Every logistics and retail executive today hears the same thing: AI is going to transform the last mile. But where does last-mile AI actually move the needle? And what’s the path to build an operation that captures value without derailing what's already working? For most organizations, the gap between AI interest and AI impact is wide, and the bridge to connect them isn't clear.
Benjamin Gordon, managing partner at Cambridge Capital, is one of the best people in the industry to close that gap. He has advised Fortune 500 companies on technology adoption and invested in companies that have put AI to work in real logistics environments. He has also hosted over 400 of the top supply chain CEOs at the BGSA supply chain conference for 20 years. Benjamin has seen what works and what doesn't, at scale, across the full spectrum of the sector.
On this episode of Deliver: The Last-Mile Performance Podcast, hosts Bringg CEO Guy Bloch and Sales Engineering Lead Raquel Zanoni sit down with Benjamin to cut through the noise about AI’s real value across the most complex segment of the supply chain.
Watch the full interview to learn:
- Why one logistics company’s 56% stock gain in a down freight market came down to two AI agents and two specific problems
- The three areas where AI has the highest ceiling in last-mile delivery
- How to experiment with AI in a mission-critical environment without putting your operations at risk
- Whether to build, buy, or do both, and why the answer depends on your company's size and install base
- The one metric shift that separates last-mile leaders from the rest
Key takeaways
Three areas where AI has the highest ceiling in the last mile
- Workflow automation: logistics is a series of connected process steps. AI is well-suited to string those steps together, remove manual handoffs, and reduce the cost of execution.
- Pricing intelligence: feeding large volumes of freight data into a trained algorithm produces faster, more accurate, real-time rate decisions than legacy benchmarking tools.
- Cost reduction: AI enables smaller, more focused teams to do more. The gap between what one person can accomplish with AI today versus five years ago is significant and still widening.
Successful AI adoption requires a defined problem and a clear use case
- The organizations that succeed define a specific problem before making any investment. The ones that fail start with the technology and look for a problem to attach it to.
- Be precise upfront: does the business want AI to optimize its DC network, improve inventory decisions, or reduce response time to customer requests? The answer shapes everything that follows.
- Working with partners that specialize in the use case accelerates both the result and the learning. The best AI companies want educated customers.
Build and buy at the same time
- Small internal teams experimenting with AI can build institutional knowledge. External partnerships with specialist companies deliver speed and domain depth. Both matter.
- Fortune 500 companies should treat AI spend the way public market investors treat R&D spend: as a percentage of revenue that reflects a commitment to future competitiveness, not a one-time cost.
- The companies that do both move faster and learn more than those that choose one path exclusively.
Experimentation in the last mile requires guardrails, not avoidance
- Logistics is mission critical. Handing AI control over live operations without testing is a significant risk. But avoiding experimentation entirely is also a risk.
- Start analytically: use AI to surface patterns in the data, model routing scenarios, simulate DC placement decisions, and run hypothetical supply chain configurations.
- Test operationally in one market, one store, or one variable at a time before any broader rollout. The goal is to build confidence in AI-driven decisions before scaling them.
The data that matters most to the C-suite
- Holistic cost: the fully loaded cost of the supply chain, including inventory carry, working capital, and warehousing, not just transportation spend. AI enables the kind of cross-functional optimization that wasn't computationally feasible before.
- On-time delivery and customer satisfaction—these are directly correlated. 80% of consumers who have a bad delivery experience don't come back. Accurate, proactive communication is one of the highest-leverage inputs.
- Resilience: the ability to identify and reduce delivery risk before something goes wrong. AI analysis of last-mile patterns surfaces failure points, from carrier reliability to day-of-week delivery risk, that only appear in the data.
Last-mile leaders benchmark against the best, not against themselves
- Most organizations measure progress against their own past performance. The market, however, judges them against the best available alternative.
- The right question isn't whether you're doing better than last year. It's whether your consumer experience is better than your strongest competitor's in the eyes of the consumer.
How last-mile technology investment has changed
Guy: Ben, you've been at the intersection of logistics and technology longer than most, from founding 3PLX in the late '90s to investing in the next generation of supply chain leaders at Cambridge Capital. From that vantage point, how has the conversation around last-mile technology investment changed over the years, and particularly now with AI?
