Flexport CEO Ryan Petersen declares 'code red' on AI: custom broker error rate cut 10x, new tariff refund product in the works

trunks back.

Yes. Yes. Have a porter with you that brings your trunk with you.

Four porters.

Four porters. Is it like Louis Vuitton that's known for doing this huge trunks? Is that it? Yeah.

Yeah, we got to do that. Uh, anyway, our next guest is Ryan Peterson from Flexport. He is now in the reream waiting room. So, we'll bring him in the TV room. Ryan, how are you doing? Oh, look at this setup you guys. You look like you look like president material. You look presidential. That's my president.

This is my president, Ryan. Great to see you. How are you doing?

What's I'm great, man. I'm back in my This People don't know this is my hometown. I'm in Washington DC.

I didn't know this was your hometown.

Yeah, I'm son of uh son of a government economist.

Really? No way.

There you go. born to be president forced to handle global logistics for fast growing companies.

Okay. Well, I mean the first the the first question that everyone's got to be asking you is is like how are how is the geopolitical conflict affecting your business because a lot of the flexport stuff that happens I imagine across the Pacific across the Atlantic you're not moving crude oil but is it affecting you? What are the knock-on effects? Have you been processing the last uh couple weeks? How you're holding up? We got to print out we got to print out a a a picture of Ryan with his face on Ben Afflac, you know, smoking a cigarette because I feel like you should just put that up in the in the office like every day.

There's so many there are so many completely insulated from there. Yeah, I run a SAS company that you know the 99% retention and you're like whatever happened on the front page of the journal affects me. But take me through

and like a month and a half ago I told my comm team, hey guys, I'm going to take a break. I'm not going to do any press. I just want to lock in on AI. I'm just gonna we got to apply AI everywhere in our business.

And then since then, we had the tariffs get overturned and

now this straight of hormos and I've been asked to do press like more than in my whole life ever combined almost. So um

but I you know impact the the impact on container shipping which is what a lot of people think about flexboard is relatively negligible like prices have gone up. Um but other than that it's not that big of a deal.

Yes. Um the reality is that straight of Hormuz and the Persian Gulf is a culde-sac. It's it's really not that important that you go in there. Yeah. For container shipping. It's a huge deal for oil and

course the price increases due to fuel costs.

Fuel costs and like people get away with what they can get away with in logistics. So some of it is just like ah you know disruption let's raise the prices.

Sure. Sure.

Um the bigger impact is actually on the air freight market.

Okay. where the Middle Eastern airlines own an up about 18% of the world's air cargo capacity. So air freight prices have doubled and we're a big air freight provider too. So that's been

um helping companies navigate that especially air freight uh Asia to Europe is really disrupted. A lot of that would fly into Dubai and and kind of get planes would refuel, get um trans shipped and and then sent on. So, we've actually built like a new product that's actually working really well where we ship cargo via ocean from Asia to Los Angeles and then hot shot truck it to LAX and put it on a plane to Europe. Um,

which is very strange, but apparently I mean it saves people a lot of money and time versus air freight uh direct from Asia.

What does hot shot mean?

It's like it's this concept where the truck doesn't make any stops. It just goes directly to the

Okay, I love it. I might have made that up.

I know. It sounds good to me. uh talk to me about uh the the the tool that you rolled out for uh tariff refunds. Was that the first AIdriven product? I mean, I imagine you've been using AI for years, but was that the first like, wow, okay, we built something new very, very quickly moment for you because I know you've been doing a ton of investing in AI tooling, AI software development, all sorts of things. And I want to get into that, but I want to focus on the the tariff tool first.

Yeah. Well, we've been um there's a ton of stuff that we do with AI that a lot not all of it's customerf facing. In fact, most of it's just automating work and making ourselves cheaper and more reliable. But the tariff tariff refund calculator, not that much AI actually. What what we do is basically just take your data from the government. Whenever you import something, you have a record that sits with Customs and Border Protection in their technology system.

It's not a very easy to use system. They'll give it to you as a CSV file.

Okay.

Uh and then we clean that CSV file and give you a report showing you exactly what you're owed.

Yeah. It's not um not rocket science, but we were able to build it really quickly. Thanks AI.

Uh so we launched it uh I think it came out the terrorists were announced on a Friday. We launched this on Monday. I was very proud of it because

Monday we worked all weekend and Monday morning at 7 a.m. I was on CNBC talking about our new tariff refund calculator, which is pretty showing you, I think, how how the modern world's going to look. like you got to be really good at responding to what's happening in your industry and if you're that fast, you know, I don't I think it'll be a long time before our competitors have anything comparable.

