The Latest Technology in Freight Transportation with Samsara

- Welcome to another episode of "Supply Chain Frontiers," the MIT CTL podcast where we explore the trends, technologies, and innovations shaping the future of supply chain management. I'm your host, Mackenzie Berry. Today's episode is in a different format. It features a panel at Samsara Beyond 2026. The panel is moderated by tech journalist Jason Del Rey and includes our own Dr. Angi Acocella, research scientist in the Freight Lab at the MIT Center for Transportation and Logistics, and David Gal, Vice President of Product and Engineering for Connected Equipment at Samsara, which has been working with the Freight Lab. Good afternoon, everyone. Thank you for joining us today. We're gonna shake things up a little bit. So we're gonna something that we've never done at Beyond before, which is a live podcast recording. Surprise. The topic at hand today is gonna be about AI's transformation and limitations in the supply chain, cargo theft, dynamic pricing, kind of a whole gamut. And to walk us through that, we have several wonderful guests, and we have Dr. Angi Acocella from MIT. We have David Ga, who is our VP of Connected Operations and the mastermind behind the tracking label that you all have heard about. And last but not least, we have the famed tech reporter, Jason Del Rey, from Recode, Fortune, and now his own media property, "The Aisle." So I will hand it off to Jason.

- Well, I'm so excited to be here for many reasons, and I appreciate Samsara having me here. Two brilliant people to have this conversation with. Just a little bit of background about myself. I've covered as a reporter and an author e-commerce and retail for about a dozen years, largely the intersection of technology and retail. A lot of time that's meant focusing on the consumer side of things. Now at my new publication, "The Aisle," AI's impact on online shopping, on recommendations, on consumer search. But what often is dismissed among the sources and companies I talk to is the hard parts happening behind the scenes once you click Purchase and once you step away from that, maybe that AI agent giving you a product recommendation, and you go on and actually complete that before it shows up at your door. So this is gonna be a fun conversation. Angi, I'm gonna start with you real quick. You just came out with some very important research that I think would be relevant to a lot of people in this room. Why don't you give us sort of the top two or three major findings and then we can jump off from there?

- Great, so just a little bit of background also, Angi Acocella, a research scientist at MIT's Center for Transportation and Logistics. What we do at CTL is we work with companies to try to understand what are their issues within transportation supply chain. And one of the things that kept coming up for me was AI, right? I had this question of, well, what's actually happening on the ground? So there's a lot of talk about these end-to-end solutions. And so because we have great interactions with companies, with partners, I decided, well, let me go out and ask. And so I interviewed a number of folks from all industries, right? From manufacturing, food, retail, and those that kind of have their own assets in transportation, those that outsource 3PLs, 4PLs, all sorts. And I was asking them, what do you do with technology in your transportation process? So from the procurement to the forecasting to the actual assignment of providers to measuring and tracking and things like that, and then closing the loop and seeing how that all ties back. And so we can get into some of the details, but I think that a couple things that came out for me was the implementation of digital technology. So it wasn't specific to AI, it was kind of all technologies. So it's not this big transformational process that a company's gonna go through and say, you know, we're gonna change everything. It was very much problem-specific, and here's a problem within our processes. How do we implement some type of technology that automates it or improves some efficiency? But the second-

- That's actually a refreshing point, by the way. Someone who spends a lot of time focused on consumer technology companies where often, we see even the best companies develop products in search of a problem that may or may not exist.

- Yeah, 'cause I think a lot of folks are wondering, you know, should I be investing in this hype? Is it a hype? Or is it something that's really gonna add value? And so being a little bit more cautious about how they're implementing. The second thing that was also somewhat surprising was that, you know, Excel email is still predominantly the way that a lot of transportation is happening. And so the question is, why is that? And I think currently, that's because everyone knows it, right? So that change management is very slow today.

- And if Excel works for you, it works for you, right?

- Right.

- Any Excel lovers in the house? Yeah.

- Yeah. No, you could be proud . Okay, go ahead.

- Yeah. The last thing was that, you know, the biggest bottleneck still is data quality, data aggregation. Data is siloed across different functions within the company, and who owns it? How are you filtering for things is still the main thing that's coming up across all of my interviewees.

- I don't know if you have any reaction to anything Angi just talked about there. We're gonna dive into some of the stuff you're doing here specifically, but I'm curious for your thoughts on any of those takeaways.

