The Freight Pod
The Freight Pod is a deep dive into the journeys of the transportation and logistics industry’s brightest minds and innovators. The show is hosted by Andrew Silver, former founder and CEO of MoLo Solutions, one of the fastest-growing freight brokerages in the industry. His guests will be CEOs, founders, executives, and leaders from some of the most successful freight brokerages, trucking companies, manufacturers, and technology companies that support this great industry. Andrew will interview his guests with a focus on their life and how they got to where they are today, unlocking the key ingredients that helped them develop into the leaders they are now. He will also bring to light the fascinating stories that helped mold and shape his experiences.
The Freight Pod
Ep. #82: Alan Holland, Founder & CEO, Keelvar
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Freight buying has a dirty secret: most bids do a great job of finding the lowest number and a terrible job of finding the best outcome. When shippers reward price without measuring service, good carriers get punished, bad actors slip through, and procurement teams spend weeks in spreadsheets trying to guess what suppliers actually want.
We sit down with Alan Holland from Keelvar to break down a better approach using sourcing optimization, mechanism design, and AI agents. Alan explains how incentive compatible auctions can pull truthful preferences to the surface, why package bids and conditional discounts create space for small fleets to win the lanes they can run best, and how the Google Ads auction is a surprisingly useful model for modern freight procurement. From there we get practical: connecting transportation management system performance data to the sourcing event, weighting on-time performance with price, and designing feedback loops that reward reliable execution instead of “cheap and shaky” promises.
We also zoom out to the bigger AI shift. Massive compute, LLMs, and code generation tools like Claude Code are changing how software gets built, which means logistics technology will evolve faster than most teams are ready for. We talk about risk, uncertainty, penalty cliffs, rebate targets, and the long-term “holy grail” of multi-shipper combinatorial exchanges that could unlock network-level efficiency.
If you care about freight procurement strategy, logistics automation, AI in supply chain, or the future of brokers and carriers, hit play. Subscribe, share this with a friend in logistics, and leave a review with your biggest question about where AI agents help most.
Follow The Freight Pod and host Andrew Silver on LinkedIn.
Thanks to our sponsors:
Stuut Technologies: Your AI coworker that collects your cash automatically.
https://www.stuut.ai/
Cloneops.ai: Not just AI. Industry-born AI.
https://www.cloneops.ai/
Rapido Solutions Group: Nearshore solutions for logistics companies.
https://www.gorapido.com/
GenLogs: Freight Intelligence on every carrier, shipper, and asset via a nationwide sensor network
https://www.genlogs.io/
Sponsors And Quick Show Start
SPEAKER_01Before we get started today, I want to give a quick shout out to our headline sponsor, Stute Technologies. Every week on this show, I talk to operators who are crushing it on the service side but still chasing down payments. That cash flow gap is brutal when you're trying to scale your business. That's why I wanted to tell you about Stute. Stut is the AI platform that does accounts receivable work instead of just assisting with it. Their CEO, Tarek, actually came from TQL, so he gets how logistics companies operate. I've known him for years and I truly trust what he's building. What Stut does is simple. Their AI actually collects your receivables for you, not some software that just helps you chase the payments. It literally does the work. It's up and running in days, not months. As an example, Bishop Lifting is using Stut for a 35% reduction in overdue invoices, millions more in working capital, and a reduction of manual work by 50%. If you're tired of your AR eating up time and cash, check out Stute.ai. That's S T-U-U-T.ai. Tell them you heard about it on the Freight Pod. They'll get you collecting in days, not months, so you can focus on actually running your business. And let's get this episode started. I'm Mr. Ellen Human from across the pond. Welcome. How are you, sir?
SPEAKER_00Very well, Andrew. Thank you for having me. And where are you joining us from? I'm based in Cork on the south coast of Ireland. So it's just after St. Patrick's Day here, so I'm good for well-rested.
SPEAKER_01Are you feeling the after-effects of a well-celebrated holiday?
SPEAKER_00I'm I'm looking at I've got young kids, so uh just celebrated differently than when I was younger. Let's put it that way.
From AI Research To Freight
SPEAKER_01Yeah. Well, I'm excited to have you on the show. Uh your company Keelbar is one that I'm fairly familiar with from my days as a freight broker working with a ton of the enterprise shippers who use your product. Um and there's just so much that we could talk about from running a business in Ireland that serves uh thousands of global enterprises across the world to um kind of AI in today's world, which I know is something that you're very familiar with. So to get started though, I'd love to learn as much as I can about you, Alan, the individual. Um take me back, like you know, before you got into the freight world, um give me some of your your your journey of how you got here.
SPEAKER_00Um so uh I guess I I started off my career, uh I I did a degree in computer engineering and then went to an AI research lab. So I I I was lucky that there was one of the world's top AI researchers, set up a new lab uh here in Cork, right on our doorstep. So he he was he was Marvin Minsky's student, right? So he he was right at the he came from MIT and uh one of the godfather one of the godfathers of AI, right? So he was bringing the world's best top AI researchers here with him, and I was the first PhD student, so I really just got lucky um uh in in that setting. But I wasn't always I don't think I was your classical academic. Uh I came from you know my family had their own business. There was a chemical, they had a they had their own start startup, it was a chemical company, and you know, I used to have to fund my my PhD student days by working night times for them, and David as a as a as a as a poor PhD student, and uh so I kind of got to see what the freight world looked like by seeing how bulk chemicals were being moved around and how much it cost and how essential to profitability it was to do that well when you're a manufacturing business. So by day I was in an AI research lab collaborating and and having trips to Palo Alto, collaborating with Google on auction platforms for for selling sponsored search. So you got to see how AI was being embedded in monetization of the world of e-commerce, collaborating with Amazon on how to price uh goods they were selling in their marketplaces. So very sophisticated e-commerce uh solutions by day, and then by night helping my parents respond to procurement teams who are buying chemicals and trying to build in freight costs into your pricing. Um so that's how I kind of I'd say that's the yin and the yang, if you like, and then night and day, literally, of the world of e-commerce and B2B commerce. So kind of got an insight into just how different they were, and in some ways, how e-commerce had this huge advantage in having kind of uh just a green field site in the early 2000s. How are you gonna design your business strategy in a way that's efficient? Whereas B2B was had solutions that were designed to digitize paper-based processes. Um and I could see that I've got to solve this. This this is this is an itch. I'm gonna have to scratch. And so I started scratching it by looking at okay, what what should the world of B2B doing? You know, given that it is more complex than e-commerce, you know, you're not just selling books or sponsored search advertising slots. You're you're selling B2B, you're selling dangerous chemicals or it could be perishable fruit, or what you know, whatever your line of business is, it's got different um constraints, and there's a lot more constraints on uh the physical world that you've got to respect. So, how do you bring speed and efficiency of e-commerce to the real world? Um atoms, things moving around. Uh so that became my life's work, really. Um, first of all, starting researching how should it be done, uh, and then seeing it not done well, getting frustrated looking at software vendors who are not doing this well and saying, uh, I'm gonna have to go out there and show them how it should be done. So that's that's how Kielbar started.
SPEAKER_01So so this was the early, this was like around 2010. Like, what was the timing of all this that was going on between you and the other?
