Case study · Engineered metal components · United States, 2026

Two weeks with Trade Insight AI saved $10 million in duty.

A US importer of castings, forgings and machined parts ran its hardest classifications through Trade Insight AI. The engine confirmed its HS codes, showed the duty at stake on every difference, and moved hundreds of parts off an HS code that an antidumping petition named four months later.

See how it works
1The problem

Three HS codes for one part, and no way to check any of them

With Section 232, Section 301 and the IEEPA actions stacked on base rates, the same product family could carry anywhere from 45 to 75 percent duty. Which stack applied turned on details that were easy to get wrong: where a casting was machined, how finished a gear blank was at the border, whether a part belonged in Chapter 87, 84 or 73.

Take one gear blank, one of hundreds of similar parts the company imports. It had three HS codes, from a consultant, a chatbot and its own staff. None came with reasoning anyone could check against the tariff. A production shipment was a week away.

ONE PART OF HUNDREDS ·THREE ANSWERSOutside consultantHS code Ano reasoningAI chatbotHS code Bno reasoningLogistics staffHS code Cno reasoningDUTY AT STAKE45 – 75%SHIPMENTIN 7 DAYS
FROM 200 SKUs TOTHE ONES THAT MATTERED200SKUs classified in minutes,each with its reasoning6highest-risk items flaggedas likely misclassifiedSix to reclassify.Everything else confirmed.A written reasonable-care findingfor the broker, counsel or CBP.
2What we did

200 HS classifications in minutes, and the six that mattered

The company's 200 highest-duty, highest-value and highest-volume SKUs went through Trade Insight AI in a matter of minutes, each coming back with an HS code and the reasoning behind it. The only wait was for product details from the company.

Out of those 200, the engine flagged the six highest-risk items most likely to be misclassified. Everything else held up, with a written reasonable-care finding the company can hand to its broker, counsel or CBP.

3How it saved $10 million

Hundreds of parts, moved off an HS code that an antidumping petition named four months later

HUNDREDS OF PARTS ·TWO HS CODES · FOUR MONTHSCOMPANY'S ORIGINAL HS CODEParts of linear-actinghydraulic cylindersordinary duty: freeTIA'S HS CODE,WITH THE REASONOther hydraulic motor partsordinary duty: freeThe six reclassifiedhundreds of parts movedfour monthsSeptember 2026AD/CVD investigations initiatedPETITION SCOPE NAMESthe original HS code85 – 371%dumping margins at initiationWHERE THE CLIENT STANDSOn the correct HS code,technical basis on file,binding-ruling request ready.
Annual savings of $10M

When the six were reclassified, one of them, a duty-free hydraulic motor part, moved off the HS code for parts of linear-acting hydraulic cylinders and onto the one for other hydraulic motor parts, with the reason cited. The same reasoning applied to hundreds of similar parts that were never in the original review, and they moved too. Both HS codes were duty-free, so a review that looked only at rates would have skipped all of them.

Four months later, in September 2026, the Commerce Department opened antidumping and countervailing duty investigations on linear hydraulic cylinders and parts from five countries, the client's source country among them. The petition names the old HS code, with margins of 85 to 371 percent.

Where the $10 million comes from
  • On the correct HS code, with the technical basis on file and a binding-ruling request ready to go.
  • Scope turns on the written description, which is exactly what the enriched description and the cited rationale give them.
  • Without that second pass, hundreds of parts would have been sitting on the named HS code when the petition was filed, facing cash deposits of 85 to 371 percent on every entry.

Trade compliance AI you can explain, defend and trust.

Trade Insight AI classifies products for customs from the General Rules of Interpretation, the section and chapter notes, and the rulings. Every HS code comes with the headings considered, the notes applied and the rulings cited, so it can be handed to counsel, a broker or CBP. Teams run it themselves on their own product data. The platform does not train on customer data.

See how it works

Case study: engineered metal components, United States, 2026. Duty rates shown are the ordinary rates quoted in the review.