❓ Problem

Until recently, most people never met a customs bill. The $800 de minimis threshold meant your parcel from Shenzhen, Seoul or Berlin just turned up. That era is over. De minimis was suspended for China in May 2025, extended globally that August, and eliminated for every country in February 2026. CBP made the suspension indefinite by regulation in June, the trade court upheld it in August, and statutory repeal lands in July 2027. The EU is following, with a €3 flat fee on the roughly 4.6 billion sub-€150 parcels arriving each year.

The practical effect is enormous. CBP processed 1.36 billion de minimis shipments in FY2024, about four million parcels a day, and every single one of those now needs a formal entry: a 10-digit HTS classification, an origin check and a duty calculation. Brokerage runs $25 to $75 per entry, merchandise processing fees add another $2 to $9, and a $12 product from China can now carry $14 to $24 in combined duties and processing charges.

Here is the kicker. Those classifications are being made at machine speed by people with no incentive to get them right, and misclassification is already the single biggest cause of unexpected duty bills. Businesses routinely pay 10-30% more than they owe because a broad category got used instead of a precise subheading, or a Chapter 99 overlay was applied when an exclusion existed. The recourse exists on paper: a protest within 180 days of liquidation in the US, form BOR286 in the UK. Almost nobody uses it, because the recourse is shaped like a 1970s filing cabinet and the person holding the charge label has no idea whether the number on it is right.

And the money is provably enormous. When the Supreme Court struck down the IEEPA tariffs, it created a $166 billion refund pool across 330,000 importers and 53 million entries. Even with a purpose-built government portal, CBP rejected nearly 30% of filings, thousands of approved claims stalled on bad banking details, and the payouts skewed hard to large importers with proper customs teams. Ford booked $1.3 billion while households, who absorbed roughly $1,000 each in tariff costs last year, got essentially nothing. Filing skill, not eligibility, is what decides who gets paid.

✅ Solution

An AI customs agent that audits an import charge and files the claim when it is wrong.

• Upload the paperwork, get a verdict in 30 seconds. Charge label, commercial invoice, courier receipt. The agent extracts product description, declared value, origin and the code that was actually used.
• Re-derive the classification from scratch. It independently determines the correct 10-digit code, cross-references Chapter 99 overlays and active exclusions, checks the duty maths, and separates legitimate duty from the courier's own brokerage and advancement fees, which are a different fight with a different counterparty.
• File through whichever door actually applies. A post summary correction before liquidation, a protest after, BOR286 or C285 in the UK, or a direct courier dispute where the carrier acted as importer of record. Routing is most of the value, because the average person has no idea which of those four situations they are in.
• No win, no fee. Free check, percentage of anything recovered.

The consumer product is the wedge and the data engine. Every audited charge label teaches the system which SKU categories, which lanes and which couriers generate errors, which is exactly the dataset you then sell back to e-commerce brands, forwarders and brokers as pre-clearance software.

📊 Key Numbers

Market size

• US customs duties hit $194.9 billion in FY2025, and $264 billion in calendar 2025 as the effective tariff rate climbed to 7.7%, the highest since 1947. At a conservative 3% overpayment rate, that is a $6-8 billion annual pool of wrongly assessed duty in the US alone.
• The parcel slice is the addressable wedge. 1.36 billion formerly duty-free shipments at $20-30 of duty and fees each is roughly $27-40 billion of brand new, parcel-level charges. If even 5% is miscalculated, that is well over $1 billion a year sitting with consumers and small importers who will never chase it.
• Adjacent and immediately upsellable: about 80% of eligible drawback goes unclaimed every year, roughly $15 billion left with CBP.

ARR potential

• Consumer alone is a small business. 500,000 audited parcels a year, a 15% credible claim rate, 60% success and a $60 average recovery is about $2.7m returned, or under $1m of revenue at a 25% fee. Treat it as marketing, not the model.
• The SMB importer tier is where it pays. A business paying $300k a year in duty with a 5% error rate has $15k to recover, worth $3,750 at a 25% contingency. Two thousand of those is $7.5m, and it recurs annually because the errors recur.