Borders Are Becoming a Software Problem

Here is my prediction. Within a few years, most of the paperwork that moves a parcel across a border will be prepared by AI agents working on a seller’s behalf.

The border will not disappear. It will become a question of data, and the sellers who win will not be the biggest. They will be the ones whose data is clean enough for an agent to trust.

If yours is a mess, an agent will not rescue you. It will just fail faster.

Here is why I believe this, and what it already looks like at Ergode.

The Border Is Already Asking for Data

Two changes in the past year show the direction.

On August 29, 2025, the United States ended its $800 duty exemption for low-value packages. U.S. Customs and Border Protection estimates that 1.36 billion packages claimed the exemption in fiscal 2024. That is nearly 4 million a day.

Since July 1, 2026, the EU has charged a flat €3 on small parcels, counted per kind of item and identified by tariff sub-heading. The EU’s own example is one silk blouse and two wool blouses: two kinds of item, so two charges. The material changed the bill. The interim charge is meant to last until a new EU customs data hub takes over, expected in 2028.

Different governments, same question: what exactly is in each parcel?

The answer lives in a code. The Harmonized System, run by the World Customs Organization, has more than 5,000 commodity groups, each with a six-digit code. More than 200 countries and economies use it, covering over 98 percent of the merchandise in international trade. Then each country adds its own digits. The first six are shared and the rest are local. The U.S. goes to ten.

A product title written for a shopper, something like “Cozy Winter Set,” answers almost none of it. The shopper does not need to know the fabric. The border does.

Where I Think This Goes

This part is my view, not a forecast from any source. A question with a precise answer is exactly what AI agents are good at: which code fits this item, which rule applies in this country, what the shopper owes at checkout.

For a long time, crossing a border well meant knowing people: a broker, an agent, someone who had done it before. That favored big companies, because they could afford to hire it. When an AI agent does the lookup, the advantage moves to whoever’s product information is cleanest. A small seller with accurate materials, weights, origins and codes can clear the same border as a giant.

Rules will keep changing. But a rule that has become data can be updated. A rule that lives in a handshake cannot.

People do not disappear from this picture. They handle the odd product that fits no category, and the customer who needs a human. That is where judgment lives.

What This Looks Like at Ergode

We already see this with the other border: language.

Our customer support team answers customer messages in more than 80 languages. Most of the team speaks English. AI helps them read and answer each customer in the customer’s own language, and the team handles what software cannot: the care, the tone and the judgment.

One story stays with me. An older gentleman, Mexican and a Spanish speaker, had ordered from another seller. That seller spoke only Dutch. The two never found a way to talk, so when his order went astray, nobody could help him. The product never arrived, and he was left in the middle of a mess.

Ordering things online did not come easily to him, and the first attempt had gone badly. But the product was not available locally, so he had no other choice but to order it again.

This time he reached us. We did not have a Spanish speaker on the team. We had English speakers and good AI tools, and it did not matter. Our team spoke to him in Spanish, and between them and the AI they made the whole thing feel seamless.

He was extremely happy. After a bad first experience, someone was finally speaking to him in his own language about something he really needed. That is a feeling no tracking number can give.

The product was not what failed him. The language was.

That is what I mean when I say the border is becoming a software problem. The AI does not replace the team. It lets a person who cares reach a person who needs help, across a gap that used to be too wide. Customs is the same story at a larger scale.

What Sellers Should Do Now

Five things, none of them glamorous.

Treat product data as an asset, not a chore. It is what the agents will run on. Material, composition, weight, size, country of origin: capture them once, at the source, and fix them there.

Show the full cost early. A shopper who sees duties and taxes at checkout is not surprised at the door, and surprise is what turns into refunds and bad reviews.

Give someone ownership of the rules. With customers in many countries, a rule is changing somewhere all the time. Someone should be watching, and the answer should land in your systems, not in an inbox.

Answer customers in their own language. When something goes wrong with an order, the customer needs to be understood more than anything else. Make that possible before the problem arrives.

Keep a person for the exceptions. Software handles the routine product well. The odd one, the item that fits no category neatly, still needs judgment.

The Case for Being Optimistic

I do not think borders are going away. If anything, governments are looking at parcels more closely than they used to.

But a border that asks clear questions is easier to live with than one that asks vague ones, and clear questions can be answered by anyone who prepares. That makes the game more open than the old one, where the answers belonged to whoever had the right contacts.

The next great cross-border company may well be a small one, with very clean data and a team that cares, working alongside AI. I would back that company over the biggest one.

Agree or disagree? Tell me where you land on LinkedIn. And if you sell across borders, tell me how clean your product data is today.

Regards,
Rupesh