matter of months. Is that fair to say?

Yeah, I think so, just because you're seeing real value being delivered and players with solutions out there that are helping companies overcome some of these core pain points. It's gone from a place of maybe being a little more uncertain to "hey, this is happening, should we do this?" — and the question now becomes more "how?" and "by what path?" and "how do we prioritize?" Because none of this technology is inexpensive, and demand is outstripping supply — so it's more a question of how you optimize across impact, speed, and cost based on your business objectives, as opposed to "this is where the industry is going," which I think everyone more or less believes at this point.