Flexport’s Ryan Petersen: tech’s trade expert on solving tariff chaos with AI agents & winning again

18:13So we've always had plenty of proof points that we know and what customers say about us. I would say last year, the way that we handled, there was just this stark difference in how Flexport responded to the tariffs versus the competition in building technology to help you figure out exactly what amount was owed on each product. There were 53 changes to the tariff code last year in 52 weeks, so more than one a week. And the rest of the customs brokerage industry basically failed to keep up. And in mass, we've been taking especially large industrials and automotive companies because they have steel and aluminum.

18:51And the government put these new regulations out for steel and aluminum for how tariffs are calculated. It's not that hard to do, but it requires database technology that they didn't have, for lack of a better term. And so we just started rolling up even like German car companies taken from German freight forwarders. Like this American California company has no business doing that other than the tax.

19:15Ryan Petersen:Well, so let's talk about the tension there. Is something like tariffs and chaos in the market there, is that good for your business? When all of that was announced last year, what's going through your mind? From a market share perspective, we doubled our volumes in our customs business and in our e-com fulfillment businesses last year, 2x growth, and driven by that. So we take market share like crazy, just actually just absolute growth in this sector. I never want to say it's good because it's really hurting our customers. Like you definitely don't want to celebrate that. But certainly we're relatively much better than other providers for these companies right now.

19:51So, yeah, it's been a net positive for us. But like I'd rather the market grow. Like the market has probably shrunk for total number of entries and total amount of customs or total amount of freight being moved. So it's better to be part of a growing market. One of the framings I've always used is, so global trade has grown 4 % annually since the Mongol invasions around 1200 AD. Like you've just had this like, and it looks like a crazy ex-Silicon Valley hockey stick, like even though it's 4%. And the biggest freight forwarder in the world has less than 4 % market share. So you're like, actually, I don't need to compete with anybody.

20:27If I just take the growth in the market, I can be number one and do that every year, right? It's much better to be part of a growing industry than it is to like, oh, let's fight over the knife fight over a shrinking pie. So I wouldn't call it like long run. It's not good for tariffs or bad for business. But short run, we've managed to be nimble and better at helping customers with the problems.

20:49Ryan Petersen:When I launched Upstarts Media last year, I had big ambitions and a lot to figure out. I knew I needed a bank that could move at my speed and grow with me too. With Mercury, I was able to get started in minutes without leaving my desk. Mercury made it easy to send and receive wires and manage the cash flow of our growing business from day one. Now, I can spend less time with busy work and more time with the founders I like to cover. Building a media startup comes with new challenges every day. But with Mercury, banking is one I don't need to worry about. You don't accept outdated tools like floppy disks and fax machines anywhere else in your work and life.

21:23Ryan Petersen:Why accept them for something as important as your money? Visit Mercury.com to learn more and apply online in minutes, like I did. Mercury is a fintech company, not an FDIC-insured bank. Banking services provided through Choice Financial Group and Column NA, members FDIC. When you think about the hardest moments of the business as a leader or intellectually, is it the bold times where you're trying to keep up? Or was it the harder bear times where you're fighting for business? We on the show talk about an upstart moment where you're really back to the wall, trying to change the status quo. What comes to mind first for you?

21:57Ryan Petersen:with that? Well, every great story follows the hero's journey. The story that's great in the sense of like, oh, this will resonate with humans. Humans crave a story like this, which is boy, like ordinary boy, girl, ordinary person, world starts to go well, life is getting better and better, amazing, we're on top of the world, and then everything goes to hell, and then you get the comeback story. You can do that over and over and over. Yes, all the great movies and stories and like that is what you should do in your journalism as you write stories and stuff and I think startup stories should should be like that too your sales pitch should be like that a little bit too like how do I tell a compelling story for the customer and Flexport's definitely had our our climactic moment I mean you put you put me on the cover of Forbes we raised you know several billion dollars eight billion valuation valuation was there customers like profitable lots of greatness happening.

22:54And then we had the fall. And for us, that looked like, well, the COVID boom of containers undid itself. And the price of freight fell from 20 ,000 to 1 ,500. And our revenue is quantity times price, at least in the freight part of our business,

23:10Ryan Petersen:is kind of an easy calculation to be like, oh, wow, the revenue tanked. We had also operationally made some mistakes in terms of we structured ourselves. We try to make it like in a pursuit of efficiency. We built this really great workflow system that assigns work to operators. How do I move a container? And it's very intelligent actually, but was the humans on the side, the management screwed it up where we then assign individuals to own each task in the system and go, you do this task over and over again for everybody. And it was, first of all, it's boring work. And, but more importantly, lack of accountability of like something goes wrong in this container, you're like, I don't know, it was like 50 people that did something.