Ben: Well, since you referenced the beginning, I'll tell you. When I started my first company 27 years ago, 3PLX, which was a SaaS TMS company, last mile wasn't even part of the conversation. When I went to Silicon Valley and talked to venture capital firms about what we were doing, I said, "Our name is 3PLX because we're execution software for 3PLs." And the most common question I got? "What the hell is a 3PL?" So it's miles and miles apart from then to now.
Today, the awareness, curiosity, and desire to solve problems in the last mile, whether you're talking about venture capital, Fortune 500 companies, logistics software firms, or anybody else in the ecosystem, is greater than it's ever been. And that's because, number one, we all became painfully aware of how important it was. You don't know what you're missing until it's gone.
Five and a half, well, six years ago now, when we all experienced the COVID shutdown, the last mile became really important because you couldn't get anything, you couldn't go anywhere, and there were shortages of everything from semiconductors to chicken to toilet paper. All of a sudden, the last mile went from being a curiosity, to a subject of cocktail conversation of all things, to a vital need that everybody cared about. The last mile is more important than it's ever been. It used to be something people didn't even know about or think about. Now it very much matters.
Where AI investment in logistics is actually flowing
Guy: With AI investments in logistics accelerating, there's still a wide gap between what gets funded, what gets piloted, and what actually improves last-mile performance. Sorting signal from noise is one of the biggest challenges executives face right now, and we see it firsthand. As an investor, you see where capital is actually flowing in logistics technology. Where are the most serious AI bets being placed in the last mile right now?
Ben: A couple of parts to your question. First, let's talk about venture capital and logistics, then AI, then the last mile.
Venture capital in logistics: about $12 billion is being invested in the logistics ecosystem. And if you compare that to 27 years ago, the number was a fraction of that, less than a tenth. So the magnitude of capital and interest in logistics and supply chain software is just off the charts compared with where it used to be.
Second, there's the issue of AI. Five years ago, AI existed, but it wasn't a major focus or major part of the dialogue. Today, something like 65% of all venture capital is in AI, and that's true in logistics as well as other areas. Some of that is skewed by large companies, like OpenAI and Anthropic, which in aggregate skew the data. But that's true at all levels, including the earliest stages, and it's true for the companies we see in logistics and supply chain.
Third, you have AI in the last mile. There are plenty of interesting early-stage startups we see at Cambridge Capital, and there are plenty of successful companies that started out as SaaS and are figuring out how to infuse AI into what they're doing. But any software company today has to be thinking, and hopefully more than thinking, about how to pivot from traditional to SaaS to AI. That's true for just about every company we look at at Cambridge Capital.
Raquel: Out of that $12 billion you mentioned in venture capital logistics, do you know how much of that is specifically focused on the last mile?
Ben: The piece that's last mile is less than 10%, but that's not surprising because last mile as a percent of total supply chain spend is in line with that. If you think about it from a macro standpoint, you've got warehousing and inventory management, you've got transportation, and then you've got other forms of supply chain analytics. There's over a trillion dollars in US supply chain spend, of which about $600 billion is transportation, $100 billion-plus is warehousing, and about $300 billion is in the realm of inventory management, working capital, and the supply chain capabilities around that. So last mile as a percent of total supply chain is less than 10%. It wouldn't be surprising that the same ratios hold true for venture capital.
Where AI delivers the biggest ROI in supply chain
Raquel: Maybe that helps frame the next question, which is about AI, though we can probably expand it to technology spend in general. From your perspective, how should companies think about what to focus on across their supply chain when they're investing? The ultimate goal being driving measurable returns. Where do you see that technology spend today, and where do you think it should be?
Ben: At our annual conference in January, which we've hosted for 20 years with over 400 of the top CEOs and leaders in the global supply chain arena, this was a huge topic of discussion. A couple of observations.
Number one, when you think about where AI has the biggest impact, supply chain is number one. There was a Stanford study that showed that of all the different verticals where you could apply AI, the biggest cost savings are coming in supply chain. That makes sense because there's a lot of low-hanging fruit. Whether you're talking about the last mile or inventory management, deciding where to put various things, various analytic decisions involving lots and lots of data, supply chain is ripe for that.