SPEAKER_00I did a PhD in 2002 through 2005, and I became a lecturer and I was teaching AI AI in university, and um but also starting to collaborate with big business. Like I was starting to publish papers on saying, look, this is how you should run more advanced kind of auctions for buying vehicle fleet or electricity uh or um facilities management. And then I started to get companies knocking on my door, and they were saying, You can see how you helped the city council in Cork get their six million on vehicle fleet spend down to five million by just having package bidding and conditional discounts. Uh one company had hundreds of millions in vehicle fleet, but there was and they wanted to do the same thing. So it was a very large corporation, one of the world's top 50. Um and so I did a pilot project with them when I was in university and showed how you you could actually get a six billion spend down to five billion. And uh Corksy So project number one was getting six million spend down to five million, project number two was six billion. Six billion down to five billion. So it was at the same, I would say, the same percentages in terms of efficiencies by just allowing small players play to their strengths, medium-sized play to their strengths, and the large players play to their strengths. So if you get allow everyone play to their strengths and win what they really want to win, not win stuff they're not they know they're not good at and actually prefer not to do. But oftentimes they just feel feel beholden on to putting their arms around everything so they can keep people out. So if you can kind of design your negotiating mechanism well, you can actually make space for small, medium, and large players. Because um I in the Cork City Council's case is a good example of companies just focusing on electric vehicles. This is early days, so it was like golf carts and these mini cleaners that would go down small side streets and sweep. And that's all it is. But they've never been able to win business with the city council. Uh and now they suddenly did because you're kind of breaking it down into small and small, uh smaller contracts and allowing businesses to carve it out, like the tipping trucks are do that, uh, commercial vehicles and the vans and uh cars and specialist players would do that. And and I would say on balance, everyone was happy. Suppliers were happy, the council was happy, and the same thing with uh these very large corporations where they they can't have that level. When you're buying when you've got hundreds of thousands of items you're buying, you you just can't find form a human relationship with suppliers in the same way, you can't understand exactly what their preferences might be. So you you've got to set up a system so that they're allowed to play to their strengths, and and even more importantly, tell you what they don't want because very often you end up bundling, and this is a kind of I would say a trap that a lot of procurement teams fall into. They end up bundling the big bundles in ways that suppliers go, oh, I love A, B, and C, but why do I have to do D? I'm no good at D. And then they prefer to kind of resell it to somebody else. But what ends up happening is the just the friction of doing that is that it's like you they just do a bad job and just say, that's it.
Mechanism Design For Fair Bids
SPEAKER_01What's really interesting is that example you use, you know, while I know nothing about you know the Cork City Council and that procurement event, the example is really relatable to a freight broker. And just thinking about the number of times I've gone through procurement events where a couple of the things you said feel so true. Like one is, you know, I just remember thinking, I want this shipper to understand what I'm really good at and what my desires are and what I don't want. But and you said hundreds of thousands, even in the case where they've only got 40 or 50 providers, it's really hard for them to have a systemic way of understanding what do I want versus what does he want versus what does that broker want? What are we really good at? So I guess part of my question is you you talk about it like the solution is or the end game is that all the providers, small, medium, large, are getting the piece of the pie that makes the most sense for them. Um, but I guess my question is like, is that a product of creating the right system that allows that information to come to the forefront, or is it about changing the behavior of the players like myself and how we show up into the event?
SPEAKER_00Yeah, I I it's a bit of both, but I would say the most important thing is because the buyer tends to set the rules, right? Around this is this is the rules of the game. They define what game theater is called, they're the mechanism designer, right? And mechanism design is like inverse game theory. How do I set up the rules of a game? So those that are playing in in their in game theory we call it self-interested profit maximizers, right? Which sounds very uh utilitarian, but that's what its business is, right?
SPEAKER_01It sounds like the perfect way to describe it. That's what we're doing. Self-interested profit maximization. Maybe you could argue it's not necessarily just self-interested, because I know that for my self-interest to be maximized, I need to maximize your self-interest too, so there's some kind of duality there, but you are kind of describing it perfectly.
SPEAKER_00So like that that's the premise of game theory, right? And that I used to teach that in university, right? So it's it's like if you just assume that everyone is going to be a self-interested profit maximizer, then you should set up the rules of the game so that they have an interest to communicate the the truth or is close to the truth. It's what it's what game theorists call incentive compatible. So Google did that really well, right? And this is this is this is why Google data is instructive to the rest of commerce, right? So they set up an auction scheme where let's say you're bidding on legal services in Chicago, right? And you were bidding ten dollars a click, and you would you you could spend this on Yahoo or you could spend it on Google. Now on Yahoo, the rules were you pay what you bid. So you'd you'd be paying$10 per click, and then you'd figure out, oh, the second highest was$9. So I'm gonna put it down to$9.01. But the second guy saw that you put it at$901, he's going$9.02, and then you go$9.03 every day, right? So everyone is coming back every day changing their bids. So there's lots of noise in the Yahoo system. What Google did was said, we're not gonna charge you ten dollars, even though you bid ten, we're actually we're gonna charge you the second highest. So we're gonna charge you the nine, even though you bid ten, right? So this is an example of an incentive compatible auction where you could just set and go, look, 10 is actually the maximum I'll be willing to spend. So what advertisers did was they said, Yahoo, that's a noisy mess. The cost of actually running our ad campaigns there is is high. So I'm gonna shift it all over to Google. So Google defeated Yahoo and all other search engines. They had good quality search results, but that's not where they really defeated the rest of the market. It's the auction protocol they used, which is a smart auction protocol, which was uh it it revealed truthful preferences for small, medium, and large players better than the others. And what they also did was they gave incentives for quality adverts. So they ranked who would win the top based on your your bid amount times the expected click-through rate. So now if you had an ad for let's say it was marketing services, and you're also you're just bidding on legal services, people wouldn't be clicking on it, so it would fall down the rankings. So it even if you were paying a high price per click, it would still fall down the rankings if it wasn't relevant. So it's a little like say if a freight provider, you you're being very competitive in your pricing, but you're not delivering on time. You're not your performance is poor. So you need a mechanism when buying freight so that you shouldn't have to you shouldn't have to expend a lot of effort when bidding. Instead, you should make it simple. And if you're good at what you do, you win. But if you're not if you're c if you're cheap, but you're not operationally effective, you don't win. So what you need is optimization of the auction format. So what so I would say our mission has been to bring that type of speed and efficiency of e-commerce to various spend categories in the world of B2B, Freight being among one of the best adopters. Uh I and I think that's because they recognized early on they couldn't use Excel for the big strategic sourcing projects. It would it would break, it was just too slow, time consuming, and error prone. And you couldn't get the type of package bids and conditional discounts. So buy side started of all the spend categories, this logistics teams, logistics procurement teams became the best at applying sourcing optimization. So we took some of those teams who got really good at applying good standards in their strategic negotiations to say, now your agents, right? Your agents should apply, agents should do your tactical bids as well, and do hundreds or thousands of these, all the spot bids, and what they should do is start applying intelligent mechanisms in their negotiations with your um with all your carriers, and you should equip your carriers with bidding agents to respond. So just like advertisers for Google, they're not in every single uh keyword working out what we should bid for legal service in Chicago, legal service in San Diego and uh and whatnot. They they set up rules around their budget, their capacity, um, their strategy. And uh that's the direction of track. We're not there yet in terms of machine-to-machine negotiations, but it's getting there, it's getting there quite quickly. Uh I would say this year there will be um I would say by middle of the year, I'd be disappointed if it was much later in September this year. We'll have quite a number of customers doing very equally sophisticated clearing of um uh freight matching um as Google do with their adverts.
