When I began working in exports, a shipment generated a small mountain of paper. Quotations, orders, invoices, packing lists, certificates, insurance documents, bills of lading and letters of credit passed between companies, banks, carriers and government offices. A document might travel separately from the goods and still have to arrive first. A discrepancy of a few kilograms could delay payment for weeks.
Much of that information can now be created, checked and transmitted electronically. Artificial intelligence promises to take the next step: reading an order, assembling the required data, preparing documents, identifying inconsistencies and monitoring the shipment without someone retyping the same facts at every stage.
How close does that take us to an export process without people? The answer depends on whether we mean paperless, automated, or genuinely human-free. They are three different things.
The digital foundations are already here
The UK’s old CHIEF customs system has gone. Import and export declarations are now made through the Customs Declaration Service (CDS). Electronic submission is established, although the declaration still has to be correct and the goods still require customs clearance. (HMRC on CDS; export clearance)
Internationally, the Harmonized System gives customs authorities a common starting point for classifying goods. It is maintained by the World Customs Organization, and its first six digits form a shared international structure. Countries can add further digits and apply their own tariff measures and restrictions. An HS code is therefore a useful digital key, but it is not a complete product description or a guarantee that the classification is right. (WCO on the HS; HMRC commodity code guidance)
Booking, schedules and tracking have also moved online. We can receive updates on a vessel, a container or a delivery without waiting for a telephone call. Ports are exchanging more information electronically: since January 2024, International Maritime Organization member states have been required to use maritime single windows for ship-clearance information. That requirement does not make every commercial document or customs process paperless, but it is an important part of the infrastructure. (IMO Maritime Single Window)
The UK Electronic Trade Documents Act 2023 allows qualifying electronic trade documents, including bills of lading, to have the same legal effect as their paper equivalents. The International Chamber of Commerce’s eUCP rules provide for electronic presentation under documentary credits when the credit is subject to them. The legal and banking foundations are developing, though individual traders, banks, carriers and countries must still be able and willing to use compatible systems. (Law Commission; ICC eUCP)
Even our conversations have changed. An exporter can meet an agent in Seoul, a buyer in Dubai and a shipping line in Rotterdam from an office in Yorkshire. The journey that once took days may now begin with a video call. The relationship, however, still needs building.
What AI could do with an export order
Imagine a British manufacturer receiving an order for equipment destined for a new customer overseas. An AI system connected to its approved company records could compare the order with the quotation, check the agreed delivery term, retrieve specifications, identify possible commodity codes and flag whether an export licence or product certificate may be needed.
It could ask production for an available date, compare that date with packing requirements and shipping options, prepare a draft commercial invoice and packing list, and check that quantities, weights, values, addresses and descriptions agree. As the shipment moves, it could monitor milestones and alert the export manager when a vessel is delayed or a document has not arrived.
Under a documentary credit, it could compare the draft documents against the credit’s stated requirements before presentation to the bank. That would be especially valuable to anyone who remembers discovering, too late, that the invoice, transport document and credit described the same goods in slightly different ways.
None of this requires AI to make every decision. Much is ordinary software automation: transferring validated data between systems and applying clear rules. AI becomes useful when it must interpret correspondence, extract information from varied documents, suggest a classification, detect an unusual pattern or explain an exception. It can shorten the time spent finding a problem, but the underlying facts must come from somewhere reliable.
An invoice generated from an incorrect order is still an incorrect invoice. A system that copies the wrong weight consistently across six documents has created consistency, not truth.
The shipment is more than its paperwork
The physical side of exporting refuses to disappear into a screen. Somebody must make the goods, select the correct items, inspect them where required, protect them against the journey, mark the cases and load the vehicle or container. A driver collects them; terminal staff handle them; a carrier moves them. Customs and other authorities may select goods for examination. At the receiving end, people unload, inspect and sometimes discover that what arrived is not what was ordered.
Machines can assist with many of these tasks. Warehouses can scan and sort; cameras and sensors can record condition, seals and temperature. Yet a digital record of a packed case is only as sound as the link between the record and the contents. If a component is substituted on the factory floor after the system has prepared the documents, someone must catch the change.
This is why the export process remains a chain of skills and disciplines. The packer, the driver, the freight forwarder, the customs specialist, the banker and the export manager each know something that the others may not. An AI system may connect their information more effectively, but it cannot assume that one participant’s entry tells the whole story.
Where judgment and responsibility remain
Classification is a good example. AI may suggest an HS heading from a catalogue description. The correct classification can depend on material, function, composition, technical details and the legal rules for interpreting the tariff. The destination country may then require additional digits and impose measures that the six-digit international code does not settle. A confident answer based on an incomplete description could be expensive.
The same applies to origin, valuation, licences, sanctions, product standards and the identity of the customer or end user. A machine can screen names and prompt questions. It cannot discover an undisclosed end use merely by filling in a form. Nor should an exporter assume that employing a software supplier or customs agent transfers away its duty to check the information: HMRC says businesses remain responsible for due diligence when an agent acts for them. (HMRC guidance)
Commercial decisions are harder still. Should we accept a new buyer’s payment terms? Can production genuinely meet a delivery date? Does the agent understand the market? Is the buyer’s requested change innocent, or does it alter the risk of the transaction? An experienced export manager brings together contracts, finance, logistics, compliance and human relationships. No single specialist knows every discipline in equal depth, which is precisely why the manager must know when to call one in.
There is also the question of trust between systems. An electronic bill of lading must be usable by the carrier, trader, bank and eventual holder, sometimes across several jurisdictions. Industry members of the Digital Container Shipping Association have committed to full electronic bill-of-lading adoption by 2030; that is a goal, not a claim that the job has already been completed. (DCSA commitment)
The UK’s planned Single Trade Window illustrates the remaining gap. Its ambition is to let traders submit information once for border processes, but development for the 2025–26 financial year was paused. We should not describe a single, seamless UK border interface as something an exporter can already rely on. (HMRC Single Trade Window update)
A more useful destination
For routine, well-understood consignments between established partners, we may be much closer to a process in which people handle only exceptions. Verified product data could flow from the manufacturer’s system into the invoice, declaration, transport booking and payment process. AI could check each stage, explain a hold-up and bring the right person in before a small discrepancy becomes a serious delay.
Other consignments will continue to demand close attention: a new product, an unfamiliar market, a controlled item, a complex package assembled from several suppliers, or a buyer whose instructions change after production begins. The more unusual the transaction, the less sensible it is to let a system proceed simply because all its boxes have been filled.
So I do not expect human-free international trade to be a useful target. I do expect far less repetitive administration, faster correction of errors and a substantial reduction in paper. That could give experienced people more time to judge the matters that really require them.
There is a condition. We must first understand the export process well enough to tell the technology what the facts mean, who is authorised to change them, and when a decision must be referred to a person. We need practitioners from every part of the chain to design and challenge these systems. Otherwise, we risk moving a mistake from desk to desk at electronic speed.
The skills and disciplines of international trade are already an international asset. AI may make that asset more valuable by connecting knowledge that is now scattered across departments, companies and borders. The future export expert may spend less time producing documents, but more time asking whether the documents describe the right goods, the right agreement and the right journey.