23:46Who's going to just own it and make sure it goes right? And that led to real quality problems. It's classic. If I'd have been more of a student, which I am now of Toyota production system and lean, that entire universe is all about like the operators trained across all the tasks and like follows the part down the line. Instead of like in the American automotive world, they just do the same job over and over again. We got this problem wrong. So long story short, we took a real fall. We burned too much money. I brought in the CEO, famously, it didn't work out.

24:17Ryan Petersen:I wrote the story at the time about you riding off into the sunset to become a venture capitalist at Founders Fund. And it lasted, what, a couple months? It was not my... Yeah, I was the chairman of the board for about six months. And yeah, so that was sort of like that downward swing that we took. And now that was two and a half, almost three years ago. And now we're like back on that, finally, like the comeback story that should make like sports story the true hero's journey. Was coming back the hardest part for you? Yeah. Yeah, it was. We had to cut costs pretty hard. But the bigger problem was our quality was really suffering.

24:55And like customer relationships we built for a decade had gone in this pursuit of efficiency to like the thing that they were here for was quality, reliable logistics, on-time performance delivered by tech was like it was failing. Right. And if it's not better, they might as well go back to the old guys, right? 100%. The tech is only valuable if it delivers KPIs they care about. On-time performance, savings, consolidation benefits. So, like, that means, like, if I can basically play Tetris inside your containers, how do I load more stuff with algorithms to do that? Yeah, being on time, being more reliable, more predictable, better able to predict when is this cargo going to arrive.

25:32All of these things is what you care about. that's what and like over the last two years two and a half years we just like complete turnaround to where we're now back to being best in class especially in um in customs where we've got ai auditors an ai auditor actually audits every customs entry before we file it and we have humans that also sample and do compliance reviews our error rate in customs is we went from 1.8 which is already quite good in the industry to 0.2 which is just unheard of 0.2 error rate as just one of many, many examples where AI is suddenly like, oh my goodness, it's night and day what we can do on a quality basis.

26:11So revenue growth is really good. Net promoter score is a really good way to measure this. It was 73 last month.

26:18Ryan Petersen:What was it at its nadir? It went down during COVID. I mean, during COVID it was pretty ugly. I think it got to a low of 17 before someone decided that net promoter score was not the reliable metric that we once thought it was, and we don't need to study this metric anymore. And then you guys stopped, and then now you care about it again. And we came right back. I mean, again, I was only the chairman for about six months, but we brought it back and then made that the core focus and have spent the last two and a half years just like completely dialing quality, reliability. And then strategically, like if you're just a freight forwarder, and so freight forwarder coordinates freight moves, like containers most famously, but air freight, trucking.

27:01You're just kind of this coordination layer to get freight moved from here to here. You're on some level of middleman, buy low, sell high. And there's two kinds of middlemen. There's those who just buy low and sell high and add much value. And then there's those who coordinate complexity and streamline things. We've always tried to be that kind. But if you want to be a data provider and you want to be a system of record and an orchestration layer and provide other value here, analytics and AI, Like you can't just do that on the subset of containers that they happen to broker with you. You need to be a technology provider for the enterprise, which means we've had to make our tech over the last three years work regardless if you ship the freight with Flexport or not.

27:41That we still are able to give you this orchestration layer, the algorithmic consolidation. So how do I load a container for you? You can ship the container with whoever you want, but our algorithms will coordinate that loading. Our algorithms will now route containers for you and choose what's the best way to move this cargo. And so that's like the big evolution is like really, really, truly living, taking the tech that we built for ourselves and then opening it up so you can now use it even if we don't, even if you don't buy freight from us.

28:08Ryan Petersen:You recently announced all these AI agents. Were the promise of those agents, was that a light at the end of the tunnel as you're kind of trying to rally the company to maybe feel positive momentum again? Or was AI more of a lucky sort of technological breakthrough that fueled that? When we kind of think about how you get things moving again, what was sort of your thinking? I would say it's lucky. We feel blessed also to be in San Francisco so that we know these people, you know, Sam Altman is one of our Flexport's first investor. He was still a partner at YC when I went through YC. He was a partner there.

