Calculate duty: what software can take over and what still needs a person

Software can already look up tariff rates, convert currency, build the customs value from captured costs and check broker statements against internal figures. People still need to own classification of new goods, origin decisions and valuation judgments. None of it works until the product and purchase data are trustworthy.

Where software does the work well

Once a product carries a confirmed tariff code and country of origin, the arithmetic is mechanical. A trade system can pull the current rate for that code, apply any preferential rate the goods qualify for and add anti-dumping or countervailing measures where they apply. It can convert the invoice value using the exchange rate customs publishes for the period, which often differs from the rate treasury uses.

Building the customs value is also rules driven when the inputs exist. The system starts from the invoice price, then adds freight and insurance up to the border or removes them, depending on the delivery terms on the order. It flags lines where those terms are missing.

Reconciliation is the other strong candidate. Brokers file entries and send duty statements. Matching each entry to the purchase order, the receipt and the supplier invoice, then comparing the duty charged with the duty expected, is tedious for people and easy for software. Variances above a tolerance go to a person. Everything inside it posts.

Accruals and landed cost follow from the same calculation. When expected duty is computed at receipt, finance can accrue it before the broker invoice arrives and push it into inventory cost without waiting for month end.

Where AI helps but should not decide

Classification is where AI gets the most attention. A model reading product descriptions, specifications and images can propose a tariff code and show its reasoning. That speeds up a classifier working through a backlog of new items. It does not replace the classifier. A wrong code is the importer's liability, and the model cannot weigh the general interpretive rules against a ruling the business received years ago.

AI is also useful for reading supplier origin declarations and certificates, pulling out the fields and spotting ones that have expired or name the wrong product.

What still needs a person

Origin under a trade agreement often depends on the bill of materials, where each component came from and how much processing happened locally. Deciding whether goods meet the rule is a judgment, and someone must be willing to sign for it.

Valuation between related companies is similar. Transfer pricing adjustments made after import can change the customs value. Royalties, tooling supplied free to a manufacturer and buying commissions all raise questions that need a trained person and sometimes an advance ruling.

When an error is found in past entries, deciding whether and how to disclose it to customs is a human call with legal weight.

Duty relief schemes, such as inward processing, bonded storage or drawback on re-exported goods, need someone who understands the conditions and keeps the records that prove them.

What has to be true about the data first

The product master must hold a tariff code and origin for every item that crosses a border, with a named owner and a date of last review. Descriptions need to be specific enough for a classifier to work from. A short internal name is not enough.

Purchase orders need delivery terms on every line. Freight and insurance costs must be captured against the shipment they belong to, not lumped into a monthly carrier invoice.

Related party suppliers should be flagged. Bills of materials should show component origin wherever preferential origin is claimed. Supplier declarations need to sit somewhere the system can read them, with expiry dates.

The broker's entry data must come back in a structured form. Scanned PDFs alone leave the reconciliation manual.

Questions to ask the people who run it

• When a new product arrives, who actually picks the tariff code, and do they write down why?
• Does the broker ever classify goods themselves because the code was missing or looked wrong?
• Which exchange rate is used in the internal calculation, and does it match what appears on the entry?
• How are freight charges split across shipments when one carrier invoice covers many?
• What happens when the broker's duty figure differs from the expected one? Who chases it, and how often is it simply accepted?
• Are preferential origin claims made on every eligible shipment, or only when someone remembers to attach the certificate?
• Where do supplier origin declarations live, and who notices when they lapse?
• Have transfer pricing adjustments ever been reported to customs?
• Is there a spreadsheet that sits outside the trade system and holds the real logic?

The answers usually reveal a manual workaround nobody documented. That workaround is often the thing an automation project will break first.

APQC's Process Classification Framework® (PCF) is an open standard developed by APQC, a nonprofit that promotes benchmarking and best practices worldwide. To download the full PCF or to view definitions and measures, please visit www.apqc.org/pcf.