- I mean, I think we see that with our customers, not from a research perspective, just from an anecdotal perspective. Our customers are busy doing really important work. It takes time to adopt. So being practical is kind of at our core ethos. When we listen to customers, you kind of hear about a dynamic range of where people are at in this journey. The folks that are getting started, they've just digitized today. There are folks that are, you know, pushing the boundaries of what you can do with AI, but it's very tactical, and it's to solve a specific problem. And I think when we build products, we do this in concert with them to make sure that we're actually solving a meaningful problem because anybody can build a product, but building the right product's much harder. We see it across the journey, not just from sort of digital transformation, but, like, which assets are you gonna transform? And then what problems do you solve for those assets? And then, okay, we solved this one. Now how do we go to the next one and the next one? So it is definitely not a step function.

- There's so much talk in both the public markets and just technology in general right now about productivity, right? Productivity ROI when you're talking about AI investments. And that often leads to discussions of layoffs, cutting staff, finding more efficiencies. I'm more curious, like, where humans still have value. I'm not just talking society at large, which is another question, but in businesses. So I'm curious, for companies that are integrating some AI solutions, like, where is the human just paramount and not being replaced?

- Yeah, I think that was one of the main takeaways, was that humans are still in the loop in so many ways. So first, I'm seeing AI technologies being used as kind of an analyst tool, not to replace them. It's helping understand, well, what are the things that can help me understand my job better? It's helping be better at looking into the data or summarizing. But in terms of actual decision-making, that's not where AI is gonna be, at least today, is not being implemented. So there's these critical milestones, and at critical steps, folks really wanted a human still in the loop to say, "Okay, am I gut-checking what the solution is saying? Or am I gut-checking what the recommendation is?" And there was also this element of making sure that what the AI is recommending actually makes sense to the people that it's being recommended to. So I was talking to one manager, and I remember he said this solution that was being provided to me showed, you know, seven different solutions, and I was like, "Don't show it to my people 'cause they may think you're telling them to do all of this." So they're still trying to figure out how AI can help make decisions, but still have that human be the one that makes the decision.

- Was there any sense in the surveying of folks being overwhelmed on where to start? And if so, I'm curious if there are any takeaways about, you know, best practices.

- Yeah, it was find the problems that you need the most help with, right? And see if there's a solution there. The other thing was there's certain areas and, you know, I was specifically looking at transportation, but there were specific areas where some type of digital technology made sense. So things that are easily automated, right? You can think about in kind of the transportation in your shipments, right, across your network. There's about 80% of the lanes that only cover, like, 20% of the volume, right? That's difficult stuff, right? That's where there's maybe a manual input that's needed. That's where there's inefficiencies, uncertainties. That's where AI automation, auto-tendering, things like that are working really well. And so finding areas where it's a process that's pretty well known in your organization, that's maybe a good area for technology.

- Dynamic pricing. Big conversation point on the consumer side of things in retail and e-commerce, obviously, with some companies. I don't know if any of you follow this, but a company called Instacart, which is grocery delivery. They got sort of caught a few months ago in doing dynamic pricing. Same product, same store, different customer, different price, different flavor of this, obviously, in the B2B world. But what did you find or what came up in conversations around where dynamic pricing does and does not make sense right now?

- So there's a couple elements here. I think the last 8 to 10 years, what a lot of our work has been on at MIT has been what transportation should go through a contract, right? So that just very stable demand, contracts make sense, and what should be more of a dynamic, kind of transactional, spot-type of transportation. And we've talked about portfolios of spot versus contract for forever. Now we're starting to think about the stuff that should go to a spot or dynamic price. Typically, it's like I said, that you know, low-volume, inconsistent lanes, those, you know, 80% of lanes that's low volume. That should be a more dynamic price. Now, the question today is, well, what are the mechanisms by which you should be having a dynamic price? So actually, this is some new research that we're gonna be doing over the next few months and having some roundtables at MIT about is, from the shipper or the carrier perspective, how do you actually utilize potentially agents or things like that to have an algorithm that backs what your pricing might look like? So should shippers be setting the price? Should carriers be looking at the price? Auction mechanisms, things like that. So I don't have an answer for you now, but maybe six months from now.

- Anyone using any technology like that today?

- Yeah.