Within that, a great case study is in the public markets, because there's a lot of data around this. In US domestic transportation, we're now in year four of what's normally a one-year recession. Freight cycles typically go up and down over a period of a year or two, but the freight recession that began in early 2022 lasted four years. I actually think it's over now, and the data over the last three months reflects that. The point is, four years of record drops in prices: truckload and drive-in truckload rates dropped over 50%, record bankruptcies in trucking, logistics, and other areas. A lot of pain all around.
Despite that, one of the top performers in the entire S&P 500 last year happened to be a truck brokerage and logistics company, C.H. Robinson. They were up 56% last year. How did that happen? Well, it turns out AI was a big part of the story.
Ben: C.H. Robinson made a decision to focus on AI. They launched 30 agents, and one very practical problem they wanted to solve was faster and more complete responses to requests for quotes from customers. Customers communicate the way they want to, not the way you want them to. So a lot of it comes in by email.
C.H. Robinson has thousands of employees, so you can imagine the hundreds of thousands of emails that come in. It turns out they were only responding to 60% of requests for quotes. Think about that: people are asking you to do something, offering to pay you, and you're ignoring them 40% of the time. The first goal was getting that response rate from 60% to 100%. The second was cutting the response time. The average was about 20 minutes, and in the freight industry, 20 minutes is not good enough when you're making real-time decisions.
The agents focused on those two problems. They drove the response rate from 60% all the way to 100%, and they took the response time down from 20 minutes to less than 30 seconds.
So no surprise that in a year where the freight industry was getting crushed, C.H. Robinson had revenue up, gross margin up, bottom line up, metrics across the board up, and the stock price reflected all that: a 56% increase. To me, that's one of the best examples to show that AI isn't just something techies talk about. It's something smart CEOs are using to create real value and real impact in transportation and logistics.
Raquel: I really like that response because it frames up something we see a lot as well, which is that AI doesn't necessarily need to be this huge strategic insights play, pulling all this information out of your data. For a lot of these companies, especially in logistics, it's about removing manual processes, streamlining or automating workflows in a way that improves the customer experience and the worker experience. So I really like that you highlighted a seemingly simple but very effective use for AI.
Three areas where AI has the highest ceiling in last-mile delivery
Guy: It's indeed a huge opportunity. Thinking about all the areas where we can really drive performance in the last mile, from routing optimization and demand forecasting to delivery visibility and automated dispatching, there are just so many areas where you can identify potential investment. Where do you think AI has the potential for the highest ceiling over the next few years?
Ben: There are several areas where there is tremendous opportunity. I'll give you three.
One is automating workflow. Logistics is all about stringing together a series of steps to move something from point A to point B. Those processes might be manual, they might have some software in them. AI is a fantastic tool for taking those process steps and automating workflow. We see that in other areas of logistics. Palette, for example, is an AI company doing workflow automation in freight forwarding and truck brokerage. At the BGSA Supply Chain Conference in January, we had a shark tank where companies competed to show they were the big winners from an innovation standpoint. We had over 60 applicants this year. More than 40 of them were in some form of AI, and most of those were some form of AI workflow automation. So workflow is clearly a big opportunity. That's number one.
Catch up on the latest episodes of Deliver: The Last Mile Performance Podcast
Number two is using AI to make better pricing decisions. You can take massive amounts of data, feed it to AI, and train the algorithm to make faster, better, real-time decisions. A great example is GreenScreens, a company we invested in and that I was involved in as a co-founder six years ago. What GreenScreens did was say, "We will take your data, typically from a logistics company but also from a trucking company or shipper, run it through the algorithm, evaluate it, look for patterns, and ultimately, based not just on your data but on an aggregate of over $40 billion of freight data, tell you in real time what a truckload of freight should cost from point A to point B."
