A Foreign Boss Got Eliminated by a Machine

A while ago, a friend who exports industrial equipment vented to me.

He had spent three years rebuilding his company website, product brochures, and customer case studies — all in English. Traffic went up. Inquiries dried up.

He thought it was a pricing problem, so he went and checked his competitors. Their prices were higher than his.

He thought it was a copywriting problem, so he hired a native speaker to rewrite everything. Still nothing.

Finally, he asked an American friend to open ChatGPT in front of him and type in: "I need a supplier in Asia that makes XX equipment."

His name wasn't in the answer.

Three entire years of English-language materials he was so proud of — and in the eyes of that machine, they might as well not exist.

When I heard this, my heart sank.

Not because I felt sorry for him. Because I immediately ran the check on myself: the things I've built — in the eyes of that machine, do they still exist?

---

The Buyer You Think You're Talking to Isn't the First Audience Anymore

The way we used to understand "cross-border" was simple.

I'm a foreign supplier. You're a buyer in the destination market. Whether you'll buy from me comes down to three things: is my product any good, is my price right, and do you trust me.

That logic ran for fifty years, and it never changed.

But now — hold on a second before you make that call.

Because something has slipped in between you and the actual human buyer.

What is that something?

It's ChatGPT. It's Claude. It's Perplexity. It's the AI procurement agents embedded inside large American companies' purchasing processes. It's Google's generative search box that just hands you an answer.

Gartner has a prediction: by 2028, 90% of B2B purchasing will run on AI agents.

Think about what that means.

It means that before the procurement manager at that American company ever sees your quote, a machine has already read through — on his behalf — your website, your product brochures, your case studies, your press coverage.

And after reading, that machine does a round of filtering on behalf of that person.

The filter produces two outcomes.

One: your name shows up in the machine's answer. The procurement manager sees it — oh, there's a company like this, let me click in and take a look.

The other: your name doesn't show up at all. The procurement manager never even gets the chance to know you exist.

You see? You're not competing with your peers anymore. You're competing with whether that machine is willing to "say you out loud."

---

It Can Translate, but It Can't "Do Business"

At this point you might be thinking: fine — I'll get my English right, localize the website, and I should be okay.

Let me tell you something.

There's a founder who exports medical devices. His English website is impeccable — you can't find a grammatical flaw. American buyers read it and think, "Reads professional, but it doesn't feel like a real American brand."

Why?

Because a machine can translate words, but it can't translate business.

You write, "Our products meet FDA standards." The machine translates that into English. The grammar is perfect.

But when an American buyer reads that sentence, what he cares about isn't whether your grammar is correct. What he cares about is: in my purchasing process, what is this thing called, which category does it fall into, which approval pathway does it go down, who calls the shots, who carries the risk if something goes wrong.

The machine doesn't know any of that.

It has language, but it has no business language.

It knows how to spell the three letters "FDA," but it doesn't know that inside a US hospital's purchasing process, the "FDA clearance list" and the "reimbursement code" are two completely different systems.

It knows "global shipping," but it doesn't know that across the US, the EU, the UAE, and India, the compliance boundaries, purchasing habits, and trust signals are four completely different sets of rules.

So you keep running into this awkward pattern. The website is grammatically flawless, but after the American buyer reads it, he walks away thinking: this guy doesn't understand my market.

When a local company makes a mistake, the buyer thinks, "Oh, minor issue."

When a foreign company makes the same mistake, the buyer thinks, "I'm not comfortable with this."

Call it "innocent until proven guilty" versus "guilty even when innocent." That's the invisible ceiling of cross-border.

---

One More Thing You Probably Haven't Realized

Let's pull the camera back even further.

The hardest part of cross-border business was never marketing. It was compliance.

The EU AI Act takes effect on August 2, 2026. The fine? €35 million, or 7% of global revenue, whichever is higher.

Here's the key: this law governs AI systems operating inside the EU, regardless of where your company is registered.

You're in China. Your AI model runs on a user's device inside the EU. You're now in scope.

Think about how scary that is.

There's another regulation called DORA (Digital Operational Resilience Act). It covers any third party providing IT services to EU banks, insurance companies, payment firms, or securities firms. No matter where you are.