Sponsor Break CloneOps
Toward Machine Led Freight Negotiation
SPEAKER_01Hold up. Wait a minute. Let's give a quick shout out to our sponsor, cloneops.ai. Logistics moves fast. There are calls to make, updates to send, documents to track. Your customers are always waiting. Cloneops.ai helps teams stay ahead with a platform that can be white labeled and a marketplace of AI agents built for real logistics operations. Inside the Clone Ops marketplace, you'll find agents for tracking, documents, after hour support, finance workflow, fraud prevention, and more. These agents plug into your workflows and existing systems to handle the repetitive tasks, whether it's making calls, sending emails, replying to texts, gathering info, or writing updates. Your team stays in control with real-time visibility into every interaction. And because the AI is always turned on, operations keeps moving even when your team can't. A standout is carrier screening plus voice ID, which ties a voice to a DOT number and alerts your team when that same voice appears across different or multiple DOT numbers, helping catch fraudsters before they cause damage. No one else has this. And for companies building a long-term AI strategy, clone ops.ai gives you the ability to white label, build, own, and control every AI use case across your business. And this team is run by David Bell, founder and leader who I've known for years, a good friend of mine, and someone who, frankly, anything he touches in this industry turns to gold. I trust him with what he's building here. If you're ready for automation that supports your people, visit cloneops.ai. Now let's get back to the show. Your expectation is, and I'll I'll give you some extra time, by the end of this year, you expect that a lot of the shippers using your platform essentially will be negotiating transportation with minimal I'm trying to use the right words here, minimal human interaction. Minimal tactical interaction.
SPEAKER_00So minimal tactical interaction. Like if you look at, say, the advertising agencies on Google, they're every day they're looking at their the advertising platform and figuring out what their strategy ought to be. But they're not in their bid on this one thing or bid on this other thing. That that they shouldn't be that low down in the weeds. They should be monitoring their utilization rates and their operational performance relative to margins and deciding what's the right amount of business to be winning with these various customers. Um whether they, you know, and even strategic feedback from the system as to well, you should be you should be increasing the size of your fleet here. And others should be decreasing the size of their fleet, right? So there's uh there's I would say there's advice that should that could be given back to some of these carriers around right sizing and or relocating. There may be some regions where margins are higher and and they don't even realize it. This is the thing that you're complex networks. So there's a lot of businesses flying blind, and they're just looking at the aggregate performance of their fleet rather than optimizing uh where should our trucks be based, which customers are most profitable, which are rewarding. It's sometimes very frustrating to be really good at your business. And you don't get rewarded. And some customers are bad customers, so they're bad buyers out there who just want the lowest price all the time. Well, sometimes it's better to just cut them. And if if they're not rewarding you for being good, you go elsewhere. And go go to customers that build in feedback loops in which good carriers get rewarded. And it's almost a case of demanding that from them. They got to do that for you because if they're expecting good performance, they need to you need to be rewarded for that good performance. And in my experience, most suppliers they really want to do a good job, and they hate it when you set up a mechanism that just rewards the lowest price. Because they know that a shady character who's willing to do a poor job, it's going to win that business. And actually, they'll do more damage to the shipper and to the market. And that really it's up to buyers to be building in these kind of feedback loops where you're connecting your buying system to the transport management systems that are operating within their business so that you know who's good, who's bad, reward those who are good, punish those who are bad, and be happy to give more margin to those who are good, because that's what really drives your business success. So that's that's what we feel is efficiency in a marketing. That's how you make something win-win, and that's kind of a core value of our business. How do we make things win-win?
Linking TMS Data To Sourcing
SPEAKER_01So you just said something that I was actually thinking and was going to ask. Um, but there's a lot that you said that I resonate with. First of all, you know, when when I was running my own brokerage, you used a term earlier, or uh you said something earlier that that hit perfectly, which was how we ran our business, which was like we were under the impression that every one of our shippers was responsible for making the rules to their own game, which is what you said. And those are the exact words I used to use with my team. Is like all we can control is the service and the price we provide to these shippers. We don't make the rules for them. They make the rules and we decide if we're willing to play by those rules. It is incumbent on us, it's it's more important than ever, that we are able to learn from the outcomes of those rules and how we play within them. Because if we're gonna provide this premium level of service to a shipper and they're not gonna reward that, and they're gonna keep taking good lanes away from us because someone came in and offered them cheaper, we need to stop providing that level of service to them or provide the level of service that they want. Provide if they just want the cheapest price and they don't care that we hold that price for three months or six months, they give them the cheapest price for today, knowing that we'll change it tomorrow, if that's the rule that they want to play by. Now, I say all that to say this. Um it has sometimes been really challenging to understand the rules the shippers are playing by because they seem to change them, or it's the trend, there's not a necessarily transparent communication back and forth. You don't know why all of a sudden a load disappeared. And one of the things I was thinking about was how important the technology is to create a system where both sides have transparent communication and transparent understanding of the rules. And the thing I was thinking was like, well, Kilvar in my mind is a sourcing system, but isn't necessarily, and this is where I hope you can tell give me more info, um, isn't necessarily clued in on whether Molo's the best provider or Arrive or Knight Swift or whoever. And so I'm curious, like you mentioned the the sourcing system connecting in with the TMS, and I'm curious, like, has that been an evolution in in the business where now when sourcing events are happening, there is a lot more tied in information and data that tells you how to stack rank players and who's good at what?
SPEAKER_00This is this is I would say only it's it's a relatively small number of shippers that have started doing this. And this is where we're trying to get the market to. We've got to hook up the operational performance metrics with the sourcing system so you can kind of jointly optimize right and reward the good the best shippers, right? That's that's the key. They've been trying to do that a bit too manually. And for most big shippers, it's it's just an overwhelming, overwhelmingly large volume of data, and they don't necessarily do it well. And what they need to be more focused on is kind of the policies and aggregate levels of performance rather than a lane-by-lane analysis. Um so but that is that I would say that's a key part of the feedback loop now. So um, and that's that's what's kind of I would say on the buy side, that's what's becoming that's what's happening. It's it started happening a few months ago, but uh as we move through the year, we move to more of our shippers. Some some of them haven't even started using agents yet, you know, to to automate. Um, so there's some laggards who um still running things very manually. Um but it's so obvious now that agents are better at doing this for all the tactical uh kind of fast-paced bids that it doesn't make sense anymore. But you know, some companies just just are playing catch up on that. But the leaders are are integrating with TMS systems and then starting to, I would say, have an enriched view. And that's the importance of having having agents run sourcing events over a source and optimizer because it's the optimizer that can an optimizer is not about optimizing cost, it's actually about optimizing a balance between uh price and non-price objectives. So it's about being deliberate in weighing up the value of on-time performance. And for most big shippers, on-time performance is it's more important than the price. Because if you got things like perishable goods or you've got shelves running empty because you're loaded, you're out of stock, that's that's what that's why it's so important to get that data integrated into your decisions, because they'll be willing to pay a premium, but they need to have, I would say, the calculation engine all hooked up to automate the um consideration of on-time performance, in particular of all the non-price objectives that on-time performance is so critical to many big shippers. Um, and then there's some somewhat durable goods who care if they happen to care less about infantry, you know, so they'll be more price sensitive. But there's many shippers out there that are more time sensitive.
