Flexport founder & CEO Ryan Petersen on Al, tariffs, India & the state of global trade

Flexport emerged at a time when global freight was still heavily dependent on fragmented paper-based systems. I can't stand paperwork. That's why I started Flexport, by the way. I think data and Excel sheets are turning out to be fine. LLMs are really good at processing these things. Being able to pair a customs expert with an engineering team is the secret sauce. Reduce the size of our team. Reorganized to be more customer-focused.

The plan for India, I mean we're frankly late to the game in India. What, 3,000 years late? We'd like to take the company public. Yeah, we should be public in the next 5 years. I would say long-term the future of Flexport is much more nimble.

Yeah, it's great. Thanks for that nice intro.

Yeah, it is. It is.

Well, there's so much that goes into it. I guess the number one driver is, as you mentioned, tariffs. What happens with the US administration and their negotiations around the world, that's by far the biggest driver. We've seen it drive a huge amount of volatility because in fact even when they impose a new tariff, it leads to temporary surges as people try to get freight in before it takes place. And so we've had a lot of bullwhip effects. So we're going to see, we've got not just the administration but the Supreme Court is going to rule soon about whether those tariffs are legal and then what happens next. So I expect a lot more volatility there. The next big one is probably the Suez Canal. I mean, the fact that that's not the biggest story in logistics is sort of crazy when you think about it. The Suez Canal has been closed now for, mostly closed for container traffic for what are we going on? This will be the third year at the end of the year. And then third is probably overcapacity that we might see. The ocean carriers have ordered a lot of ships and we're going to figure out if there's too many ships here by the end. We'll see if they reopen the Suez and there's not a big surge in demand. If the global economy sort of plods along, then I think you're going to see a good environment on that side of things, low prices of freight. And then the backdrop for all of that is AI and what AI can do for supply chains in making better decisions, lowering costs, improving responsiveness. I think we're on the cusp there. And not to mention on the demand side the buildout of AI, like the data centers that are getting built, the amount of logistics that's needed to do this sort of generational capex. So lots of interesting stories. As a transport reporter, you picked a good domain to have plenty of things to write about.

Yeah, I like the electrical grid analogy or the cloud computing, but electrical grid is like, we really take it for granted that when you flip a light switch, you're controlling a power plant. It turns on and off. You're actually physically, that power plant is generating a little more electricity for you. It's burning a little bit hotter on demand. So it's sort of crazy to think that that's how responsive it is to customer demand. When you order something online, I think a lot of us just don't think about it. So mentally you're like, 'Oh yeah, it's probably what happens. I clicked it, all these things took place and one more unit was produced somewhere in the world or like some statistical model said how many more units to produce in response to that demand.' But in reality, it's not. It's a lot of Excel files, a lot of PDF attachments and phone calls. Really, it's not a well-orchestrated system even in the biggest, best companies. I mean, if someone tells me that their supply chain is really well-run, I know it's just because they don't trust me to tell me the truth because I have not seen one yet that's really well-run. We work hard, people are doing their best and people can make improvements and there are well-run on a relative basis certainly, but compared to where it could be, there's infinite problems to go solve. But from a vision standpoint and where we're executing technology-wise is really starting with the unit of inventory, with the customer demand and saying, okay, how do I forecast that demand better on behalf of a merchant so a merchant can say how many items am I going to sell of each of my products? What's my demand look like? And therefore, what should my replenishment plan to bring more goods in based on real data for how long does it take to produce the goods? How much do they cost? What's the margin? How long does it take for transportation? What's the cost of capital? What's the conversion rate? If I can ship, if I can deliver the product in two days, one day, here you have Zomato in five minutes, or maybe it doesn't matter at all and I sell the same number even if it's two weeks shipping on the website. All of those need to feed into this model that tells you how much inventory to have and what's your risk tolerance because if you end up with too much inventory, you're going to have to discount it and your margins are going to get crushed and your brand's going to suffer because you're seen as the discounter. If you don't have enough inventory, you're losing sales, you're not making money that you should make. And that's a very hard, these are hard sets of problems that today a lot of the models for this exist from the 1970s, like the mathematical models. It's not brand new math. Yes, there's going to be some breakthroughs in AI and machine learning and ways to better calculate this, but at its core, you barely need that. What you need is real-time data from all of those data sources I just mentioned, the product dimensions is one other, the price of air freight and ocean freight. How does that all feed into a model that will then tell you based on our prediction of demand how many units to buy, when to buy them, how to ship them, where to ship them, where to position them. That's what I mean when I say the electrical, you know, you want to make it really responsive to customer demand. Then you buy something and the models are figuring out what should I do about this order. Of course, the first thing to do is ship the product. I'm talking about the replenishment cycle.

