The One Thing AI Still Can't Do in Your Supply Chain

"The one thing that AI can't do is accountability." — Blake Tablak

Beyond the AI hype, Blake Tablak, CEO of Trax Technologies, sat down for a live conversation on where AI is genuinely moving the needle in supply chain, and where it still runs into a wall.

Watch the Conversation

In this discussion, Blake Tablak, CEO of Trax Technologies, breaks down where AI is creating real value in supply chain today, what it still can't do, and how executives should separate genuine opportunity from hype when deciding where to invest.

Resources

Full Transcript

[00:01] Host: Hi everyone, and welcome to the Trax channel. Today I'm here with Blake Tablak, CEO of Trax Technologies. Hi Blake, how are you?

[00:01] Blake Tablak: Wonderful, wonderful. Thank you for having me.

[00:19] Host: Good. So today we're talking about beyond the AI hype, how technology is creating real supply chain value. This is a hot topic. Everyone is talking about it. Everyone's posting about it, especially CEOs, because you're the head of the company and the training trickles down, and your standards for continuing your education in this. So the first question is, AI is dominating conversations across industries. Where are you seeing AI create the most meaningful value in supply chain today?

[00:59] Blake Tablak: Yeah, this one's a great question, because there are two parts to this. I think there's where is it adding value, and then where can it not add value at this point. I guess this is important context, especially as a CEO. We talk to our CTOs, our heads of engineering, our product teams, and AI can do anything. Okay, that's great, and I agree it can do a whole lot more than previous technologies. An incredible amount of innovation, and the pace of innovation, great time to be alive.

That said, there are some things it cannot do. So what I would say, generally speaking, is rapid insight, or the ability to look through large swaths of data quickly and get an insight back. This one's really interesting. If you're a CEO, a chief operating officer, or a CFO, and you're not using a Claude or an OpenAI for something, I think you're really missing out.

The reality of these things is they're very capable, but they are not infallible. I've done many an analysis, I'm probably a daily user of Claude, and to run deeper analysis across information I have access to, financials, what have you, for our business, oftentimes I get an incorrect answer. And because I understand the context of what I've asked, and I understand how the data works, I know it's wrong. But if you didn't, you may look at that and go, "Well, that looks right." And then make a bunch of decisions that are incorrect. So the first thing I'd say is it's really good at analyzing large swaths of data, but you still have to have the knowledge about how that data works.

The second thing it's really good at is operational execution. It's the next evolution of robotic process automation. There was a lot of hope in robotic process automation, and it delivered a ton of great ability to automate. Now it's, "Okay, I don't need to know exactly how the process works, I just have to be able to explain generally how it's supposed to work, and where the decision trees are," and continue to train the model. So it will do repeat operations. I've been engaging with more and more support agents in my personal life, asking questions, and I know they're agents, but the communication's really great. It's instant, and it's 24 hours a day.

So executing on operations and giving you insights, I think these are the two things right now that, more than anything, an executive and a company should be able to execute on.

Here's what it can't do. One of our top account executives here, she and I were talking the other day, about a major account, and she said, "The one thing that AI can't do is accountability." And in supply chain, accountability is so important. The way supply chain is designed, the inherent nature of it: multiple vendors, multiple different modal types, multiple carriers, multiple suppliers. It is not designed for perfect execution, which means the relationship and the accountability still have to be there.

We provide program services to some of our customers, and one of the things I tell them is, look, it's not that you can't do this in the platform, totally self-service, a hundred percent you can do it yourself. The big reason you would want me to do it is the twenty billion plus that I am managing. I have a lot more ability to get somebody to do something, or to see past just that small problem and address a broader problem that carrier may be having. And so I carry more weight, I can help them be more accountable to themselves as well.