- No? All right, maybe you're gonna-

- Yeah, we'll talk.

- Invite some to your round table.

- Yeah.

- Okay, David, let's talk a little bit about tracking label announced. People excited about that?

- Yeah.

- All right, wow. We're ready to party. Tracking labels are for everyone. Why was now the right time versus a year ago or two years from now?

- So tracking labels, just a reminder, it's a Bluetooth-based disposable tracker versus shipment visibility. And I think tracking shipments is not new. People have been tracking shipments for a long, long time. Today, the state of the art is roughly barcode scanning. So a cross-dock or a loading dock, somebody or a machine literally scans a barcode and we know where that thing is. And the challenge is that those barcode scans can be super sporadic, both in time and space. So you might have, you know, most recent barcode scan from several hundred miles ago and several hours ago. And so the question is, like, where is it right now? Because I have a critical shipment that needs to show up. The landscape kind of looks roughly like RFID on one side of the equation and cellular connectivity on the other side of the equation. And RFID is super appealing for certain applications. Very low-cost tag, so everybody gets excited about that, but there's a CapEx component here, which is really, really burdensome. And frankly, I think only maybe the top one or two customers can really go implement something that's leaving. Notably, UPS just put a really tremendous capital expenditure towards RFID scanners. Nonetheless, RFID still has these challenges where you have to be in close proximity to an RFID reader in order to have visibility. And so one of the challenges is shipments fall off the back of the truck literally and proverbially. And so the other option has been-

- I'm from Staten Island, New York. If you know anything about Staten Island, we know about the proverbial part. Yeah, sorry.

- No, the other is the spectrum. There's cellular connectivity in GPS-based devices, which are ubiquitous and robust, but expensive. And so people have to be super judicious about where do they apply that technology. And so Bluetooth has been something people have kind of naturally gravitated towards because it's relatively low cost. The challenge is that you need to have a network. And what's happened over the past couple years is that Samsara's network has been deployed to real scale. And just for context, our network is comprised of, you know, millions of buses, bulldozers, trucks, trailers that are in residential areas. They're in intermodal yards, they're in airports, on GSC. And so that network is actually what enables a product like the tracking label. So that's why now it's possible, and it wasn't even, you know, two years ago.

- I know how much companies love talking about limitations of new products. But we have to go there. Like what should a customer who's thinking about engaging and using these tracking labels like know about where maybe they do and don't make sense?

- Yeah, there's a couple things. So first of all, the product is still warm, right? It came out of the oven yesterday. So whenever you launch a new product, you're gonna learn new things and they're gonna see new stuff. The other one, though, is, you know, we're really excited about this product. It's brand new and like any new technology, it's gonna scale with cost over time. And so at least today, we view this as a product for critical shipments, and we view critical as either really important in time and must be delivered. Otherwise, there's significant downtime and cost. So think about, you know, a data center being shut down because something's missing or super high value and prone to theft. So that could be scrap metal. That could be, you know, expensive jewelry. All sorts of things that you might imagine. Kit Kat bars, most recent one. So that's where we sort of think of the sweet spot is.

- Is there anything you learned during this process of your team building about cargo theft that maybe was either surprising or might help folks who've been challenged by sort of the ramp-up in theft over the last few years?

- Most surprising thing is just how ubiquitous theft is. And frankly, how sophisticated it is and how it's everywhere. You read the headlines about this stuff, you probably get to read about maybe 1% or less than that of what's actually going on in the real world. And the more we dig in, we've spoken to experts on this stuff, like it is a sophisticated operation. There are real organized crime rings out there that are participating. We can call it an ecosystem. Yeah, I mean, I think it really is at this point. So it is sophisticated, organized, and it's tactical. The organizations that are really proactive about this are preparing against that stuff. They're taking measures. Yes, they're tracking their cargo. Yes, they're tracking their vehicles, but they're also importing workflows to avoid double brokering. They're really thinking about sort of the holistic picture here.

- Last thing, for now, on this, what are the keys over the next few years so that the cost drops down, that it becomes something that is not just maybe the most critical items but widespread either inside an organization or just in, you know, across the sector?

- I mean, the key is volume. So it's an electronic product like any electronic products. There are economies of scale. Inside of our device there is a Bluetooth radio, there's a battery, there's some chipsets. Those will come down in cost as our volumes go up. And then manufacturing techniques will also improve over time. So I think we'll see that, you know, for the next two, three, four, five years, we can scale costs, which will be fun.