28:43So he was one of our first investors and friends. And being in Silicon Valley has been a huge advantage in my view. But it was lucky. Like we, that Chachapiti was what, November of 23 or something like that. We had been going, it wasn't immediately obvious to us that this would be, it should have been, we can just apply this to automate all the tasks. In fact, a lot of the tasks that we have automated, last year, we automated 50 % of all the work. A lot of it we did without AI, just, okay, write a script, move data from here to here, like fill out this form, parse this spreadsheet. And yes, we've been using machine learning for a decade, but like the new forms of large language models and generative AI really take hold.

29:26I would say December of last year was where it just changed everything for us, realizing that it was better at writing code than the engineers were, that an engineer writing code with Claude and Codex are 10 times better and maybe 100 times better than if they didn't have it. We're still finding out the limit. And kind of the auditor that we had, the AI auditor, we actually launched in October of last year. And it was kind of in December, November, where we started to see the results. We knew it would be better, but when we saw like, oh my God, our error rate went down to 0.2%. That was the wake up call for us to go look at everything that we do and say AI can be 10 times better than the human doing it.

30:09It's not just that it's cheaper. It can be 10 times better.

30:12Ryan Petersen:And so now you have the AI agent making the container ships more efficient. You have one doing translation. There's a bunch of them, right? Yeah, we've got lots and like many more. I mean, I think the really interesting how it changes the roadmap, and we've probably pivoted like 30 to 40 percent of our engineers now just to building agents and and then they like there's way less planning than there used to be at flexport in the roadmap like used to be like a one-year strategic plan like you would lay out okay we're gonna do this we're gonna do that now you're kind of like all right let's go see if you can make this work come back to me two days later and it works because of AI coding just like we for example we built an AI agent this week, which we did not have at the beginning of the week.

30:56And now we have, which audits every bill that comes in and reviews them against what actually the bills, meaning from the ocean carriers or trucking companies, and it reviews every bill and determines if it's accurate against what they quoted us, against what the procurement database said, and against what actually happened on the shipment. So it checks the message logs, it checks the event logs of the shipment and says, is this the right amount? And did we invoice the customer correctly? And then, you know, I was just like, we didn't have that at the beginning of this week. And now we do. And I didn't know last week that we would have that.

31:29I thought it would be cool. I told them to work on that. And so we have a list of like probably 50 agents that are in development right now, of which some will end up being customer facing products. A lot of them will just lower our costs and improve quality. But I hope to actually like be able to in the future just go, hey, this suite of agents here just does work for you in your logistics. like don't even worry about logging in because your operations is dialed right now, you know, thanks to AI.

31:57Ryan Petersen:You mentioned Sam Altman being an investor. Have you used the speed dial or taken advantage of that to kind of help Flexport be more on the cutting edge? With Sam? Yeah. Oh, sure. Yeah. Actually, he was a hackathon, Flexport AI hackathon. We got Sam to show up and say some words of encouragement to the team and, you know, judge a couple of things last year. So yeah, Sam's, he's very busy, so I don't waste a lot of time. But I did help him recently. I want to say I'll take credit for this. If you're out there and use Slack and ChatGPT, I'm the reason now that ChatGPT bulleted lists properly get posted to Slack.

32:32They were terrible at this and I fixed that. And Sam is very thankful.

32:36Ryan Petersen:You're welcome, Sam. So it's a two-way street for real. And maybe someday you'll make him money. Benioff responded to my tweet. Speaking of having a powerful Twitter. He was like, oh, let's fix this. And he texted Slack's head of product and like, yo, fix this. That's amazing. With employees, after you had to do these hard decisions, like several rounds of layoffs, maybe the energy isn't great. When you think about sort of what connected or what got people excited again to be doing hackathons, besides just the technology, were there any approaches or messages you had for the team that you feel like in hindsight helped get the company moving faster again?

33:13Yeah, I mean, the first is that Flexport has always had a very compelling, like, mission. We want to make global trade easy. And that is, like, it's not like this contrived thing. I think it's, like, hard if you do some SaaS for insurance or something. It's a little bit, like, few orders removed from, like, how does this impact people's lives. And so we've always had, like, really committed people who just, like, think that this is both an important mission, but also fun, interesting, like, fascinating. I don't know. You get to see the ships and how it all works. You get to meet the entrepreneurs and stuff.