Are Flexboard employees allowed to buy tariff refunds?

Um, I don't know if our employees are, but Flexport is. We'll be we will be announcing

No, I can't announce it right now, but we will be announcing a program to buy people's refunds at a discount and get them cash up front. Um,

makes sense for businesses. So that's something we're we're actively working on. It's a very hot market right now. I think we can build something unique because you know this the the secondary market for this is very interesting. It was trading in the 20 cents on the dollar range 20 25 cents and then the day of the Supreme Court ruling it went to 50 cents

and it's now in the 60s and even I'm hearing some trades in the 70s if your claim is big enough

uh 70 cents on the dollar to get paid now. But I think those guys are going to run into real problems because they have an agency problem. Like if you if you're a hedge fund, you bought someone's tariff refund. How how hard is that guy going to work to actually go get the refund when push comes to shove? Like you're going to have to file paperwork and the government's going to reject your claim and then you got to file more paperwork and like

oh because the original the original owner of the refund still has to go fight for it, but they've already sold it so they're not incentivized to go work. Whereas like it's not like you're trading the claim and then the hedge fund can go fight it because I feel like the hedge fund would be very motivated to get that claim, but they don't have the right to it.

Ultimately, it's the original importer that still has to go get the claim. And so we're going to be there to hold your, you know, hold the hand and make sure we actually go get the claim. So I think we're uniquely positioned to help with that and

create the kind of create the trust across the market. So we yeah, we're going to try to launch that soon. Uh we're very active right now.

Yeah. Talk to me about the other AI initiatives. Um,

yeah. What's What's your what's your approach? Is it to just tell every employee like buy as many AI products as you can, do do a million demos or or is it more like build focus on I'm I'm joking there. That sounds like something you would never do, but uh is it more like, you know, you have like an internal team that's like focused on like shipping experiments? Is it I'm I'm assuming it's like more more internal tools because Flexport customers don't necessarily care what's happening. They just want the they want things fast to run on time and all that stuff.

Yeah. Um we have a Yeah, we're we're in this incredible moment. I I I've declared code red a few months ago

around AI because what we saw we built this product in um for AI audit of customs

back in Octo we launched it in October of last year and basically what we do we we we're a customs brokers we help people clear customs. So we built this and as a customs brokerage you have to review we review about 5% of all customs entries a second time with a human compliance team.

Yes.

And our error rate in that process was 1.8%. It doesn't mean that we violated customs law 1.8% of the time. Just to clarify it's like oh maybe we didn't declare like some free trade agreement that it was eligible for sure etc. Those types of things. Um so we then built in in October we launched this AI auditor. So now we audit 100% of entries

before we transmit it to the government.

Yeah.

And then by November we started to see the data. We were we had gotten our error rate down to 0.2%.

So it's 10 times better

than the humans are.

That's crazy.

And and that was like the first giant wakeup call. You're like, "Oh my god." And there's a lot of work at our company that looks like a customs audit.

Totally.

Like you're like wrangling unstructured data, doing some reasoning about it. Yeah.

Submitting, you know, moving documents around. submitting data to a government or some other party. So we we by early Q1 this year we're like oh it's code red like we we have to be the company that takes AI and automates and does everything with AI and with AI agents in particular in this industry and we have

massive like if we don't do it it's only because of ourselves like and I'm I'm paranoid as hell

sure

that we're like a boomer company that can't get this done because we've got a lot of people who don't have the skill set and like a lot of people who are setting their ways So, I've been just like leading from the front on this. We've got small teams that are kind of really cracked um

and building stuff and some proofs of concept where the AI agents can do most of the work that people do. And we're telling everyone at Flexport like, yo, this is the new reality of our business. And if you're not at the head of the curve in learning how to use AI, how to automate work, how to make yourself more productive. Yeah. Like, let's just live in reality. Everybody's job's at risk, including mine. Like if I'm not able to lead from the front and be the best at this, then someone else needs to be the CEO of the company.

Yeah. Some someone's going to build the AI AI native truly AI native flexport. Like you you're not going to let it be be anyone else.

I think it'll be us, but there's a chance that it's not right. And like we're either become the most valuable company in the world or we're worthless. Like it's crazy crazy fork in the world.

