Process trade preferences: what to automate, what to keep human, what data comes first

Software can already run the origin calculation for each product and chase suppliers for their declarations. A person must still decide borderline tariff classifications and sign anything that goes to customs. None of it holds until bills of material and tariff codes are accurate and every purchased component carries an origin with evidence behind it.

What the work involves

Claiming a preferential duty rate under a trade agreement means proving that a product qualifies under that agreement's rules of origin. The proof rests on what the product is made of, where each input came from and what happened to it in the factory. Once a product qualifies, the exporter issues an origin statement or certificate, the importer claims the lower rate, and everyone keeps records in case customs asks later.

The finance link is direct. Duty saved is real cash, and a claim later found invalid becomes duty owed, often with penalties attached.

Steps software handles well now

The qualification calculation is the clearest candidate. Given a clean bill of material, tariff codes for each component and their origins, a rules engine can test a finished good against tariff shift rules or regional value content thresholds far more consistently than a spreadsheet. It can rerun the test whenever a supplier, a cost or a recipe changes, which is where manual programs tend to drift.

Supplier solicitation also suits automation. Portals can send declaration requests, track who has replied, flag expired declarations and escalate the silent ones. AI models are useful for reading returned documents in varied formats and pulling out the part number, the stated origin and the validity period for a person to confirm.

Generating the origin statement on a commercial invoice can be automated once qualification is settled. So can matching shipments against qualified products and assembling the record file that supports each claim.

AI can suggest tariff classifications from product descriptions and drawings. Treat these as a first draft. The suggestion speeds up the specialist; it does not replace the specialist's decision.

Steps that still need a person

Final classification of anything ambiguous stays with a trained classifier. A single heading choice can decide whether a product qualifies at all, and customs will hold the importer or exporter responsible, not the tool.

Someone has to interpret the agreement itself. Rules differ between agreements, contain exceptions and get amended. Deciding which agreement to use for a given lane, and whether a cumulation provision applies, is judgement work.

Signing certificates and declarations is a legal act. The signer needs to understand what they attest to.

Verification requests from foreign customs need a human response. They arrive with deadlines and specific questions, and the answer usually requires pulling cost records, production evidence and supplier support together into a coherent argument.

Supplier relationships matter too. When a key vendor refuses to provide a declaration or provides one that looks wrong, a buyer or trade specialist has to have that conversation.

What the data has to look like first

The bill of material in the system must match what the plant actually builds. Engineering versions that never reached production, or substitutions made on the floor, will produce confident and wrong results.

Every component needs a tariff code assigned at the level the rules require. Codes copied from an old import entry or guessed by a buyer are a common weak point.

Origin per component must come from evidence. A country of origin field filled in from the supplier's address is not origin. The record should link to a valid declaration or a documented analysis.

Cost data needs to be traceable if value content rules apply. Standard costs, transfer prices and freight treatment all affect the result, and finance should agree on which figures are used before anything is automated.

Part numbers must connect cleanly between purchasing, engineering and the trade system. If one component has several numbers across these, the engine cannot see that a declaration covers it.

Retention matters as well. Supporting documents must be stored so they can be retrieved by product and shipment for as long as the agreement requires.

Questions to ask the people who run the process

What the procedure says and what the team does are often different. These questions tend to surface the gap.

• When a supplier never returns a declaration, what actually happens to that part in the qualification?
• Which products are claimed every time without anyone rechecking them, and why are those trusted?
• Where do the tariff codes in the system come from, and who last reviewed them?
• How does anyone learn that engineering changed a component or that purchasing moved to a new source?
• Is there a spreadsheet outside the system that someone relies on to make the final call?
• Which customs verification was hardest to answer, and what was missing when it arrived?
• Who decides which agreement to claim when more than one could apply?
• Are there claims being skipped because qualifying seemed like too much effort, even though the product probably qualifies?

The answers usually show where the data breaks and which judgements live in one person's head. Both need fixing before any tool is switched on.

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.