SPEAKER_01How much more challenging is it to understand the like you can't understand the true cost of a lane unless you understand on time performance in some cases? So if you've got uh any vendor of Walmart, for example, it's the best example, largest example, but any vendor of Walmart, if you're not hitting a certain threshold of on time, then you're paying a percentage of a fine that is equivalent to like a percentage of the retail value of the goods. And so I'm curious how your system can create how you can systemically create a process that takes into account that kind of hypothetical. That's like if you're at 95% on time or 98% of time, there's zero fine costs. But if you're at 94% on time, there's a fine cost equivalent to 3% of the retail value of the truck. Like that's a those those are made-up numbers. But I'm curious, is how do you think about that in terms of like creating the right optimization system?
Pricing Risk Under Uncertainty
SPEAKER_00When there's I would say there's and this is I would say within AI, there's most people think of AI as like the LLMs we've got present, but there's there's lots of I would say other pillars of AI around managing uncertainty. Uh, a kind of key pillar of AI has been optimizing under uncertainty. And there's a lot of work done in terms of the algorithms that an AI agent should be applying in when dealing with this kind of uncertainty. So if you've got a cost model for if my on-time performance drops below X percent, I'm going to have this penalty factor of X million dollars or whatever it may be, right? It's um if you're trending below that, now your buying decisions should be different. You need to have a higher level of confidence that the remaining shipments you've got are going to get there on time. So you should be willing to pay a premium. I think people have an intuition, like say, during a sports game, even right. So if um if you're if you're down by a certain number of points in some 10 minutes to go, you're gonna take more risk and you have to you have to shoot three-pointers, or you have to, you know, whatever it is, you your strategy adapts. So you've got you've got these models within the field of AI, like Markov decision process and other stochastic modeling techniques that will inform you let's say who I how uh what you should be willing to pay. It's a form of uh uh insurance, right? So it's to deal with the heightened risk you have now, given that your your on-time performance has been, if you're a carrier, that is, uh sorry, if you if you're um if you're it depends on whether you're the the the shipper uh sorry the or the carrier in these cases and who who is buying the freight, but uh it's you can basically embed these types of algorithms within an agent that knows how to change its attitude to who you deal with based upon performance to date. Um because usually the penalties are based on aggregate performance levels over a certain period of time, and uh there's there's a capability we call campaign optimization, right? So uh let's say on from a buy side, these shippers are also interested in they often have clauses within contracts where if they spend X million dollars with a carrier that they're entitled to some kind of rebate. Now, as the year goes, uh as the year goes, maybe they're underspending and they're not trending in the right direction. So a similar challenge is to know how much you should buy in favor of this carrier in order to get their uh spend above the five million dollars so you can get this rebate. So there's this I would say a set of related challenges around campaign optimization that um again it's it's not similar to uh I'd say Google's AdWords auction, right? Where you've got a budget, someone's got a budget, an advertiser's got a budget, and they want to spend up to$10,000 per month on these keywords. Google's deciding on a daily basis which which of their ads to place. And they'd like to use up the budget, but sometimes they can go over and sometimes they go under, and then they have to adjust their strategy throughout the month. Likewise in logistics, that if there's certain targets that are keys that are key, then you've got to adapt your strategy. Uh it gets it gets quite technical in terms of how you can adapt. There's different algorithmic approaches to model that uncertainty. Um, so what we see happening is different flavors in terms of risk appetite. Some people would leave it very late to adapt, and some people would like to course correct earlier. Um so having controls over those campaigns, uh how how to rebalance and trend in the right direction it's it's it's a matter of taste. And that's I would say agent management becomes a skill then, right? So you have to watch and observe your agent and give it direction to say you're not you're not course correcting quickly enough. Or you're course correcting too fast. We'll sort this problem out in the last week if we if we need to.
SPEAKER_01You know, it's interesting, for one, you know, it feels like you're kind of the perfect person to be doing the job you're doing. Um just given the familial background in what you got to witness with your own family and your chemicals business, and then you know, I I don't know how many people were studying AI 24 years ago, but uh it it certainly is a tenth of a tenth of a tenth of a tenth of a percent of what it is today.
SPEAKER_00Yeah, it's probably less than 500 worldwide. I I I probably named you know first name terms with many of the top 100, let's say put it that way.
SPEAKER_01Let's let's stay there for a second because that's just so interesting to me. You know, you were studying AI 24 years ago. Four years into that study path. What's the right way to ask this? So it's 2006. You know what did you think the next 10 to 20 years looked like for AI and its advancement? And how close was what you thought to what is now what's what we're seeing?
SPEAKER_00It's funny because I uh for uh like I I didn't set up Kilo R till late 2012 because uh I I and we didn't go after agents initially because I knew this is just too it's just too futuristic for most. And then when we started doing it in 2020, it was only the most sophisticated businesses that we targeted uh who were kind of consciously trying to be on the frontier. Um but it wasn't until 2025. It was well, I suppose ChatGPT was a moment, right, where everyone kind of decided, geez, this AI is great, and we need we need more AI. But they also had maybe somewhat of a misconception about what AI was, because that was one branch of AI that was extremely impressive, but it was you it was token prediction at its uh uh in its most basic form, right? It was give it a set of input tokens and it will predict uh the output tokens, right?
SPEAKER_01So what was staggeringly impressive to the laymans like myself was completely infantile to someone like you who was like this is this is as expected at the lowest level. I think what or am I exaggerating?
SPEAKER_00No, because I think what was so surprising is I suppose we I what I was used to is the AI community not really having the amount of computing resources that open AI had. And that was like nobody got to see what you could do with a vast amount of computing resources and how would that change the game? And it it's difficult to extrapolate, and I think for every human mind, it's uh it's actually whether you've been in the field of AI and computing or mathematics your whole life or not, extrapolating what happens on the y-axis as you go out. We were used to supercomputers in the lab, and and that but this was tight, they were tiny compared to what open AI had used to an access to. So when you take and you go way out on the x-axis, then of computing power, and then go up, it's very hard to guess where that lands, and uh when it's on an exponential, and it is that like when I was using Chat GPT first, like I was predicting to my kids what I should be able to do, and then it's like more sophisticated things. And you go, Oh, wow, actually, this surprises me. This is I wasn't expecting it to um be able to help me with asking questions about say medical diagnosis and very technical subjects, like, wow, it's it's got access to the whole they really did get access to the whole web here. You know, they did things that uh and they got every textbook. So these were things that uh like copyright law was like out the window, and they'd ingested everything. Um so I guess there was assumptions or incorrect, right? That you you'd never be able to get that much data and that much computer power that has now changed the game. So now as soon as they did that and just like crossed red lines and so on, that if they if they'd done it slowly, somebody might have stopped them. But they did it suddenly, and when they released it, it was like it's a feta complete, and then there was all these other labs going. Um, and a lot of the transformation, transformers, um, literature on how to do this was kind of publicly known. So now it was a kind of engineering task to scrape vast data sets, push it through architectures that um so it was it was obvious then that oh, there's gonna be lots of models. Um, so it was gonna be all these models competing against each other, and there's investors throwing so much money at this, this is gonna move fast. But another major leap was, and I think Enthropic did a really smart thing here, they recognized that code is highly structured tokens, and in general, prose and uh literature there's less structure, but in software code, it's very highly structured, so the predictability uh is greater, and uh, when you see how some of the new models can generate code so fast that now it this is a change in gear because you can you can generate code that will generate new data and new insights, and now now you've got executable software code that is like another uh another level of speed and advancement because now we can move the whole software industry faster. Um, so it's uh like I might have been I might be in a world of AI 20 25 years, but I find it very hard to predict where it's gonna be in the next year or two because there's so much resources and things are moving so fast, but the software industry is gonna be changing quickly. Um I I think I I think it's it's it's foolish to almost predict what's gonna happen. Just it's in everyone's interest to be moving quickly with it. Don't don't try and swim against that, current. Uh it's because you'll be wrong.