- Any questions for David? Oh, yeah.

- I actually, I have a question to ask you about.

- Yeah.

- So one of the things that come up from the research was about the data quality and then integration and the infrastructure underlying all of that. And you guys have tons of data, right?

- Yep.

- How are you managing all of that and then being able to share that with your customers to make it valuable?

- Yes, if you think about our typical customer, they've got a fleet of vehicles, fleet of equipment, and we collect a ton of data off of those, whether it's geolocation, fault code information, utilization, whatever it might be. And we really build on top of that. So we aggregate that data into things that are useful. Then nobody wants to see an individual GPS data point. They wanna understand utilization or productivity of that asset or health of that asset or whatever it might be. So we build that out natively in our cloud platform, and that is sort of interesting new innovation over the past couple years has been that we've been baking more and more customer customization in with Hector AI into the product. So the customers can just go answer their own questions and make sense of this. So you can imagine in short order asking which shipments are at risk because of the vehicle it's on has an active fault code kind of thing.

- Yeah, so you're taking care of that background data.

- That's right, yeah, I mean, our objective fundamentally, we want our customers to enjoy their morning cup of coffee without grinding away at other things and multitasking. And we want the system to go handle the grinds so that they can do what they're great at.

- Question over here.

- My name's Morgan Parrish. I work for Spartan Companies. We're an energy construction company. We're building data centers and oil and gas for BP. A lot of things that we build that are out in the middle of nowhere, except this network is really helpful with the 5, 600 trucks we have in that area. When we install some of our equipment, when we leave those job sites out in West Texas, if you've ever been there, no one's there to protect it. And so this would be great to be able to do that. Can you be the only ones to read those Bluetooth devices? Or is there some type of protection so that it's not detected by thieves or other groups?

- Oh, I see.

- Is there some technology to only allow those that are printing and scanning and activating to read those?

- Yeah, so I mean, the way we think about Bluetooth security in general is we take a pretty different approach to consumer-grade Bluetooth. So we baked in security and anonymity into that Samsara network from day zero, you know, two years ago. Ultimately, if something's in the air, a motivated thief can listen for it. The question is, how do you figure out what that is and trace it down? All those kinds of things. And we've taken significant measures to make that pretty difficult. You know, what I can say today is we have asset tags that use the same technology. Typically, when we see those things being stolen, it's inside jobs. So it's folks that know actually where those were installed or what they're looking for rather than a third-party kind of independent thief. So, you know, unfortunately, theft and loss is a cat-and-mouse game. We think we're maybe a couple steps ahead right now, and we're gonna continue to invest to make sure that we stay a couple steps ahead.

- I just have one follow-up to that. So someone may be able to detect, if they're committed enough, the general area, but knowing exactly, or is that not what you're-

- Well, I'm saying that, you know, if you take out a wireless scanner and you listen, you will see radio transmissions in the air no matter where you are, no matter what the device is. There's an electromagnetic wave in the air, you can detect it. How do you go pinpoint something? How do you know what that thing is? How do you determine whether that's a Samsara label versus anything else? That's a different question. So we try to make that step as challenging as possible.

- My name is Kelly Sutherland. I'm here at Samsara, but I really, I have a question for David, and that is, what are some good use cases that you've seen from customers that's been in beta for a little bit? So what are some good use cases where you've actually been able to find something unique or that was helpful for our customers?

- Yeah, we've talked to chemical distributors that are looking at one-way shipments of mission-critical chemicals. If they're not at the job site the next day, things get shut down. So that's been a good use case. We've seen super high-value shipments where people just need visibility on them. That's been a good use case. We've seen pharmaceuticals. We've talked about biological samples. You can imagine if that biological sample doesn't make it to the lab on time, you have to go take another biological sample. So real cost associated with those. We've seen auto transport. So car carriers, for example, they wanna track a Ferrari from, you know, dealership to dealership, something like this, and there's theft involved. So literally, I mean, I won't say we've seen it all. I think we've seen 1% of what we've seen, but we've seen a lot. But really, it is all about criticality. It's like, does time matter or does dollars matter? And so what's at risk?