And I think that's fascinating, because it can do great alerting, it can give you early warnings, it can tell you all kinds of things, but the accountability piece it can't do. I don't know when it will be able to do that, many years from now, maybe, or maybe not, who knows. But that piece it can't do. And I think it's an executive really understanding, can I free up my team so they can go do higher level work, but can I hold myself and my partners more accountable? I can't do that with AI. It's not designed for it, and I don't know when it will get better.

[04:54] Host: Yeah. So how can companies separate real AI opportunities from the hype, and identify where technology investments will have the greatest impact?

[04:54] Blake Tablak: Yeah, I would say the first thing is don't think about AI specifically. I can go off-roading more effectively in a Ford Raptor than in a Kia Sportage. The tool itself and the solutions themselves, don't think about it in the context of those, think about what does the business need. We present, monthly, quarterly, to some of our largest clients, operational insights through our intelligent operations agent. And it's constantly monitoring what's going on and coming back and saying, by the way, there's a more effective way to do this. There's a more effective way to ship these kinds of products. There's a more compliant way to do this. That's awesome. But that's us, we knew what questions we wanted to ask, we knew the things that would be most interesting to our clients. Most of them get it. And so we designed an agent to constantly monitor these things, and look for insights in the data, and give specific means and modes, here's where to start, and that's all great.

That's the place to start as an executive. What are the questions you're trying to get answered? What are the operations where you think you can find some insight or value? What are some suspicions you have? You've been doing this for long enough, you're suspicious of something. I was a chief revenue officer for many years, and I'm very suspicious that X or Y is actually happening. Good news, I can use AI to find out if it is or it isn't now. And I think that's where separating real AI opportunities from fake ones matters, and if somebody says, "I'm going to AI everything," I think your alarm should go off. I think it should, especially in supply chain. I think you should be very wary of that statement. Be like, "Hey, I want that to happen, I'm all in." That said, I need to see some use cases that are actually going to pull themselves out of this and accomplish a problem. Cost, quality, speed, emissions. There's your four biggest drivers of any executive's life when you wake up in the morning and you're in supply chain. So which one matters most to you? Pick that one right now and say, what can I do to drive that down, and what are some suspicions I have about our ineffectiveness or inefficiency? And use AI to go uncover it, because it's a much better solution for uncovering that kind of detail.

[07:11] Host: Absolutely. So what role does data quality and visibility play in helping organizations successfully implement AI solutions?

[07:11] Blake Tablak: I gave a talk to some of our executives many months ago. I've been working with ThoughtSpot for a long time, and one of the things I said is that if you're going to go look at doing some work on your own desktop, the first thing you need to do is find the data set you're working with. Make sure it's complete. And you're like, "Well, how do I know if it's complete?" Well, upload it and start interrogating the data with Claude. Pick low, medium, high, depending on what model you're using, and how important the outcome is. Is it directional correctness, or is it precision darts in the middle of the target? Pick that mode, and then start interrogating it. Does this data look complete? What areas look to be missing? Are there any anomalies in this data?

That's the next thing, once you get a complete data set, then start thinking about cleaning it. Talk to me about the anomalies you're seeing. Are there gaps in this data? Is there other information that would be required for me to get answers to the questions that I need? That's the beginning of your journey when you're sitting using Claude or something like that, that data cleansing. If you don't have the data to start with, or it doesn't exist, AI won't be able to help you. If you can infer it, if it's just a gap or a missing piece and there's something else you can use to infer it, that totally works, we do it all the time. But if it just doesn't exist in the first place, manifesting it out of nothing, there's nothing that does that.

So what I would say is data quality matters a ton. Our entire platform is designed around using artificial intelligence to ingest, map, and normalize. The reason for that cycle is that normalization allows our clients to do all kinds of interrogation, because carriers call something different every time. They call things unique things every single time, and that makes it impossible, or very difficult, for a company to gain value out of the data coming back from this process. So I think data quality is huge, but data completeness is the other one, and you really have to come up with some processes to make that work.