Sponsor Break Rapido
Compute Power And Broken Rules
SPEAKER_01Yeah, exactly. Okay, let's take a quick time out and give a shout out to one of our sponsors, Rapido Solutions Group. Rapido connects logistics and supply chain organizations in North America with the best near shore talent to scale efficiently and deliver superior customer service. Rapido works with businesses from all sides of the logistics industry, which includes brokers, carriers, and logistics software companies. Rappido builds out teams with roles across customer and carrier sales and support, back office administration, and technology services. The team at Rampido knows logistics and people. It's what sets them apart. Rappido is driven by an inside knowledge of how to recruit, hire, and train within the industry and a passion to build better solutions for success. The team is led by CEO Danny Frisco and COO Roberto Icaza, two guys I've worked with from my earliest days in the industry at Coyote. I have a long history with them and I trust them. I've even been a customer of theirs in Molo, and let me tell you, they made our business better. In the current market where everyone's trying to do more with less and save money, solutions like Rappido are a great place to start. To learn more, check them out at goropido.com. That's gorapido.com. Now, let's get back to the show. It's so interesting. And and one of the things you said I'm curious about, you mentioned how it almost was like unfathomable to go that far on the X-axis with the compute power. Like you almost couldn't reasonably guess if you were to do that, what would happen? Now that you've seen that play out, I'm curious if you were to go back in time a few years and someone gave you like this unlimited compute power, would you would you expect that you would go down the same path that OpenAI went, where like it's it's a lot about search and and d ingesting all this data so that guys like me can just plop a question in and get. Every answer I need right in front of me? Or do you think there may have been other uses that are more interesting or applicable if you had that much computer power?
SPEAKER_00It was it was the problem to go after, for sure. You know, it it because the the founder of our AI research lab, that's uh Eugene Freud, uh this is the guy who came from uh from the US, and he said the vision for AI was declare the problem you want solved, and AI will figure out a way. Um so but there's a lot of AI researchers looking at different different aspects of intelligence. Like there's there's there's let's say there's the real world of robotics and the physical world, interacting. Like that's that's a subfield of AI. There's uh optimization, which is modeling combinatorial structures and finding finding the best combination of things to do. There's there's learning and adaptation, and then there's language. So um I don't think that most AI researchers um fully appreciated the power of understanding language, because I would say there was a maybe it maybe it's that researchers tend to be rules abiding, right? There was an assumption that you couldn't just r go into the library, rip the covers of all the books, toss them into a machine that ingests every piece of data, throw every copyright rule out the window, and just harvest the entire sum of human knowledge written on every book and everything on every internet webpage script, everything. I think university researchers would have had a natural like, but you'd be breaking laws. I said most people I've seen criticize modern AI have been some computer scientists and researchers who said, but you can't do that. That's against the rules. It's like, well, it's done. It's done now.
SPEAKER_01Yeah, the it's it sounds like what you're saying is like someone had to be willing to just pretend the rules didn't exist to make something this gargantuan and substantial and meaningful happen. Because if we follow the rules, it'd be 30 years before we were even talking about a path. That's why it didn't come out of Europe.
SPEAKER_00We're more of a rules-abiding culture here, so it needed a combination of like that they had fantastic AI researchers on their doorstep in Stanford, but there was a lot, I would say, uh lots of AI researchers around the world, but they probably needed to go to Silicon Valley to get the capital, so you need a lot of money and a lot of computer resources, uh, you know, and the appetite to go after something really big. And I would say an American culture of just let's get shit done and go for it, right? And and and it's done now. And I think everyone in the world can now look and go, yeah, okay, rules are breaking, but broken, but like there's so many benefits to this that the world isn't going back to pre-LLM days. It's it's now like let's just work with this and find a way. And maybe there's got to be some some type of compensation. I'm sure there'll be there will be some lawsuits, no doubt, going on, so but it's almost like this is too good and too valuable for humanity. Um, and that we've got to accept that when you combine this with LLMs pointed at software code bases and to generate new code and the harnesses around this, like claw code. There's going to be more clawed code. The irony of like developing a tool like Claude Code is that you can use a tool like Claude Code to build cloud code harness, right? So it's it's um it's funny how the the their only moat really is is speed and how fast how long can they keep running this fast? Because as soon as they slow down, somebody else is gonna go, hey, you know what? We've got a cloud code look like this, um, can do something just as good, and or it's got some kind of unique advantages. So I don't know what that'll be, but whatever it is, you gotta look at and go, oh, how can we how can we leverage this now and use it, use it more effectively?
Fear Jobs And The Real Upside
SPEAKER_01So I think it's reasonable. The the structure you talk about, where you know the the kind of American way of just put your head down and go and see what happens. And and it when you're willing to try new things, good things happen. This is true. There's another side to that though, and and and I've seen it. You know, one of the benefits of the business we built was we did that. We were like, let's just run as fast as possible, take off and see what happens. And there was so much good with that. We got to hire a ton of people, we grew really fast, but there were slippage, there was there were issues, there were things that did not run as effectively because you're moving that fast. Um, there's always that other side of the coin that's like when you go this quickly, it's just natural to expect that there are more issues on the backside that you're not managing perfectly because you're focused more on speed and growth than anything else. Now, add-in, we're talking about maybe the most powerful thing to ever be built. It I think it's reasonable to assume that the downside is is just as potentially powerful and fearful. And I'm curious, like, what is your general take on like our what our fear should be of the negative potential outcomes of all of this?
