What happens when you hand customs professionals a stack of paperwork and a 15-minute clock
Many people in senior customs roles, and many who build solutions that plug into the customs process, don't always have a feel for the work on the floor. They see that people perform a lot of manual work under stress, that clients push to work faster and cheaper, and that businesses run the risk of fines caused by human error. But seeing it and feeling it are two different things.
So we built a hands-on experience designed to make those pain points tangible, and then to show how AI customs classification has moved well beyond the hype, delivering real, applicable value.
We opened with a parallel between technological evolution in aviation and in customs, and a rhetorical question: would you board a plane flown entirely by AI and robots, with no human in the cockpit? It set the tone for the core idea: AI as a co-pilot, not a replacement.
Then the hands-on part. Each participant received a realistic document set: an email, an invoice (8 lines), a packing list (9 lines) and a bill of lading. Alongside it, a subset of the customs tariff, with goods codes to 10-digit precision and their descriptions.
The task: count together all goods lines sharing the same tariff code and the same country of origin, and record the results in a table. Where goods were shipped across multiple containers, those lines had to be split per container number. The tools: pens, highlighters, and their own smartphone as a calculator. The clock: 15 minutes. The winner: whoever got the most lines correct.
Participants came from varied backgrounds: customs managers, compliance managers, and software vendors active in customs management, including both salespeople and people who support projects at shared clients in practice.
The first reaction was always the same: open the pack, explore the documents. Some dived straight into filling out the answer sheet; others first worked the puzzle of which lines belonged together across the different documents. The highlighters came out for everyone, and so did the smartphone calculator. What was striking across the room was the drive to get it right, and the sighing under the pressure of the task and the clock. Spontaneous collaborations sprang up, on the logic that two pairs of hands work faster than one.
When the 15 minutes were up, the results were remarkably different from person to person. Some had formed 2 lines, others more than 10. The correct answer was 7 declaration lines.
Then we ran the demo, and showed the same case completed 100% correctly in under a minute. A few reactions stuck:
"Wow, the business impact of this is enormous."
"Your solution looks incredibly user-friendly."
"This is a game-changer for compliance."
The insight participants carried home: embedding AI is no longer really a choice. Those who still aren't engaging with it will struggle to keep pace with competitors who are.
But the hands-on element combined with the demo made something else clear too. AI doesn't have to be an opaque chatbot that introduces new compliance risks. Done right, it's an enabler. It gives people more room for the compliance work that matters, and it brings calm to teams by flattening the huge variability between client jobs. The gap between large and small files shrinks from hours to minutes.
That's the difference between basic extraction and what Customaite does: auditable, compliant extraction that keeps the declarant in control.
Did this spark something? Curious how customs classification software like Customaite compares to manual work? We'd be glad to talk, no strings attached. And if you'd like to run a workshop like this to make your own organisation feel what customs work really involves, we're happy to help you set it up.