Ben: The old way to do it, you might have used DAT [a legacy freight rate benchmarking platform] or another legacy system, and they would have told you, but that data tends to be lagging, not forward-looking. It tells you what last month's data was, not right now. And it's only as good as what's in it. Garbage in, garbage out. You really want the full market, not just a handful of data from a handful of companies. GreenScreens was built specifically to solve that problem. It cut the error rate for pricing from 20% down to 5%, based on a study of one of the largest market competitors. As a result, it became much more valuable for customers, allowed them to push decision-making down to the front lines, make faster and better decisions, take out process steps, and make more money. Triumph, the largest bank serving the logistics sector focused on truck brokerage, bought the business last year. That's a great illustration of how you can use AI to do a better job with pricing, or really anything that involves sifting through large amounts of data.
The third thing is cost. We all know the numerous examples. Anyone studying AI knows the areas where you can take out costs. The big talk in Silicon Valley is about tiny teams, and when we're going to have the first one-person unicorn. That might seem provocative, but the truth is you can do more as one smart, talented person with technology today than at any point in history. AI empowers that. Those are three great opportunities for AI.
How to get started with last-mile AI
Raquel: I'm sure our listeners are hearing you describe these three buckets and some of them are thinking, "We kind of do that today," or "We want to do that but we don't know how to get started," or "We tried that and it failed." How do you think about framing this for large logistics companies that are dipping their toe in AI? How do they get started, and how do they evaluate which bucket makes the most sense for them? Because often they try something, it fails, and they think it's just not for us.
Ben: There's a story about a guy who tried to swim across the English Channel. He swam 99% of the way and said, "I can't make it," and swam back. Don't stop, keep going. Keep your eye on the prize and keep moving forward. That's number one.
Number two, work with partners that specialize in this. You don't have to do it all yourself. In the case of C.H. Robinson, they did several things. They did their own internal experimentation, they hired a team of engineers, but they also worked externally with great software and AI companies. It's not either/or. Do both. Try your own internal experimentation, but also work with great companies and partners. What I've found is that the best AI companies want to share their knowledge and educate their customers, because the view is, the smarter your customer is, the more they're going to value what you can do for them. Ask for the help.
The last thing is be precise about what the use cases are. What's the problem you want to solve? AI is not a technology looking for a solution. It's much more effective when you know upfront what problem you want to solve. Do you want to use AI in the last mile to do a better job of mapping out your network and deciding where your DCs should be? Do you want to use AI to assess your inventory optimization strategy? Or do you want to use it for some other specific thing? Take the time to define upfront what you want. The ability to use AI to get there is unbelievable.
Why AI initiatives succeed and why they fail
Guy: Let's talk about successes and failures in the world of AI, specifically in last-mile performance. Many logistics and retail organizations have launched AI initiatives over the past two years, but the outcomes vary widely. Some are very successful, some fall behind. What we see is that the difference between success and failure often comes down to implementation discipline, clear KPIs, clear use cases, and whether the organization built or brought the right capabilities. Based on what you've seen at BGSA, talking to executives, and looking at your portfolio companies, what are some of the reasons AI initiatives across the last mile actually worked, and where did they fall short?
Ben: Let's take both sides. Where do they work and where don't they work?
Where they work: number one, and C.H. Robinson is a great example, is when you do a great job of starting with the problem you want to solve. Robinson didn't start by saying, "Let's go spend a lot of money on AI and tell the street we're cool." They started by saying, "Let's figure out how to take that 60% response rate to customers and get it up to 100%." The most successful companies are those that are really good at defining the problem first. To quote Deng Xiaoping, let 100 flowers bloom. Try different ways of doing it, but start by defining the problem.
An example of a company that didn't do that, and this is true not just for AI but for last-mile and software in general, is Pier 1 Imports. One of the Cambridge Capital portfolio companies did a lot of work with Pier 1 about eight years ago. The problem was straightforward: if you're a customer with the choice between buying something from Amazon, getting it same day or next day with full tracking visibility and a great customer experience, or buying from Pier 1, where you might wait a week or more with little to no visibility into where your order is, you're going to choose Amazon. You might be buying the exact same product, but if you have the choice between the Amazon experience and the Pier 1 experience, we all know what happened. Pier 1 ended up going bankrupt. Not because they didn't have good products, but because they didn't have a good supply chain process.
Who knows, if Pier 1 had been a better adopter of AI and last-mile technology, could they have given customers the same kind of experience that Amazon customers got? Could that have been the difference between success and failure? I certainly think so. Part of this is about technology, and part of it is about how good you are at defining the customer problem you're solving.
