Tarifflo Turns the Most Boring Paperwork in Global Trade Into a Refund Button

Ford Motor Company once paid a $365 million penalty over a single wrong number. Not a rounding error in the billions, not a factory recall - a tariff code. One digit in the Harmonized Tariff Schedule sat in the wrong place, and the difference between a 2.5% duty and a 25% duty became a nine-figure lesson. If you want to understand why Tarifflo exists, start there: the code that moves the goods is the code that can sink the balance sheet.

Every year, more than 400,000 companies import roughly $4 trillion in goods into the United States. Nearly all of it has to be classified - assigned an HTS code that tells customs what the product is and how much duty it owes. For most of history, a human did that. A licensed customs broker, paid around $100 per classification, working through a schedule thousands of pages long, sometimes taking weeks. And still getting it wrong an estimated 30% of the time.

Tarifflo, founded in 2024 in Provo, Utah, and now part of Y Combinator's Summer 2026 batch, was built to do that job with agentic AI instead - and then to do everything that happens after it.

01 / THE FOUNDERFrom a KPMG desk to a coding contest

Bryce Judy did not read about this problem. He lived it. As a licensed customs broker doing trade compliance at KPMG's Salt Lake office, he classified products by hand for large importers - including work on names like Dick's Sporting Goods. He knew exactly how slow, expensive, and error-prone the process was, because he was the process.

He met his co-founder, Tanner Helms, at a coding competition while at Utah State University. One knew the domain cold; the other could build. That split - deep compliance expertise paired with engineering - is the reason Tarifflo reads less like a generic AI wrapper and more like software written by someone who has been fined for getting a code wrong.

02 / THE PRODUCTAn AI that reasons like a broker

Most classification tools lean on traditional machine learning, which has a fatal flaw in this market: tariff rules change constantly, and a model trained on last quarter's schedule is already stale. Tarifflo's approach is agentic. Its AI agents work through a classification the way a human expert would - weighing product attributes, consulting customs rulings, and producing not just a code but the legal rationale behind it. The underlying database refreshes nightly, so the answer keeps pace with policy.

Speed is the part that reads like a magic trick. Whether you hand it 500 products or 10,000, Tarifflo returns classifications in roughly the same 200 seconds. Compare that to a broker's per-item fee and multi-week turnaround, and the value proposition stops being subtle.

03 / THE PLATFORMBeyond the code itself

Classification was the entry point, not the destination. Judy has said the goal is to be "the one-stop shop for anything trade compliance related, supply chain management." The platform now stretches across the full lifecycle of an import - finding overpaid duty, filing directly with customs, clearing agency paperwork, and mapping where goods actually come from.

04 / THE MODELOnly paid when you get paid

Here is the line that makes people lean in: "1 in 8 products overpay on tariffs. Find it free." Tarifflo will audit a sample of your past entries at no cost, and on the recovery side it only charges when it actually claws money back. There is a clock on this - overpaid duty is recoverable for roughly 300 days - which turns the pitch into something urgent rather than aspirational.

For a beauty and personal-care importer, Orly Corporation, that math produced $1.3 million recovered after Tarifflo audited more than 10,000 items. The company also runs a more traditional SaaS side, with annual plans historically ranging from about $2,500 to $75,000 depending on size and volume. Two doors, same building: subscribe for ongoing classification, or let the audit pay for itself.

05 / THE TIMINGA billion new problems, all at once

Trade compliance spent decades as invisible back-office work. Then it landed on every front page. Tariffs shifted almost daily through 2025, and the revocation of the De Minimis exemption - the rule that let low-value shipments skip formal entry - created an estimated one billion additional classifications a year, effectively overnight. A chore became a crisis, and a crisis became a market. Tarifflo was already building for exactly that world.

As co-founder Tanner Helms puts it, the stakes are unforgiving: even a single incorrect code can trigger unexpected fees, shipping delays, and steep fines. That is the same edge Ford fell off of - and the reason a fast, defensible, auditable classification is worth more than it looks.

06 / THE RECORDFrom student prize to YC

07 / THE MARKETWhere it fits

Tarifflo sits between three incumbents that never quite fit together: licensed customs brokers who do the work by hand, legacy trade-compliance software that stores rules but doesn't reason over them, and a new crop of AI classifiers. Its bet is that domain-trained agentic reasoning, a nightly-updated rules database, and a pay-on-recovery model are hard to copy from the outside. The customers - importers, freight companies, brokerages themselves, and public companies - are the ones who feel every wrong code in cash.

Whether Tarifflo becomes the platform for global trade compliance or one strong player in a crowded field, the underlying observation is sound: an industry that moves trillions still runs on manual paperwork, and someone was always going to automate it. The founders just happen to have felt the pain from the inside first.