33:44So that's probably the most important thing that kept us all together, that it was really compelling work, a lot to learn. And then mostly culture, I always think, is basically a function of winning, sometimes not always correctly reflected in your share price. Your share price goes up. People are happy, it turns out. So, yeah, getting back to winning, which for us, we didn't talk that much about culture in my first year back. It was like we're going to cut costs. We're going to improve our unit economics. And most importantly, we're going to improve our quality for customers. And that's all we talked about for the first year.

34:16And it was only the second year of the turnaround where it was like, all right, great. We did most of those things. It's going really well. Now let's start talking about culture. And invested a lot last year in leaderships and summits and training. Took all the team to Vietnam earlier this year. Last year we talked, not the whole company, but the leadership team, the top 150 people or so. doing kind of like brought maybe 400 people to Las Vegas for training, enablement, stuff, like culture, building, those things. I mean, lots of things you can do.

34:51Ryan Petersen:So you want it to be fun to work at Flexporting. You want to be fun. Fun is like you want to be winning. Like, okay, how do we win? It's like everyone understands our strategy and our mission and how to win. Like what are we doing to win? Well, as a former competitive chess player, winning is fun for me. I can't really imagine the two not connecting. Exactly, exactly. But just, you know, and the other really, we do these engagement surveys, kind of like how do people feel about their job and stuff. And I write all the questions for those myself. But the one really interesting thing is that the team that had the highest improvement, the biggest improvement in their engagement survey at Flexport was our customs team.

35:29Even though they worked the hardest last year because of the tariffs, you know. And so it's actually like some people think like, oh, working really hard will like burn out the team. I actually think it's the opposite. it, if you're working on something that's important, like working really hard, you can see it in the metrics, like it makes people happier. So it's about showing people that the work is important and that they're getting rewarded for the value they create more than, I don't know, parties or something like that.

35:52Ryan Petersen:So as you think about the future, you're the man of the moment again on tariffs. You're saying, you know, your customs agent is 10x better than a lot of the competition. Do you think that that is going to be long lasting a major business and and point of sort of differentiation for you? Or is this just the moment? And two years from now, it's not tariffs. It's something else that we're talking about. And it's just a series of these kind of spikes. Global trade is here to stay. You know, like I said, since the Mongol invasion, there have been a lot of bad things that happened. You know. They did spread dumplings around the world.

36:25Yeah. Pierogis and poli. Yeah, you're right. There have been some bad things. The 30 years war. There's been ups and downs. Global trade has gone on. And I don't think Trump is the worst thing that's ever happened at global trade by any means. And so I predict there will be more trade in the future and not less, and that you'll need – and that AI will do more and more of the – eventually all of the substantial decision-making and coordination work of trade. It just seems – basically what we've seen in the last three to four months, it makes it obvious and inevitable that it will not only be able to do most of the coordination, but it will also be able to optimize things at a level that humans could never do.

37:02and so I think you could take out probably 20 % of the cost of shipping things around the world because of AI and that's just like massive and if you're the people that do that you have the biggest company in the industry and we're very well positioned to be that company if we don't do it it's purely because of ourselves like there's no one is a better position to execute

37:21Ryan Petersen:that vision so going from like Flexport your guide to tariff chaos to Flexport your AI companion for more efficient shipping, something like that. Yeah. And we're all in on that right now. And I think like if you're not – every entrepreneur I've met with in the last two months is like working harder than we've ever worked and is more excited about work than we've ever worked because of this, because of what's obviously now possible with agentic systems with AI. And I just feel – yeah, I haven't been this energized since like – since we were in high school coding like games and working like around the clock because it was so fun.

38:01Ryan Petersen:So you're having more fun or you're more energized now than you were in your first stint as CEO? Yeah. Oh, yeah. By far. And like even now more than in the past decade at any point. I mean, just what's possible. And every day, I mean, this morning Perplexity announced a Perplexity computer. It looks insanely good. I think we can use that to automatically send it to like a bunch of teams. Let's go test this and see if it can solve this problem we've been working on. So, yeah, I'm incredibly energized. And I think even the tariff problems, they're not going to go away. You will always have to clear customs, right?

38:31And so I think we can be cheaper and better at doing that thanks to AI.

38:35Ryan Petersen:Awesome. Well, Ryan, I could chat to you for hours, but we'll just have to come back in a couple seasons and see if the tariffs are still going. For sure. Yeah, we'll see. We'll see. I don't predict anything's going to get too easy too soon, at least not as long as the current president's in office. Well, we'll have to follow you on X just to keep up then. At Types Fest. Types Fest. All right. Thank you.