- I had a quick question for Angi Acocella. When you did that study and you identified data being one of the constraining factors, right, in space, did you control for what kind of technology or systems they had in place already? It's purely selfish question from a Samsara extent. Is there a differential?

- I love being upfront about that.

- Yeah, no, that's fine. Yeah, so this was basically just interviewing folks and asking kind of generally what's going on. So not a deep dive into the types of technologies. But I did hear it from across the different interviewees that were from all sorts of different companies. So it can come from the transportation managers, it can come from different industries, it can come from the transportation providers. And in fact, we were talking before this session, before I was a research scientist at CTL, I was working with the Port of Rotterdam on digitization of their operations, and the data issue was a problem there too, right? So it's been a consistent issue for digitization in every setting that I looked at.

- Wanna switch gears just a little bit in this, you know, last 15 or so. In the coverage I do at readtheisle.com, my new publication, there's a lot of talk around, on the consumer side of things, the idea of agentic shopping. So AI personalization that will eventually know you so well that it will help you make purchases or maybe carry out purchases. There's a lot of discussion about that among the retail companies and e-commerce companies I talk to and the AI labs. Not a ton about what happens if those systems ever meet sort of agentic supply chains or really AI-driven supply chains with few humans in the loop. I'm asking us to future-cast a little bit. I'm curious what sort of problems folks who sit more on the consumer side should be thinking about as they're trying to, maybe in the most forward-looking organizations, plan for these days.

- For me, this is sort of a question that is answered by, if we look back to, like, blockchain and how that was supposed to, like, revolutionize supply chains, right? I think the issue when it comes to supply chains and transportation is that there is a physical truck that needs to show up. There's a human that is driving that truck. There's information that needs to be shared with them. There's real rubber on the road, right? And I think that's where so many challenges can come into making this their sticking points. So thinking that, like, what are all the steps and the complexities that make it such that the product that you're ordering has to come from a warehouse, has to get picked by a warehouse worker, put into a box, right? Then a driver needs to show up, gets to put it on the truck. It needs to not get lost along the way. So many things, that physical aspect of it, it's not just a digital computed type of transaction.

- I think you asked a question earlier about AI and humans, and I think this is a good example. Like, we deal with physical AI, which is a little different than, I think, consumer ChatGPT stuff. Like, Anthropic put out one of these star charts of showing the industries that are gonna get disrupted by AI. And if you got to look at where they're forecasting, it's software engineering, finance, sort of maybe carpeted office jobs. Physical AI is not going away with humans. And I think this is kind of the human-AI interaction that we sort of foresee and how you can scale teams and actually augment the people to do more and run these operations more efficiently, which then, of course, helps the supply chain and everything else around it. But this is the kind of example that I think we sort of forecast.

- Any other questions?

- Well, I was going to say a little bit about the driver experience. Our perspective, it's helping them be safer, be more efficient, do their jobs better, right? But there is an element of, you know, they're humans, and on the one hand, it's a very isolated job, but it's also sometimes drivers like that they have the freedom to kind of do what they want to, right, when they're out on the road. And so, you know, finding this balance between how do we make sure that it's helping them, they understand that it's helping them, while also recognizing that maybe there might be some pushback.

- You know, when we first put out dash cams, I don't know, eight years ago, I think a lot, and maybe customers in the room had this experience, but drivers would look at that dash cam and say, "What is Big Brother doing spying on me?" And then within that week, invariably, somebody's exonerated from an accident. And immediately, the driver sentiment shifts and understands that this is not here to spy. It's not listening to them. It's here to help and get you home safely to your family and protect your communities. And these organizations, I mean, I think as a consumer, I don't think safety's appreciated the way that our customers appreciate safety. And you're dealing with 10,000 drivers, it is a completely different ballgame than when you're just driving your Honda Civic to and from. And so these organizations have tremendous responsibility in their community. I mean, tremendous. Their names are up everywhere. Like, these are, you know, stewards of the citizenry. So they take it really seriously. And I think ultimately, the adoption curve on this stuff has been extraordinarily quick. Extraordinary, you know, and people talk, and drivers go to different organizations. And I think at this point, I'm curious to hear if anybody else has had this experience, but we see, as soon as people understand it's for the safety of themselves, it's a no-brainer.

- I think we have a question up here.