One tip instead of doing we're trying to push companies instead of doing code reds like Mountain Dew code red to do Baja Blast instead. So you declare a Baja blast when you get back to the office.

Yes.

Yeah. Okay.

It's just like a little bit more like positive messaging.

No. No. It's okay to It's okay to declare code red, but every code red must be followed by a Baja blast.

Once once once once things are really turning and you're Baja blasted, just rolling out AI feature after AI feature, you need to declare Baja blast and then everyone doubles down and goes even harder on everything. But at least you know that you're you're you're operating from an even stronger position of strength. That's the goal.

Yes.

Yeah. Well, I'll tell you I I talked to I' I've been texting one of my jobs here is to go learn like who's the best at this. And it's very interesting. There's not that much consensus. I've texted probably 10 like CEOs of kind of like really wellestablished tech companies that they don't look like Flexboard, but they have profiles where there's some manual labor and processes and compliance and things like this. And there's a lot of figuring it out right now. Everybody's the CEOs are super energized and leading from the front, but I don't think there's yet a playbook.

Yeah.

Um for how to come in and use AI agents to do most of the work in a company. But the companies that figure this out are going to run through the economy. And I think

you're going to be able to whole industries are going to get replaced by companies that are good at this or the comp, you know, the companies that are in the industry have to be the ones doing that. And that's our

if we don't do it, I really think the company could be worthless in a few years. It's it's kind of a crazy moment.

Yeah. so many assets and so much uh so much experience like it's definitely your your fight to win. Uh are you updating your timelines on uh more self-driving technology or automation in the physical world based on what you're seeing with Whimo, what you're seeing with GPT models getting better and better like it just feels like

mass driver.

Yeah. Are you going to the moon?

Um we are we are we are very close to that space. So, we operate um five sort of mega fulfillment centers. 5 million square ft of fulfillment and they're pretty manual operations right now. I've been a a robot skeptic in in these buildings.

Um I just think our our teams in those buildings are not that expensive and they're pretty good. Like if you someone I I mean I'm on the record saying like if you have a humanoid robot that has the IQ of a you know AGI of a human and it's only 20 bucks an hour like I'll hire a thousand of them in each of my buildings. And that's kind of what we've done.

Um that that said, we we it sort of messes with the aura of the company if we've got AI everything, automated everything, and then the cargo shows up at this, you know, building full of

uh workers picking items manually and putting away. So there's some argument there. We we've been spending a lot of time with our robotics fulfillments companies that that have

pitches here. It seems kind of commodified. The ROI pitches are pretty good, like maybe too good. I'm a skeptic.

Um, and uh, we'll probably be investing in that space and we'll be a consumer of it. I don't really see us building the robots like I think they're very competitive market. So, that makes a lot of sense

and and and there I think the main the main question is like how do these how do the robot companies actually price this? Are they going to basically lease them to you like it's an like like on some type of hourly rate based on runtime? Are they going to ask you to actually shell out like 20, 30, 40, 50 grand for these things? And then you have to be running the math on, okay, does like how quickly is this thing going to depreciate? You know, how how often am I going to have to replace all these motors? And I'm sure there's bunch of different things they can do to mitigate that. But

yeah, and I I don't think I don't think a humanoid form factor is right for fulfillment centers anyways. Like I mean, you just want wheels um and conveyor belts. That's like the number one demo that at least one humanoid company is doing.

I know. I I don't get it. I mean, you just you don't need the feet. Like, you just need the arm.

Uh, and and then wheels to bring you the project.

You can ride a hoverboard around. You just add the wheels and, you know, hoverboard.

$200 hoverboard gets it done. No, I don't know. Uh,

I know that. That's I mean, there there are a lot of companies working on a lot of

What happened to the hoverboards, by the way? A while. Yeah, they came and wet it so fast.

You can't commute on them. If there's any cracks in the street, you'll fall on your face. It's not good. You need you need to be in a warehouse. Basically,

there's that the only only use case.

Did Were you a hoverboard user, John?

Uh, we had a hoverboard. Do you know Rob? He actually injured himself on one. It was very dramatic at the office. Rob from

Yes. And then we had a ban issued on hoverboards in the office because we were like, I don't know if our

You have a photo of Rob on a

I think I do. text me that please. Yes.

Yeah. Yeah. Yeah. It was it was a dramatic day where we were like I don't know if our insurance policy covers this. It's mostly for like carpal tunnel