SPEAKER_00Yeah, it's there's definitely right, I I think with any major technological advances, there are downsides, right? But if you look back in the history of mankind, when there's been major advances, like I say the Gutenberg press, there were monks and monasteries going, oh lads, we're gonna lose our jobs. Who's gonna write all this stuff down? Right? There were there was people who lost, but you know, almost like those stories are lost in history because everyone moved on. There was a society became more productive, there's more things we can do. I we look back on the age pre, I would say, Chat GPT and say software was very expensive to build in that era. We didn't have enough of it because it was expensive to build, it became cheaper to build, so we're gonna have more software embedded in more of what we do. Um, and if you look in the freight industry, freight industry is definitely one that can benefit more from intelligent systems. And so we're gonna see like what I'd love to see, and uh like there's a downside in that I'd say the downside in freight will be that some will get cut out if they're not good at what they do, and they benefited from the fact that shippers were not good at interrogating who who should they be dealing with, who's who's good at their job here. If you're not good at your job, it you'll get discovered now. So that's I would say that's a good thing overall. From a societal perspective, the the small players who are good at their jobs will find it easier to get into markets and start a business with a small number of trucks and start serving the big guys because now the big guys don't rule you out because you're below a certain scale. It's just that if you do these two lanes really well, that's all we need. Just do these two lanes really well and keep doing them well, and you'll keep winning. And that to me, that's economic efficiency, and that that can be enabled by intelligent systems, because you look at a Walmart or a big shipper, like the only way they could deal with mom and pop shops for a small handful of lanes is that if the intelligent systems are running the decision making around who to give business to. And if the rules are are or if it's well designed and integrated with operational performance systems in the back end, they can keep winning. All you have to do is keep doing your job well, and you you shouldn't have to be updating your bids to uh all the time. You should be just saying, This is what it costs me and this is how well I can do, and I'll keep doing it, and that's it. That's it. It's beautiful then. Like you don't need too many middlemen if if someone's uh good at good at good at doing their jobs and the systems are set up to facilitate that. Um will there be job losses? Yeah, almost, almost certainly, but I think there'll be more jobs created than jobs lost. That tends to be with any major productivity enhancements, but we just don't know what those new jobs look like exactly.
SPEAKER_01So that's the point right there that I want to dig in for just one more minute on, because what you said, I I understand. And there's another example I remember it was like 500 years of history shows that any new advancement always results in more opportunity and more jobs than those lost. However, it is way easier for humans to understand and digest seeing what you'll lose that already exists versus imagining what will be created that doesn't yet exist. Now, I understand that why it's so hard to like envision what doesn't yet exist. I think the the example I remember reading about was maybe the sewing machine, and when that was created, how like you know, some queen in Europe was like, get rid of that thing. Like, we'll lose too many jobs. Um, but so much was you know created as a result of that. My my challenge or what I think about is like we're creating this technology that is theoretically going to be better at than humans at many of the things that we're gonna ask it to do. And it's so far and wide, and the scale is so grand that I'm like, is there really enough opportunity to create all these new jobs to counter that which will be lost? And again, maybe it's just true that because it doesn't exist yet, my brain is struggling to see it, but part of me worries that the the amount is so staggering that there really isn't a replacement rate that can support that.
SPEAKER_00I don't know if it's I think I think the the best corollary tends to uh is the industrial revolution or the agricultural revolution, right, where so many jobs were on farms. Like I think if you go back to America 150 years ago, something like 90% of jobs were on farms, or there's some but when a tractor came in, the number of jobs uh just decreased so fast, right? So I would I would say that's probably that's probably one to look at and say, well, what happened in those cases? And let's say when the cars came in, blacksmiths disappeared, and care for her oh the whole equine industry, a huge employer. So I would say I suppose economic historians will probably be able to tell us how do those people adapt. And I think there is a always there's a natural concern that this is such a revolution, it's happening so fast across so many industries. Will we be able to adapt quickly enough or uh to avoid major social upheaval? I I can't say for sure. I don't I don't know. I think that we have to accept that maybe there will be trouble from this. But I what I suspect is that we're underestimating some of the diff like AI looks like it's really good at developing new software, but where it does it in smaller scale, like it it it mini blocks, but putting those blocks together when you don't know what's inside the block, that's kind of challenging. And that there's an there's I would say greater skill around architecture needed now to put it's almost it's almost like building sites at all these small breeze blocks, and somebody's actually coming along with a giant block and placing you know just one big block to make a wall. But now it's like, okay, how are we gonna build this very large monument with fewer blocks that are not necessarily as flexible uh or as malleable as the older, smaller blocks were? That's got it's gonna have new challenges. Um and I suspect that there's new roles around software architecture, for example, uh that in putting intelligent systems together and how to get creative, and it will be a new premium on systems thinkers, those who can look at a multiplicity of different systems. You can build the individual systems quickly now, but piecing them together in a way that solves a problem better, and envisioning a multi-step process of building a larger system with more of these blocks, it's gonna it's gonna give us more options. We're gonna see more solutions on the market, um but hard to predict. We're gonna see the 2026 is just gonna be a fascinating year, right? That's that's that's for sure.
Rebuilding Software Work With Claude
SPEAKER_01So you're obviously in an interesting position as as one of the original AI researchers, but it sounds like even for you that there's been this kind of epiphany moment in the last 12 months where it was like, okay, things are changing now and they're gonna change very rapidly. I'm curious from as as a as a business leader, how do you navigate all like is it was there a moment where it was like, hey, we need to stop what we're doing and and think about our overarching strategy and how we're going to change how we run our business moving forward?
SPEAKER_00Yeah, it was uh and I was saying that was the first week of January. When I started playing playing around with Cloud Code, and for me that was like, oh wow, they've cracked, they've they've cracked a big knot here. Um and that even just playing around with that uh during the Christmas holidays, I could see that this means for our industry and the software industry, this means we need to be building software in a different way. This is this is uh this will have knock-on effects in many industries, but first and foremost, I think the software industry is getting disrupted more than any other industry. So we're us ourselves in the software industry are the examples others of this is how it might play out. And the software industry is very I would say it's a new industry, so it's quite adaptable. Uh there's a few of the older, bigger guys less adaptable, suspect they're in they're in trouble. Um, you know, uh, but how we organize ourselves around new ways of working will actually be instructive to bricks and mortar businesses um who will be able to lean on similar technologies to their advantage, too. So um I think software industry is going to be the case study for other industries and how to adapt in this new age of intelligent systems.
Leading Change Inside A SaaS Team
SPEAKER_01Holy sh that is what everyone says when they see Genlog's truck intelligence platform for the first time, founded by XCAA operatives and fueled by 15 million daily images across a nationwide camera network. Genlogs gives you the power of total market capacity while also defending you against fraudulent carriers. Holy shit is what Genlog's customers say again when they see the ROI from covering loads faster with fatter margins. Holy shippers. That's right. Genlogs unveils the locations and lanes for all the shippers in America. In the era of artificial intelligence, nothing beats actual intelligence from verified by video data. See what Genlogs can do for you. Check out a demo at genlogs.io. Again, that is g-n-l-o-g-s.io. Tell them Andrew sent you, and they'll include their carrier compliance suite for free. And if you haven't already, I interviewed their founder, Ryan Joyce, last year, and it's one of my personal favorite episodes that we've done. Check it out. Now, let's get back to the show. And and how hard is that? You know, you your business has been around now for 13 years, and you know, every month it's like you've got your team and everyone's got their job and their responsibilities, and they show up to work every day. They know exactly what they're supposed to be doing. Then one day Alan walks in in January and opens Claude Code and says, Holy cow, our business needs to change tomorrow. Like, how do you actually apply the kind of a masked, like what was the actual process for saying we need to make changes to to uh support this?