- Thomas Watson, FreightWaves. What are the things looking at adoption rates of this technology? As we're starting to see, at least on our side, the freight cycle turn, the demand for drivers is gonna continue to rise. Do you think that this will become part of a fleet's toolkit as a differentiator? You know, for example, it used to be you get, like, a brand-new truck, you get home, but are drivers now going to start evaluating these companies on these little things where, oh, yeah, you're looking out for me as well because it's my points on my license in addition to your points. Do you think that's gonna be one of the trends that we're gonna see moving ahead compared to times where, when we had a crunch for drivers, that wasn't on the table?

- I think so. I mean, I don't know exactly how it will manifest within the drivers, but we have driver recognition. I mean, I think one of the challenges that I hear about from our customers is, how do we maintain the best drivers in our organization? How do we coach kind of the middle? And then how do we make sure that for the super risky drivers, we deal with one way or another? And I think that, you know, we do a lot of positive recognition, whether that be a monetary award, whether it be, you know, a gift card or just a shout-out or a kudo. And we've seen a lot of customers do a really good job gamifying positive behaviors and therefore keeping a culture of safety where it's really rewarded and valued.

- I'll also add, I think, Thomas, you make a good point that we're going into a new tight market, a constrained market where the reasoning for an underlying structure of market's a little different than it has been in the past, where it's very much driven by a capacity issue, right? The capacity side, the supply side versus the demand side. And so, yeah, being aware of what is actually gonna maintain drivers, retain drivers, I think is gonna be at the forefront of a lot of carriers.

- Any final takeaways, either based on your research or discussion we had, that you wanna make sure folks are thinking about or as they go back to their organizations next week?

- For me, I think one of the things that I'm coming away from both this session today and yesterday's events and the research is around, what are the real problems that need to be solved? And so technology, AI can really help with that. And where I think the future is going is a little bit of not just how is AI and are these technologies a helper and, you know, helping me understand the questions I already know to ask. This was one of the things that came up a lot was, you know, the future looks like, how is AI going to help me understand questions that I didn't even know to ask? So how is AI-

- So prompting you versus-

- Yeah, more question-based intelligence, right? Things that, you know, maybe I don't know to query, what I don't know. And so that's, I think, where the future is gonna be going.

- Needs to be a lot of trust there. I'm curious how much interactions on the consumer side bleed over into someone's trust of systems.

- I think if you were to ask Gemini, "What does supply chain look like?" you get a photo of a container ship with a bunch of containers on it, and that's what people have in mind as supply chain. I think what I've learned over the past year in building this product is supply chain's actually much more complex than that. We have customers that deal with supply chain that could be taking building materials to a job site. That's supply chain. And if you think about kind of the steps, there are vehicles, there are humans, there's the maintenance of the vehicle that has to get you there. It's not just about planning; it's actually about the full operationalizing of this. And I think what's kind of fun and exciting to sort of think about, not in the long term, but I think really in the next few months, is how do all these things come together to really give you a holistic picture of how these organizations are running, how do you actually execute the supply chain? That could be, again, you know, pharmaceutical production to hospital, but it could be GPUs to a data center, and it could be car transportation, anything in between. And so you get to start to see how, okay, how many vehicles do I need? How many drivers do I need? Which ones need maintenance? How do I actually ship this package from point A to point B? And, you know, we think packages, we think 10 boxes, but we should really be thinking about so much more than 10 boxes. We gotta be thinking about, you know, copper wire and all this kind of stuff. So I basically think we're kind of at this confluence now where you have visibility on all these things. You now have this AI that can sit on top and make sense of it, and so I'm super excited about what six months from now looks like.

- I think that's all the time we have. You guys almost hit the zeros on the dot. Let's clap it up for my two fine guests that I- Thank you, guys.

- That wraps up this episode of "Supply Chain Frontiers." A big thank you to Jason Del Ray, Angi Acocella, and David Gal for joining us. "Supply Chain Frontiers" was recorded at Samsara Beyond 2026. Our sound editors are Dave Lashinsky and Danielle Simpson and David Benjamin Sound. Our producer is myself, Mackenzie Berry. Be sure to check out previous episodes of "Supply Chain Frontiers" at ctl.mit.edu/podcast or search for us on your preferred podcast platform. I'm Mackenzie Berry. Thanks for listening, and we'll catch you next time on "Supply Chain Frontiers."