GingerOSTrade compliance software for compliance teams
Every code, with its reason.
GingerOS records the basis of classification the moment a code is decided, brings back the SKUs a change touches, and keeps the final say with you.
- Answer customs from the record
5 hours5 minutesReported by the customer - Recheck the catalog
About 20%All of itIllustrative - Hear about changes in time
15 days, via a held entrySame dayIllustrative
A window listing the three trade compliance jobs on this page: answering customs from the record, rechecking the catalog and hearing about changes in time. Each row shows today’s value struck through next to the GingerOS value; the first is reported by the customer, the other two Illustrative.
Answer customs from the record.
The reasoning is written when the code is decided, not when customs asks.
Illustration with sample data: an old-style email thread about a CF-28 on eight entries, with 15 days left to reply and you searching an old archive. A pinned comment marks the only record of the reason, a 2024 email: “go with 8708 per John”.
Illustration with sample data: a GingerOS window for a sample CF-28 traces the 8 entries to one sample part and opens its classification record: the code and its earlier code, four reasons, the approver, the date, how long it is kept and an Export audit pack button.
A CF-28 means days in old spreadsheets and the mailbox of someone who left.
Time to classify one product and write its reasoning
Reported by the customerSEKISUI Diagnostics’ trade compliance team: 2 hours to classify plus 3 to write the reasoning, per product. Now one 5-minute run with the reasoning attached.
A comparison reported by the customer: SEKISUI Diagnostics’ trade compliance team took 5 hours to classify one product and write its reasoning, against 5 minutes with GingerOS.
Recheck the whole catalog.
Every SKU is checked. Only the affected ones reach your queue.
Illustration with sample data: an old-style window shows a sample catalog of 4,210 SKUs as small squares sorted by import value, with the top 20% shaded as reviewed by hand. Of the 38 SKUs a sample notice affects, 15 inside the shaded band sit in the review queue and 23 outside it stay in place, marked in red.
Illustration with sample data: the same sample catalog in a GingerOS window starts from today’s view, then the 4,172 unaffected SKUs fade and all 38 affected ones move into the review queue, largest duty impact first. The top few are listed by part number and amount a year, and a Replay button runs the move again.
Export, VLOOKUP, and a hand review of the top fifth by value. Nobody checks the rest.
Share of the catalog checked per notice
IllustrativeA catalog of 4,210 SKUs and one notice that affects 38 of them.
Illustrative comparison, not a measured result: the share of a sample catalog checked per notice, about 20% today against all of it with GingerOS, drawn as two bars.
Hear about changes in time.
A PLM change or a new supplier sends the affected SKUs to your queue.
Illustration with sample data: an old-style change log rebuilt by hand lists two sample engineering changes and a supplier change with when each was released and when compliance found out: 15 days later from a broker hold, never, and three weeks later from an OEM email. A pinned comment marks the 15-day gap.
Illustration with sample data: a GingerOS review queue of four sample changes this week from PLM, ERP and the supplier master. A material change rechecked 3 SKUs, a supplier switch needs your origin recheck, a coating change had no impact and an origin change waits on R&D.
An engineering change is approved with no field for compliance. You learn from a held entry.
Days from an engineering change to your review
Illustrative
Illustrative comparison, not a measured result: days from an engineering change to your review, 15 days via a held entry today against the same day with GingerOS.
Numbers marked Illustrative are not measured results. Your own baseline is set with you before the build.
Questions you’ll probably ask
When does a company need trade compliance software?
When spreadsheets and email stop keeping up: thousands of SKUs, several entities, plants or brokers that classify the same part differently, rules changing faster than your team can recheck, or a CBP request you can't answer from the record. Below about a hundred imported products a year, a broker and a few good tools usually cover it.
What does reasonable care mean for importers?
It is the importer's legal duty under 19 U.S.C. 1484 to use reasonable care in declaring the classification, value and other facts of an entry, even when a broker files it. In practice it means you can show how each decision was made: the facts, the rule applied, who checked it and when. Falling short can bring penalties without any intent to defraud.
What does a trade compliance program include, and where does AI fit?
Written procedures and named owners for every decision that reaches an entry: classification, valuation, country of origin and trade agreement claims, recordkeeping and export controls, plus training, periodic internal reviews and a way to correct errors. CBP's Reasonable Care publication lists the questions it expects importers to answer. AI can run the research and keep the record; the program and its owners stay yours.
Can it write the reasoning for the codes we already have?
Yes. Your current codes run through the same classification and each one gets its reasoning written down. Where GingerOS reaches a different code, it is flagged for your review, not changed. You work through them in batches, largest duty impact first.
Can ChatGPT classify HTS codes?
It can suggest one, but a chat answer leaves no record, has no trigger and none of your data. GingerOS keeps the reason behind every call, knows which rule dates applied, and brings the SKU back when your parts, suppliers or the rules change.
Will it replace my judgment?
No. You approve every final decision. An override keeps your name and reason and becomes the precedent for that part; a rejection sends the item back with your reason instead of closing it.
How accurate is AI HTS classification?
It depends on what was measured, so we measure it on your products, not ours. Before anything is accepted, we agree a test set, an answer key from your experts and a pass line, and GingerOS has to clear it.
What if data is missing, or R&D won’t share the formula?
It asks the smallest question it can, one at a time, and waits. R&D can answer with a range or a yes or no instead of the formula.
Which countries does it classify for?
The U.S. and Canada, to the full 10-digit code, and the EU in beta, to the 8-digit CN code. Other countries are scoped per project.
What will GingerOS never do?
File entries, act as your broker, give legal advice or promise compliance. It helps you document your own reasonable care.