SPEAKER_00Validate that it was as big an event as we suspect it, right? So um so first thing was let's make sure there's because we've seen hype cycles in tech before. I've been through blockchain and trying to tell customers we ain't blockchain, we're not Web3, no NFTs, no, and there's there's there's been many of these where there was hype cycles and just ridiculous. But this I knew this one was this is this is different. This has been said before, right? But this one is different. Uh but let's validate that. So we we started to look at how this can be used in anger on tough technical challenges, and validating that this real this is the real deal, you know. Um so we did that, and uh we're lucky to have some really strong, I would say, technical leaders who are able to. And I think it's natural that some people will be skeptical about just how good it could be, because it sounds too good to be true. Um but then you test it out and you go, okay, this is you test it on challenges you know are hard, and when it's able to solve those and do it a really good job, you know, okay, all right, this this does change everything. Then you've got to you've got to make sure that you're trying you don't want to leave people behind at this, right? Because there's there's gonna be some there's gonna be in every organization there's gonna be um people who want to jump on the latest thing, and sometimes you gotta talk them down and say, uh hold on. Um first of all, validate this is for real, then encourage them. Okay, let's have more examples. Uh let's encourage everyone. It's what you'd like to do is have everyone buy into it of their own volition. People it's human nature that if you apply if you if you try to apply the stick instead of the carrot for adoption of new ways of working, that there's a danger that you've you've not brought everyone along and you won't make the best of what this could be. So um we've been working hard on making sure that I think we've got everyone bought in. But it and within the space of six to eight weeks, it's a lot to get done. You know, you're you're kind of saying that you you've been building code a certain way for 10, 20 years. There's a new way now. And it's it's not like you won't you won't ever code again because I was when I was talking about those big blocks, they're like piecing them together, architecting, judging how this should be built, it really benefits from people who are experienced at coding. Like you put in someone who doesn't know how to code and start start building things. Like I'm very I I've been using cloud code, I'm very rusty at coding, right? So it's like 12 years ago I stopped coding. I don't produce good artifacts the way good coders produce good artifacts. You can it's kind of self-evident when you're looking at it. Um so uh I don't buy into the belief that someone who doesn't know how doesn't have a clue how to code will do it a very good job with this because they just won't be able to see where the flaws are.
SPEAKER_01Um so the floor is raised for everyone, but in a world where excellence is what wins you business, like that's not solving your problem. You still need the best of the best in all these areas to create a great product.
SPEAKER_00Yeah, you you do. I I think I think what I'm expecting to see is that software companies start producing more products, and it's the biggest the biggest vendors who have a suite of modules that are interconnected, and they kind of built large software companies through acquisition. They're the ones that will be most challenged because they're very often they bought they bought comp different companies, piece them together. The user experience is very different different in each of the modules. And now say companies like us, where we were specialists, we can we can build things that are adjacent to what we do, but they could have the same user experience. The same design skills. So you can kind of have Claude build your software in a way that's consistent with how you um built core pieces. And now you're going to see more software companies with more products. And so you're then you're you're judged by how good those are, right? And and what was how good was your judgment in the creating these multiple modules in this in these new suites?
SPEAKER_01So I think that's a good path for me to then ask, like when you think about Keelvar today versus where you see the business in say three to five years, what is the maturity, what is the development, what what what changes over the next few years in terms of how your business looks and engages in the market?
SPEAKER_00Yeah, I I think we'd be doing more much more for customers, you know, in that we were specialized in sourcing, but many of our customers have been asking us for years, could you do these other things? And uh being disciplined and staying in your lane was the thing to do. The best thing to do when building good quality software was so expensive. You had to be very judicious and it took more time. But now you've got to stay within the lane of your knowledge base, not the lane of your or where disciplined code creation restricts you. That it's imperative that you before you tackle any adjacent problems, if you if you have a very good understanding of what pain point needs to be solved and how it should be solved, then you should solve it. Um and uh so that's I would say it brings your three to five year plan into year one. It's it's almost like every software company's three to five years. If you didn't have a three to five year plan, you're in trouble. And I I think there's lots of software companies that didn't have a three to five-year plan. I honestly I I do talk to other software leaders, and they're not necessarily think thinking three to five years out, but they this is a moment where if you haven't been thinking three to five years out, then you'll you'd be trying to paint at the wall. You know, um, and we'll be telling you for for what's a coherent plan look like.
SPEAKER_01Yeah. Is there a product or service that you can say that you think is like super adjacent to sourcing that like is a natural next step? Or it's you still want to keep it on the promo?
SPEAKER_00I think that it's at our our connect we're gonna have a kill our connect event in Dublin in June. And we're gonna announce a range of new offerings there there. But what we're planning to do is is build a JSON application, but see what do our customers like or dislike. And if they don't like something, we just ditch it and we just say, okay. Um so but if they really like it, then that will be the arbiter for whether we go, okay, push this because go to market for software is still expensive. You know, you don't want to be um you don't want to be trying to sell or or having products where your go-to-market team doesn't understand it fully and they can't sell it and doesn't understand the market you sell into well. So even if you so if if code production costs have come down, the sales and marketing costs around it haven't necessarily. So be judicious about how you apply those kind of relatively expensive resources to.
The Next Keelvar Products
SPEAKER_01Yeah. Staying on that kind of future thinking line of thought, what what's something that you think, say five years from now, just specifically in the freight procurement world, what do you think looks very different?
Five Year Future Of Procurement
SPEAKER_00Uh it's it's gonna be much more like e-commerce in that it's machine-to-machine negotiations, but it's control you're controlling agents on both sides. So the shippers are gonna have agent managers who are seeing how how their operational operations are running and kind of swarming on why there's failures and what what you do and how you instruct your agents to avoid future failures and this notion of anti-fragility, right? So ship the shippers very often they'd like a self-healing system. So they want agents that are always learning from mistakes and making fewer and fewer and fewer mistakes, and getting better and better at predicting potential black swan events or having robust networks so that when a black swan event occurs, that you can heal your network fast. Um anti-fragility on buy site is kind of a key long-term objective for you know because I would say it's obvious that they're all moving in the in the direction of agents, but what's less obvious is that what should be the um the overarching north star for those agents? Uh what do they learn against? And that's uh anti-fragility, right? So uh then on the sell side, the amongst the carriers, then you you just want to be identified having the identifying your custom your customers of choice and prioritizing that you should have agents that are optimizing the utilization of your fleet, and the you have a finite capacity available. Make sure you're dealing with the best customers, or the custom customers you're serving their the needs that suit your business and avoiding those needs they have that don't suit your business, um, and optimizing your network design with respect to the best customers. Um, there will always be customers who don't adequately reward suppliers who do a good job and bias towards price, avoid them, right? Um, and that's the best way for them to learn, is that they lose their best suppliers. Um so that's that's I would say, and that's the complementarity between agents and both sides of a market. That it can be win-win when there's reward schemes set up so that you incentivize the right type of behaviors.
Multi Shipper Exchanges And The Holy Grail
SPEAKER_01So when I think about creating an efficient market or just an efficient network, it's important to have the most data possible from every side to plug in to create this network. And and I'm curious because this is something that's never been done, and and I've asked a number of people about the concept, but you might be the perfect person to ask about it given your business. I feel like any one of your large enterprise customers could become more efficient if there was a path to not just leverage their lane data, but also the lane data of competitors, or competitors may not be the right word, but other shippers and manufacturers who also have their own set of lane data and their own networks. And I'm curious if you see a path forward where the network of one shipper combines with the network of others to create this larger network effects, and if that's a role that your team plays.