Yeah. I'm glad you found that statement of mine because that is how I feel about competition. If we can learn from them, we need to be obsessed. When they know something we don't know, we should be absolutely obsessed with learning it. And they do have a lot of capabilities that we don't have yet. When I look at the competition in the freight forwarding world, which is, we sort of can think about supply chain in terms of three flows that are happening in parallel. And the goods flow is the obvious one that we look and we see and we cover. You know, you guys are covering and the movement of containers. And so that's the first flow, the freight flow or the goods flow. But you also have a data flow which is more important than ever because of customs compliance and tariffs. And this is all the data in and out to coordinate the freight movements, but also the government needs to know things, the models that I'm referring to, inventory financing model. And then the third flow is the financing. So you have the cash and these are all happening in parallel and they're all running in sort of a loop, like the goods going this way, the cash going this way, the data going both ways connecting it all together. In each of these areas there's probably some competitors who have some capabilities that we don't if we're honest with ourselves. In the goods flow, it's really about being local everywhere, which is very impressive. Like the biggest freight forwarders in the world have thousands of offices with local expertise in all that, you know, not just the big cities of the world but the second and third tier cities. Even a medium-sized freight forwarder in India will have 15 different offices. I mean India is so big, right? I think there's probably 15 cities with more than five million people in India somewhere around there. And if you go to a million people, probably more than 100 cities, right? So we are newer to the market, founded 2013. So it's going to take us a while to go build out, recruit the talent, build the trust to put people locally in all these countries. Flexport's in 42 offices in 25 countries now, but you know, competition is in 2,000 offices in 200 countries. So that's a big advantage if you're shipping. Having local expertise can really help. We don't believe that tech solves all the problems. You got to have elbow grease people to go do local compliance, sales, relationships with both the carriers, the vendors that own the assets that move the freight, the customer relationships. So in the freight flow, that's where I would say is the breadth of talent that they have in the number of countries that they can cover there. Physical network is really kind of overstated. It's really about the human capital because most freight forwarders don't own a lot of assets or they try to own no assets. They subcontract to, you know, they're not owning the trucks and the warehouses. Even when you see their logo, it's usually a subcontracted, you know, put the logo on the side of the truck of someone else's truck. So it's more about the breadth of coverage, number of offices. I think they have a big competitive advantage and it's a fun part of my job is to go try to close that gap. We have a plan where we'll have by the end of 2028 an office in at least one office in all of the countries that represent 90% of global containerized trade. So you get pretty close. But again, one office is not going to be enough in a place like India. Need to be in all the second tier cities as well. So that'd be the first one. In the data flows, probably there's, I don't really see someone who's, certainly not the traditional freight forwarders are not our competition there. They don't really play. They do a little bit, but they're not like at our level in terms of the amount of investment made, the amount of emphasis made on this inventory models and on the algorithms for everything that we're doing on that chain I was describing. But there are other companies that do interesting things there. There's a lot of companies we need to learn from in terms of AI, whether they're in logistics or not. The AI turns out to be pretty smart at making some of these decisions and being active here. And then financing, I mean we don't do, for example, we don't do payments, paying for your inventory. That's a really interesting area for us to learn about, be able to build capabilities and see how other companies are managing payment flows. So we don't really, like yes we definitely compete with freight forwarders but on all three of those flows actually we compete with other either tech companies or trade financing businesses etc. So we see all of those as competition on some level that we can learn from.

Did I steal that from Fred Smith? I didn't even know that. That's great. I'm honored. What a legend. RIP.

Both. I mean, I think data and Excel sheets is turning out to be fine. LLMs are really good at processing these things, especially if you fine-tune them and you build, we use LLMs to build technology that can extract data from Excel sheets. It turns out we spent decades trying to structure data and maybe actually unstructured data is fine and Excel is more structured than most data that you're going to see in this industry. So, we're seeing more and more that it does, like one of the value props for Flexport working with a company is we'll take you, we'll meet you where you are data-wise and we're working with some of the biggest and best companies in the world, really famous brands, and I won't say their names because in many cases their data is not good and I don't want to embarrass them, but it's fine. We bring them on board, take the data, you can email it to us, you can give it to us in Excel or PDF or any other format and our tech is capable of ingesting that, cleaning it and giving it back to your ERP or whatever system that you need it in in the right format and helping to manage that flow. And likely you're going to see more and more companies also do what we do themselves internally, like get a better handle on their data. And you sort of need that if you want to be good at using AI. You need to have data in a place that's with good governance and trustworthiness so that if you're going to give it to the AI to reason about, to generate reports on, you want to have it in a place that you can trust.