Build vs. buy: what actually works for AI adoption
Raquel: I haven't heard the name Pier 1 Imports in quite some time. That was a deep cut. You've brought up this idea that AI implementation requires clearly defining what your customers need and using that to drive your technology decisions. We certainly see that as where companies tend to have the most success, when there's a top-down approach. Along those lines, there's always this debate about whether retailers or logistics providers should invest in technology they build themselves or buy from outside. I'm curious what you think about building versus buying, and whether there's a more proven or effective strategy there, especially as it relates to AI.
Ben: Yogi Berra famously said, "When you see a fork in the road, take it." What he meant was that sometimes both is the right answer.
Now is a good time for experimentation. If I were advising, and I do advise, Fortune 500 companies on this very topic, I would say: on one hand, you ought to have small internal teams that are experimenting, seeing what you can do, seeing what you can create. We're doing that here at Cambridge, and we're a hell of a lot smaller than a Fortune 500 company. On the other hand, also be an early adopter of great technology and great partners that can help you, particularly in areas where there are specialists.
As an example, we have enterprise accounts experimenting with tools like ChatGPT, Claude, and others, using those to automate various parts of our own business and help portfolio companies. Just to give you a taste of the potential: we invested in a company last year, I won't say the name, but a great company solving important problems. The website just wasn't that good compared to their competitors. Their pitch deck, on the other hand, was really good. So what did we do? We took five minutes, fed the pitch deck to Claude, wrote a detailed prompt outlining what we wanted in terms of outcomes, highlighted two competitors with great sites and what we thought was great about them, and less than 10 minutes later, Claude produced an entire website that mapped out everything. Objectively, it was infinitely better than what we had 10 minutes prior. That's a tiny example, just a taste of the potential of what you can do.
Use tools, use technologies, experiment. This is true across the entire enterprise. But in parallel, go work with great companies that have great tools, whether in last-mile, returns, inventory management, or otherwise. Look for companies that are leaders in those fields and work with them. Part of the value is the results, but part of it is also the learning.
UPS has a great phrase that one of our operating partners, Remus Capescu, taught me. Remus ran R&D at UPS and then ran their corporate venture group, the UPS Strategic Enterprise Fund. He said UPS judges their initiatives by both financial returns and knowledge returns. That's true if you're an investor at a corporate venture capital firm, but it's also true if you're on the operating side. If you work with Bringg or another company, you want a tangible benefit, but you should also want to be learning more, getting smarter, and benefiting as a strategic thinker in addition to the tangible results. I'd encourage companies to think about both.
How to experiment with AI in the last mile without breaking operations
Raquel: Can I dig a little deeper on the experimentation piece? I think it's interesting you mentioned that. I'm sure we all have a ChatGPT license and we're throwing things in there. But when we think about the last mile, experimentation can be a little dangerous, a little scary, because it affects your customers and their experience with your brand. What does experimentation actually look like in the last mile? How should retailers and logistics providers think about doing that, given the risk?
Ben: The underlying point is really important: logistics is mission critical. There's been a lot of talk lately about autonomous agents that can take over your computer. On one hand, it's great. On the other hand, what if they email your most important relationship with something stupid or accidental? Take that times 100 and now you're talking about the last mile.
So yes, you don't want to experiment with something mission critical where the consequences could be terrible for your customer. But there are things you can do. First, you can experiment with the analytics. A great opportunity to use AI in the last mile from an experimentation standpoint is to say, look at my data, find the patterns, and show me different and better ways to do things. What if I used AI to change my routing? What if I used AI to change where I put my DCs, or the frequency with which I do pickups for last-mile delivery? Those are all analytical experiments you can run, and you learn from them. That's one of the easiest things everybody should be doing: use AI to analyze your data and give you ideas for how you can do things differently across your supply chain model.
Catch up on the latest episodes of Deliver: The Last Mile Performance Podcast
Two, you can apply it in discrete areas. Jim Collins likes to say, fire bullets first, cannonballs second. Start with something tangible. If you're Best Buy with thousands of stores, go pick one store, try something different in that one location, and see what happens. You're not going to roll something out across the entire network and find out later it was a bad idea. Great software companies do a lot of A/B testing. Change one variable in one location versus another, look at the data, and see what works.