[09:20] Host: Absolutely. And how should supply chain leaders balance investing in emerging technologies while ensuring they solve real business challenges? You touched on this earlier, but this is—

[09:20] Blake Tablak: We did, but I think you start at the end. What is the real business challenge you're having right now? There are a lot of companies that really need to find ways to cut costs, and the answer of just "ship less stuff" isn't going to fly, that's not going to fly with the CFO. By the way, increased costs are coming, if not here already. If that's your challenge, understand there are trade-offs: cost, quality, speed, emissions. There you go, in logistics. If I drive costs down, speed may come down and emissions may go up. As you move things around, you get secondary effects. Make sure you understand what those are. But pick the business problem you have right now, and figure out how to go attack that business problem, and then start being curious about the other questions that exist.

And by the way, when we built intelligent operations, there was a bunch of people in our company who were wildly intelligent on logistics, who'd run logistics organizations for years. We probed at what customers were interested in. Call your peers, call other people, and go, "Hey, what questions are you asking?" I always joke with my CFO, when we're talking to somebody and she says something and I don't understand what it is, I write it down and call her back and go, "Okay, you're going to think I'm an idiot, but what did that mean?" To this day I ask questions about stuff I don't understand. I really think that's super important, being curious and being okay with not knowing something. That turns out you're going to get a lot further, a lot faster. So pick the business challenge, make it real, and then uncover how peers are doing it, and figure out, if I went and invested 15, 20, 30 percent of my technology spend into this, am I going to get the return? If I have a ten million dollar portfolio and I invest three million, will I get thirty million back? That's really the question. I need a ten-time return on whatever I'm investing in. I think that's a good way to go about it as a leader of any kind, supply chain or otherwise.

[11:29] Host: Absolutely. So our last question, and I know this is broad, is looking ahead, how do you see AI changing the way companies plan, manage, and optimize their supply chains?

[11:29] Blake Tablak: Yeah, this one I think is really straightforward. Number one, generally in planning, historically you would be limited in the number of potential iterations you could do. When we're doing our annual planning, as an example, we literally have three different plans, the red, green, and blue plan. If this happens, then blue happens. If this happens, then red happens. If this happens, then green. Sometimes it doesn't matter what the colors are, it's plan one, two, and three. That's really predicated on how much processing and output power we have as a company, and that's generally why companies only did three.

Well, think about it this way, you no longer have that limitation. You can process twenty different outcomes. So what I think is going to happen in terms of planning is there's accuracy, and then there's what I'd call levers and dials. You want highly accurate plans, which means you need highly accurate data. But when it comes to the different manifestations, what do we do if this happens, that's a lever, then we're going to do this. If it only happens so much, that's a dial, then we're going to do this. And I think that's really what's changing the planning and managing.

It used to be I had to pick between resilience and cost effectiveness, as an example. Maybe I don't anymore. In fact, what I can do is analyze such large volumes of data, I may pick spots where resilience matters more, and spots where cost matters more, and I can get way deeper into the pile than I ever could before. As a leader, as the CEO of the company, I can get deeper into the information, deeper into my clients' data. I can see more, I can do more, I can suggest more, I can be more explicit about where that suggestion should be than I've ever been able to before. So I think when it comes to planning, managing, and optimizing, you're able to get down to the root cause so much faster, and you can see the change you need to make easier than historically you've been able to. And I think that's what it's doing right now.

[13:38] Host: Absolutely. Well, it's such a great conversation with you today. I've said this all week, I can't believe 2027 is almost here.

[13:38] Blake Tablak: Wild.

[13:38] Host: And so in the coming episodes with you, we'll talk about predictions for 2027 and what's going on in the supply chain world. We'll talk again soon.

[13:56] Blake Tablak: Awesome. Thanks, [Host].

Watch More From Trax

For more conversations on AI, supply chain strategy, and transportation spend management, subscribe to the Trax YouTube channel or follow Trax on LinkedIn.