The most valuable problems tend to hide in the least interesting places. Terrell Marshall spent years on products that a billion people opened every day, then walked toward one that almost nobody thinks about: the mountain of forms, codes, and email threads that stand between a shipping container and a warehouse. That mountain has a name in the industry. It is called customs clearance, and Marshall is building the co-pilot for it.
His company is Passage, part of Y Combinator's Summer 2024 batch. The pitch is short and hard to argue with. Every company that brings goods into the United States has to clear them through customs. Behind that simple sentence sits an army of brokers and entry writers who spend their days keying numbers off invoices, cross-checking documents, and answering the same emails again and again. Passage wants to hand each of them an AI co-pilot and give those hours back.
01 / THE PIVOTFrom consumer AI to container ships
Marshall's resume reads like a tour of consumer technology at its largest scale. At Google he worked on AI features inside Google Photos, the app that quietly edits and organizes the pictures of more than a billion people. He also spent time on Google Payments and on Hotels Search, the kind of surfaces where a small design choice ripples out to millions of taps a day. Before that, he was a software engineer at Workday, closer to the machinery of enterprise software.
So the obvious question is why someone with that background would choose customs. Freight paperwork does not trend. It does not demo well at a party. But it has a quality that consumer AI often lacks: the pain is specific, expensive, and impossible to ignore. A misclassified product or an inconsistent invoice can mean penalties and delays that cost real money. The people doing the work know exactly what they need. They just have not had software that speaks their language.
02 / THE PARTNERA college friendship, a patent, and a second act
Marshall did not start alone. His co-founder and Passage's CTO is Aditya Nadkarni, a friend he met in college. The two had already built something together long before customs entered the picture. They hold a patent for an AI platform in physical therapy, an early sign that this pair likes to point machine learning at practical, real-world work rather than technology for its own sake.
Nadkarni brought his own logistics scars to the table. He built AI document processing at Flexport, the freight forwarding company, and worked on doctor-patient communication software at Artera. Between the two founders, Passage has someone who has shipped consumer AI at planetary scale and someone who has already wrestled with the exact documents customs runs on. That combination is rarer than it sounds.
03 / THE PRODUCTTeaching software to read a bill of lading
The product itself is built around a handful of jobs that customs teams do by hand today. It pulls data out of commercial invoices, bills of lading, and packing lists, then pre-fills entry filings in seconds instead of hours. It cross-checks information across those documents to catch the inconsistencies that lead to penalties. It handles product classification, working through HTSUS codes, cross-rulings, and ADD/CVD order lookups so a broker can classify with confidence. And it reaches into the inbox, extracting the details that kick off a workflow and drafting the replies that used to eat an afternoon.
Read that list again and notice what it is not. There is no flashy consumer surface, no viral loop, no attempt to reinvent how people feel about anything. It is a set of quiet, high-stakes tasks where being correct matters more than being clever. That is the terrain Marshall chose on purpose.
04 / THE STYLEDriven, and allergic to fluff
People who describe Marshall reach for the same two words: driven, and no BS. It fits the choice he made. A founder chasing applause would have picked a warmer, more visible category. Marshall picked the one where success is measured in filings completed and penalties avoided, and where the customer is a professional who can tell in a minute whether your software actually understands their job.
There is a discipline in that. Consumer AI can hide behind delight. Enterprise plumbing cannot. If Passage misreads a document, a broker notices immediately, and the trust is gone. Building here forces a kind of honesty into the product, and it rewards founders who have, in the words of one profile of him, been in the trenches building, experimenting, and failing before things clicked.
05 / THE BETWhy boring might be the whole point
There is a pattern to the current wave of AI companies, and Passage sits squarely inside it. The splashy demos get the attention. The durable businesses get built one workflow at a time, inside industries that outsiders find dull and insiders find indispensable. Customs is a textbook case. It touches nearly everything you own, it runs on documents and email, and it has been waiting a long time for software that treats it seriously.
Marshall's wager is that his years on billion-user products were not the peak but the training. The scale taught him how AI behaves when real people depend on it. The customs work gives him a place to apply that lesson where the stakes are concrete and the buyers know precisely what they want. Passage is small, two people carrying an outsized ambition, and that leanness is part of the story too. It is early. But the shape of the bet is clear, and it is a serious one.
The next time a package arrives from overseas, there is a decent chance it cleared a wall of paperwork on the way. Marshall would like that wall to be a little shorter, and a lot faster. That is the whole idea behind Passage, and it is the kind of idea that only gets more valuable the more the world ships.