And then you can also use AI to create digital twins, where you run hypothetical scenarios. Here's my business now. What if I created a digital replica and tried running different scenarios against it? In short, there are plenty of ways to use AI for experimentation in the last mile without compromising the mission-critical nature of the execution.
The last-mile metrics that matter most to the C-suite
Guy: Ben, I speak with many logistics and retail executives, and we talk a lot about last-mile performance, always in the context of business outcomes. Even improving one metric by one point, at the scale of millions of deliveries a year, is worth millions of dollars in EBITDA (Earnings Before Interest, Taxes, Depreciation, and Amortization). But the dissonance is that most of them don't know what good performance looks like or what to aspire to. As we talk about AI, where have you seen it improve last-mile performance, and what are the specific metrics that matter most to the C-suite?
Ben: That is what matters most. What's the result? If we had the CEO of a Fortune 500 company in this conversation, that's exactly what they'd be asking.
So what are some specific metrics to look at? An obvious one is cost. Can I run my supply chain leaner? Can I take out cost in transportation? But cost is more than just transportation. There's the fully loaded cost, and where AI helps is in thinking holistically. What happens if I spend more on the last mile but I'm orchestrating my supply chain in a way where I'm running with less inventory? If I cut my inventory in half and spend a little more on transportation, I might be saving 10 times that in carrying costs alone. Less inventory means less working capital tied up, less warehousing, and all kinds of other ripple effects. We didn't have the computational power to make those kinds of holistic cost decisions before AI. We do now. So companies that are great at orchestration can help you evaluate not just the narrowly defined cost, but the broadly defined, holistic cost. That's one very important metric.
The second is the classic KPIs any Fortune 500 company cares about. What's my on-time delivery rate? If I reduce my cost significantly but my on-time delivery falls off and customer satisfaction drops, is it worth it? You have to make those trade-offs. Improving on-time delivery leads directly to a third metric: customer satisfaction. There is a linear relationship between the two. And as you know, the last mile is important not just because of the transportation component, but because it's also an extension of the brand. If I'm Nike and I don't deliver on time and my customers are upset, does it really matter that I saved 5% on my last mile if I've lost those customers? As Guy said, 80% won't come back. So customer satisfaction is a critical metric, and you can track it and track the inputs that correlate with it. Accurate communication with customers matters too. The only thing worse than getting something late is getting something late and not knowing it was going to be late.
And then third is what you might call resilience across the holistic supply chain. Are there things you can do to reduce the risk of things going wrong? Using AI to analyze last-mile patterns, you might find that certain carriers carry higher risk, or that deliveries on Friday afternoons are more likely to fail, for reasons that only surface through a close analysis of the data. Those are three critical things you can use AI to study: cost, customer satisfaction and its inputs, and resilience. AI can help you do that better than any tool we've had before.
How to balance AI automation with human operators in the last mile
Guy: When you do it right, you get higher conversion at checkout, happier customers, they come back, lifetime value increases, the company makes more revenue and reduces cost, and it all flows into EBITDA. But supply chain is a tricky area because it's mainly a physical world, trucks and warehouses and inventory and drivers and store associates, so many physical elements coming together to create the perfect last mile. Leaders are increasingly asking themselves how to integrate AI without compromising current operations. And there's an ongoing debate about whether AI will replace operators or augment them. From an investor perspective, where do you see the balance between humans and machines across the last mile, both today and over the next few years?
Ben: You might have the world's greatest system, but if it's hard for the customer to adopt because they don't want to get rid of their existing technology or workflow, you've got a problem. I see a few different ways of looking at it, three in particular.
One is companies that say, "Let us come in and run everything from start to finish and take over." In principle that sounds great, but it means ripping and replacing a lot of things, and that doesn't work for large enterprises. You're not going to get a Fortune 500 company to do that. It can work for SMBs, and there are interesting AI companies in last mile, returns, fulfillment, and other areas that say, "We'll be the full tech stack for an SMB and do everything." That works great because they don't have an installed base. But it doesn't work for the big guys.