Yes, so we built this tariff simulator, we launched it in June, we keep iterating, we're constantly building it so we'll have to make updates every few days. In fact, there were 52 changes to the US tariff code last year, more than one per week on average. And so keeping it up to date, helping companies navigate this has been a huge differentiator for us. Some of these are very substantial changes. They're not just like, oh, we changed the duty amount. In many cases, they changed the way that you calculate duties. And companies are really struggling to keep on top of this. So, from a strategy standpoint, it's simple: be better than the other customs brokers of the world by being better at calculating this and more helpful to the customers. It really works. So, a couple stats. We had more than a thousand users of the tariff simulator. More than a thousand companies signed up. We had actually 700 users just at our direct competitors. So, show you that competition, big freight forwarders, big customs brokers haven't been able to keep up and they're using our tech to do that. We gave it out for free to the world, I think. I don't know why we did that to be honest. About a month and a half ago, we cut it off so you have to be an importer or an exporter to use it. It's still free, but you got to be not a competitor of ours. And the online reaction to that has been hilarious. If you go on Reddit or LinkedIn, you see people, 'I can't believe they're cutting me off. This, you know, you can't do this to me, etc. This is my day job, so or this is where I do my job all day.' So, it's showing you there is a competitive differentiation between us and other customs brokers who don't make their own tech because making our own tech meant we could be very adaptive. When the rules changed, we changed, we update our tech and they haven't been able to keep up.

Yeah, the good example by the way, the one to look at is fun also is a canned beer. Because a beer can helps you understand the changes to the complexity of the US tariff code. So, a beer can, what's the value of the can? Like, say I'm importing that into the United States. The value of the beer can is mostly the beer. It's not the can. I'm just guessing like let's say it's 50 cents and the can is probably 1 cent or two cents in value. Well, under the US new tariffs, you owe 50% duty on steel and aluminum products. Well, this has an aluminum can. So you owe 50%. But you only owe 50% on the 1 cent, not on the 50 cents of the beer can. So this is a massive difference and that is a new way of calculating duties. It was never done that way. It was always the duty as a function of the HS code. So now it's the HS code plus this layer for aluminum. And if you as a company and your customs broker cannot break out the 50 cents, one cent is the can and I only pay 50% duty on the one cent, you're going to have to pay it on the whole amount and it's going to cost you, you know, 25 times more tariff if you get it wrong, something like on that order of magnitude. And so it's a simple problem honestly. Like we updated our tech in less than a week to manage this, but competition is still, what a lot of companies are doing right now is filing and paying the full amount 50% duty on the full can and then the next day they file a correction and get a refund and the refund can take months. So from a cash flow perspective, first of all, the overhead of like getting the compliance team involved to file, it's really bad. And so we've taken a huge number of large enterprise accounts from the competitors who had a 20 plus year relationship with their customs broker because they couldn't process this and it was costing them so much money. So it's just a good example of like tech and expertise. And if you were just a tech company, I don't think you'd have the expertise required to do the customs, to figure out the nuance of the rules and like be able to pair a customs expert with an engineering team is the secret sauce.

Yeah, I mean that's a big part of what the tariff simulator is about, but tech alone doesn't do it. So we have teams that are constantly engaging with customers to teach them. We also run a webinar series like every week or two there's a new webinar that we're putting out, probably once a month about customs specifically. Those are getting 1,500 plus people to attend including competitors. We haven't yet banned them from the webinars. They probably should. So yeah, there's a huge desire for education in this industry and I think news media plays a really important role in that part of keeping people informed to what's happening. But often people in the industry have more exposure to what's going on in real time and stuff. So we've tried to do from the beginning basically marketing as education, like not self-promotion but education. I think self-promotion is kind of dead in this industry, in this world. It's too easy in the internet to say I'm the best. It's free. Everyone can say I'm the best. Whereas giving you helpful information, then you will think oh this guy's the best he gave me helpful information whereas if I say I'm the best you'll be like ah he's full of himself, not valuable. So yeah, a huge amount of demand. One interesting thing though is the degree to which customers have, especially founders, like which tends to be smaller companies, where they have become experts. They've had to become experts in their own product tariffs, rules, and compliance regulations to a degree that I never would have seen before. Three years ago, I could talk to any company, I'll probably know more about the customs regulations than them. And now I meet a company and like they're teaching me about their particular product and what, and the CEO knows this stuff because it's such a big dollar amount that they're telling me. I have to be really sharp because a lot of times the company, they should know their product better than anybody, but for me to be, it's kind of hard for me to keep up.

It's, I guess there's only three words for that is growth, growth, and growth. Now growth doesn't come from wanting growth, right? Just like money doesn't come from wanting money. You have to create value for other people. So a lot of it is driven by better user experience, AI to make these decisions and help people run a leaner supply chain, lower their cost, global expansion and stuff. So all those things will lead to growth. We don't need to grow that much more to be profitable. We're targeting, I think our total target for topline revenue is about 30% for 2026 over last year. That'll make us very nicely profitable. We'll see. It's very hard with a lot of our revenue is freight revenue and so you know it depends a little bit on what the price of freight does. Higher price of freight does lead to more profits for us. So that's a little bit we always put it unknown. We're projecting very low price of freight next year though. Growth is the main thing. Cost discipline secondary but always important if you want to run a business looking at everything.

We are at least in our financial plan forecasting that it comes down.