India, China, Saudi Arabia, Indonesia — these markets all now have data sovereignty requirements. Where your AI runs, where the data is stored, whether inference can cross a border — these are market-entry decisions in their own right.

Let me do the math for you.

In the old days of cross-border, you had to nail: product, pricing, channel, brand, compliance.

Five things. Already a handful.

Now, each of those five things has an AI evaluation machine in front of it.

• Your product page — first screened by AI
• Your quote — first screened by AI
• Your channel — AI agents are picking
• Your brand — generative search decides whether you surface
• Your compliance — the law regulates the operation of AI, not your place of registration

This isn't "one more thing to do." The underlying structure of cross-border business is being rewritten.

---

Ten Concrete Changes I'm Seeing

I took all the changes I've seen, read about, or been asked about by clients over the past year and organized them. About ten points.

I won't expand on each — just feel how wide the surface area is.

One, procurement. Large American companies now use AI agents to shortlist suppliers. Your website, your case studies, your follow-up speed — they have to pass the machine first.

Two, being found. When an American buyer asks ChatGPT, "Asian suppliers who make XX," whether you surface depends on whether your structured data is done right. This is a technical craft, not something you can buy with ad spend.

Three, regulation. That EU AI Act — broader in scope than you think.

Four, financial compliance. Providing IT services to EU banks? DORA's coming for you.

Five, compliance cross-mapping. A single company has to pass SOC 2, ISO 27001, GDPR, FedRAMP, CMMC, MDR, the AI Act, and DORA. That's six sets of rules layered onto one company.

Six, data sovereignty. Where your AI runs is no longer a technical decision — it's a market decision.

Seven, M&A due diligence. In cross-border acquisitions, screening targets, reading documents, assessing compliance, planning integration — all done by AI agents.

Eight, sales tools. An AI sales assistant that works in your home market, translated into a destination market, gives you the feeling of "very confident but answering the wrong question."

Nine, register. Machine-translated English — grammatically correct, commercially off-register.

Ten, trust. The cheaper AI-generated content gets, the less cross-border buyers trust it. Local companies enjoy a "trust dividend"; foreign companies don't even get that.

Each of these ten, on its own, doesn't look like a big deal.

But put them together and you notice one thing.

The "bar" for the entire cross-border business has been quietly raised.

And the way it was raised is insidious.

Not tariffs. Not quotas. Not any written policy.

It's that machine — AI — quietly slipping a layer between you and the buyer, a layer it calls the shots on.

---

So What Do You Do?

The first judgment I can give you: this is not a problem localization can solve.

A lot of people are still running the old playbook — get the English right, get the SEO right, throw some budget at Google ads, call it done.

That playbook worked well, before 2020.

But in 2026, it's going out with the tide.

The second judgment: this is not a problem legal review can solve, either.

EU AI Act, DORA, data sovereignty — reading those clauses won't help. Legal can only tell you "what not to do." But "what to do"?

Someone has to rebuild — from the perspective of the business — your website, your quotes, your case studies, your follow-up process.

So that machine can read you.

So that the actual human buyer, after the machine hands him an answer, wants to click in and take a look at you.

This is painstaking work. Every corridor is different.

The playbook from DACH (Germany, Austria, Switzerland) into the US is different from the playbook from Hong Kong into the US.

The playbook from the UAE into the US is different again from both of those.

There is no universal template.

---

Finally

Back to that friend from the beginning.

What did he do after that?

He didn't go study SEO. He didn't go build backlinks.

He did one very simple thing.

He found three American customers, paid each of them two hundred dollars, and had them open ChatGPT right in front of him and ask the way they'd actually ask when sourcing.

Then he rebuilt his website, product brochures, and case studies around the structure of those three real questions.

Three months later, he saw his own company's name inside ChatGPT's answer.

He told me he stared at the screen for five minutes without saying a word.

When I heard that, I went quiet for a while too.

Not because I was moved.

Because one thing occurred to me.

In the old days of doing business, the goal was to make "humans" remember you.

Now, the goal is to make "the machine" willing to mention you.

Those two logics are two completely different games.

And a lot of people are still playing the new game with the old game's rules.

This is the thing that's made me most anxious this year.

I hope it's also where your anxiety starts.