Second, for companies targeting the enterprise, we see a lot more success when you add a layer on top so you don't force them to rip and replace. A good example in inventory and warehouse management is AutoScheduler. They were the winner of the Shark Tank at our conference two years ago. Their whole value proposition is going to large companies and saying, "You probably already have one or more WMS installations and rely on them for everything. It's really hard to get rid of that, just like it's hard to get rid of an ERP. We're not going to tell you to get rid of it. We're just going to point out that most WMS solutions promise labor management that they don't deliver, and we're going to be the layer on top that solves that." They come in as AI-powered labor management, integrate with all the major WMS options, and say, "Keep what you have, but use us for the labor management piece." That works for enterprise because they don't have to rip and replace.
There are longer-term questions about whether sitting on top of somebody else's platform is ultimately the path to building a multi-billion dollar business, or whether you have to become the core foundation. But it's a great way to add value, make the customer problem easier, and work collaboratively with the existing install base.
Third are the analytic tools that sit off to the side. GreenScreens is a good example: take everything you already have, but solve something new alongside it. So whether you're replacing everything for SMBs, sitting as a layer on top for enterprise, or solving a discrete problem from the side, you can win in any of those in the long run. Every company wants to be the platform, the main event, not the side dish. But there's a question of how you get there. Do you start there, or do you get there over time?
How to integrate AI into last-mile infrastructure without getting overwhelmed
Raquel: Given all the different directions we've heard about, different KPIs and metrics, different layers of AI and where they fit, for retailers and logistics providers that want to integrate AI into their infrastructure but feel overwhelmed by the options, what advice would you give them on how to get started?
Ben: Number one, there's top-down and there's bottom-up. Top-down, define what your big priorities are as an organization. Bottom-up, pick a couple of specific things. The C.H. Robinson example: cut my response time from 20 minutes to 30 seconds. You need the dashboard and the goals, but you need both top-down and bottom-up.
Number two, look for companies that are great in whatever field you're trying to solve. You could pick one, but more likely talk to three, give them some data, give them an opportunity. Most great companies will welcome that. Run some experiments, do the A/B testing. Give your data to three companies, see what they each come back with, run an analysis. That's not just making a bet, it's making an educated decision.
Number three, lay out a roadmap. Do this test in this market, and if it works, roll it out. But the biggest thing is commit to starting. In the tech world, public market investors often look at R&D spend as a percent of revenue as a proxy for innovation. It doesn't guarantee results, but if you and I have two companies and I'm spending 10% of revenue on R&D and you're spending 20%, assuming we're both spending it equally well, you're going to be more successful. You're going to have more innovation and more success as a result. Fortune 500 companies should think about working with AI companies the same way. How much are you spending on it? Because it's not really an expense, it's an investment, just like R&D is a percent of revenue.
Advice for improving last-mile performance
Guy: Ben, this has been such an insightful episode. I feel like we could keep peeling back the onion and look at so many more perspectives, but we are coming to an end. I wanted to ask you one last question. You get to see a lot and meet so many different sides of the market. If you could give one piece of advice on how leaders can improve their organization's last-mile performance, what would it be?
Ben: Most of us are internally focused. We look at how we're doing today versus yesterday. But the market doesn't work that way. The market is judging you against all the other alternatives.
Ask yourself how the consumer looks at you. Going back to Pier 1 Imports, if you were part of that management team having this conversation, you might be saying, "Let's look at our metrics now and compare them with a year ago. We're doing 5% better." But wouldn't it be a better question to ask how your customer compares you with the best alternative? How does the user experience with you compare to the user experience at Amazon? And when you invert it, as Charlie Munger used to say, when you have a problem, invert it, you realize you've got to solve a much bigger problem.
So if you're running a major enterprise and thinking about how to give your customer the best last-mile experience, sure, it's relevant to ask whether you're doing better than you were a year ago. But the real question should be, who is my number one competitor, what are they doing right in the last mile, and what do I need to do to be materially better than them? If you're a retailer, you should be thinking about what you need to do to be better than Walmart and Amazon. Above all, take an outside-in perspective instead of inside-out, and ask yourself who does last-mile better than you in the eyes of the consumer, and how do you exceed that experience.
Catch up on the latest episodes of Deliver: The Last MilePerformance Podcast
