Control exports and imports: what software can run and what needs a person

Screening parties against restricted lists, matching shipments to licences, filing routine customs entries and keeping records can run on software today. Classifying new products, deciding on licence exceptions, resolving screening matches and handling customs inquiries need a person. None of it works until product master data carries reliable classification and origin.

Where software already carries the load

Restricted party screening is the clearest case. A screening engine checks customers, suppliers, banks, freight forwarders and end users against government lists every time an order, shipment or payment is created. It also rescreens the whole partner base when a list changes. People cannot do this at any useful speed. Fuzzy matching on names and addresses is mature, and AI models now help rank likely matches so reviewers see the serious ones first.

Licence determination also suits a rules engine, provided the inputs are clean. Once a product carries an export control classification, the system can compare it with destination, end use and end user. It then decides whether the shipment needs a licence, falls under an exception or can go freely. It can also track how much of an existing licence has been drawn down and block a shipment that would exceed it. This is the same logic a funds control system applies to a budget line.

On the import side, software can build customs declarations from commercial invoices and packing lists, apply tariff rates, calculate duty and taxes, and file electronically with the broker or the customs authority. Document extraction tools read invoices in many formats and populate the entry. Record retention is a natural fit too. Every declaration, certificate and screening result can be stored against the transaction it supports.

Where a person has to decide

Classification of a new or changed product is a judgment call. A tariff code or an export control number depends on technical characteristics, intended use and how a regulator reads the wording. AI can suggest a code from a description. A trained classifier should still confirm it, because a wrong code repeats on every future shipment.

Screening alerts need a human to clear or escalate them. The software flags a possible match. Someone has to establish whether the party really is the listed one and document why.

Other decisions that stay with people:

• Whether a licence exception genuinely applies, and whether the end use raises red flags the rules cannot see
• Responding to customs audits, requests for information and penalty notices
• Voluntary disclosures when an error is found after the fact
• Origin determinations for free trade agreement claims, where supplier evidence is thin

Valuation adjustments sit in the same category. Assists, royalties and related party pricing rarely appear on the invoice, and someone in finance has to recognise when they should.

What has to be true about the data first

The product master is the foundation. Each item needs a tariff code, an export control classification and a country of origin. Each value should be maintained by an accountable owner, with a date showing when it was last reviewed. If classifications live in a spreadsheet held by one specialist, automation will simply move bad codes faster.

Partner data has to be complete enough to screen. That means full legal names, addresses and ultimate consignees, not just the freight forwarder. Ship to and bill to records should be distinct.

Transaction data must link up. The order, the shipment, the customs entry and the payment should share references so a reviewer can trace any declaration back to its source. Supplier origin declarations and certificates need to be stored where the trade system can find them, not in email.

Finally, the rules themselves need an owner. Lists, tariff schedules and licence conditions change. Someone must update the screening configuration and the determination logic, and record what changed and when.

Questions to ask the people who run it

What is documented often differs from what happens on the warehouse floor and at the broker's desk. Ask:

• When a screening alert fires, who clears it, and what do they write down?
• Has anyone ever released a held shipment to meet a deadline? Who approved it?
• Where do classifications come from when a new product launches? Who signs off?
• Does the broker ever change a tariff code without telling anyone internally?
• Which shipments bypass the trade system entirely, such as samples, returns, repairs or hand carried items?
• How are licence balances tracked today, and has a licence ever been exceeded?
• When customs sends a query, who sees it first?

The answers to the bypass question usually reveal the biggest gap.

How the controls are evidenced

Trade controls get tested like any other internal control. Auditors and regulators ask for cycle memos describing the flow, samples of transactions with their screening results and declarations, and proof that exceptions were approved. A well run system produces this evidence as a by product. When reviews turn up errors, such as a misclassified item or an underpaid duty, the correction should flow back into the product master and the rules, and post entry amendments should be filed where required. Without that loop, the same finding returns at the next review.

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.