SPEAKER_00Well, you you would need to change the clearing rules around the auction to and this this is to move it towards what's called a form of combinatorial exchange, so that you clear it, you do synchronized clearing across multiple buy site.
SPEAKER_01Um so and by synchronized clearing, you mean you're confirming that the price on A to B is approved by shipper A, you're approving you're confirming that the delivery time for B matches up with the pickup time for the next location that's owned by like all of these rules need to be met, right? That's what you're saying.
SPEAKER_00Yeah, exactly. That the commercial terms and the operational terms are all met. So you've a satisfiable uh loop um where utilization for that truck is optimized, and that there's a discount up. So that's the kind of alignment of incentives across all those actors. So you have shippers A, B, and C getting a discount because carrier A is saying if I get that loop, I'll give you a discount. And it and it's the agents working that on, right? So no human in the loop, right? So it's just requests coming in on buy site and agents from the I suppose the trucking companies who are offering these type of conditional bids. That's how you use uh optimization to the max. Uh but there's a I would say there's a quite a few things to get done as we head towards that type of that level of efficiency. Um so it's not easy because one of the things is like bice is going to get used to the fact that they can clear their spot bids in seconds. Uh and if you're saying, well, no, wait, wait till clear on the hour, because we could have thousands of spot bids that we synchronize the clearing of and do this optimization across if you were to do that, uh it's almost like the kid, you know, and the same research as Thomas Sandom, who um is a researcher in a professor in Carnegie Mellon University, who did quite a bit of work on this, on kidney exchanges, but also on combinatorial exchanges that could be applied to um uh that type of I would say multi multi-shipper, multi optimization, multi-carrier optimization, synchronized uh combinatorial exchanges. It's kind of like a a holy grail. Um hard one to get to, uh but possible. Um but you can working towards it.
SPEAKER_01It it does feel like no matter how efficient you can get in one shipper's network, it will always be less efficient than if you could find a way to add in another shipper to that network. And and there's but the the path to get that closure to make that executable is seems so gargantuan, and maybe that's why AI is like a perfect solution one day.
SPEAKER_00Um but I was curious in that way in that direction, right? Because we're getting more and more shippers optimizing within their their own myopic level of optimization, uh, but then to do the joint optimization, it I would say there's no technical reasons why it can't be done. I would say there's more business reasons why it's hard to do that, hard to synchronize some of these challenges. Uh so it it needs volume too. It needs it needs serious volume to work. It's good.
Career Advice And Starting Again
SPEAKER_01Well, let me leave you with this an easier question to end on. Um just some advice. Um actually, I'm gonna ask you two questions because one is for the general public, you know. For the for the 22-year-old kid graduating college who wants to get into supply chain and and the logistics industry, what advice do you have for them, you know, from your own experiences, from your what you're seeing in the market, on what they can do to best set themselves up to succeed in this world?
SPEAKER_00I'd say be curious about the new tools and teach yourself how to use these new tools. There's always there's always slow learners out there or people who are reluctant to adopt new tools that change the game in terms of efficiency. So when there's a big change like this, so many people are being negative saying, Oh, what will this do for graduate? I say, Well, graduates are open-minded. Every graduate I've ever met has been open-minded about what tool to learn. If people who've been stuck using certain ways of working for five, ten, twenty years tend to be the people that will go, no, that'll never work. Because they've a bias towards feeling what what they've learned thus far is something they've got to double down on and keep using as an advantage over new entrants. And to a graduate, use that to your advantage, right? That you can get things done faster if you use these new tools. Um I'm all I'm always optimistic about the future for graduates because I think that open mindedness is the key to success.
SPEAKER_01I appreciate that. Last question, and this is more advice just for me or other people who might be in my boat. Uh, I've got, as of today, 229 days left on my non-compete. Um, brokerage is the only thing I've ever done. And there's a big part of me that wants to go start another brokerage in 230 days. Do you think that's a bad idea? Do you think it's too late to start from scratch in this brokerage world that you know one of the things you said earlier was like it's the people in the middle who might have the most to worry about. That would be me.
SPEAKER_00Um I'm curious your thoughts on that. I just think there's always opportunities here. There's I think there's just so many so many people will be too slow to adopt new ways of working. That if you find an angle with this new type of tool, you can build your own tools as well, of course. Um that you look you can find an edge, right? Well, it there's a disruption happening. So I'd say old relationships will build up, will start to break down. There will be big shippers and big trucking companies who just get stuck in old ways of doing things, and it will be a dislocation and there's going to be opportunities. So it's actually a great time to be starting a business and challenging older norms. I don't know what the angle, but you should you obviously need to find an edge or an angle and do things differently. Um I'd say go for it. If you if you've got an itch, like I had an itch, just scratch it. Sometimes you gotta scratch it in 10 years later you're like, oh, I'm still scratching this itch. It's yeah.
SPEAKER_01I think anyone who knows the exact number of days on an entropy, they they have an itch that still needs to be scratched. Um, for those that still count. So you know, I I appreciate that insight though. It's it's funny because I when I think about it, when we started my business in 2017, it the path forward was so clear. Like I understood what we needed to do. I understood what I thought our edge was. It was so clear. This time around, even if we operate the same way, it just the path just feels foggier. And and I think that's true for so many people. When they look at the future, it's so hard to see clearly what that looks like because we've never done this before. Any of us. Uh, even those who have studied it for 24 years uh admit that they're learning new things. So it's it's a fascinating future, and I appreciate your mindset. That's like if you're willing to see opportunity, the opportunity will be there. I think that's true of perspective. Like it's the half-class full versus half-class empty mindset. How you view the world is how the world is.
SPEAKER_00That's it. Yeah, exactly. And uh sees the up, there's opportunities there, that's for sure. That's just it can be the as you say, it can be hard, harder to find them because it's a more complex world. You know, the world was simpler 15-20 years ago. The opportunity was like you could just describe a nice nice and little neat business plan. Now it's it's all about adaptability, I think that just learning quickly, iterating fast. Being open to saying my first premise was wrong, but I learned something new. Changing, changing direction.
SPEAKER_01Yeah, I mean, uh, it's funny because you see that now, and it's like it reminds me of one of my co-founders, Will Jenkins, used to say this all the time when we were starting our business the first two years. The word nimble was the most commonly used word by all of us. It was just like, hey, we got it was like something changed, something's happening, we have to change. It's like we don't have a choice, be nimble, like make the change and move forward. And I think if that's your mindset and business, you have such a leg up on the competition, and that's why these big companies struggle is because making change, being adaptable is so hard because there's so much red tape, it's so hard to get things approved versus a startup. It's like, hey, I see something, go do it. Exactly.
SPEAKER_00Their inertia is everyone else's opportunity, you know. That's to slightly bastardize Jeff Bezos quote. It's like inertia is like the biggest killer of big businesses. So um, yeah. Um I and I can see them being a little fearful of AI and not really knowing how to leverage it effectively. That's I think most businesses have a challenge there. Yeah.
SPEAKER_01Thank you. This this has been uh really insightful. I feel like I've learned more in this 80-minute conversation than I have in a lot of these. Um no offense to the past ones. This one was just very insightful. So I just want to say thank you, and I think my audience will appreciate as well.
SPEAKER_00Um, hopefully you can enjoy the discussion and um yeah, appreciate it.
SPEAKER_01Thank you to my audience.
