When AI Grows Into Every Cross-Border Order

A while ago, I scrolled past a set of numbers and paused.

In 2023, China's cross-border e-commerce market reached 16.85 trillion yuan (RMB). Walk back four years — 10.5 trillion in 2019, 12.5 trillion in 2020, 14.2 trillion in 2021, 15.7 trillion in 2022.

One new step every year, almost without a pause.

My first reaction: this is not ordinary. For an industry to climb by the trillions several years in a row, the force behind it can't be one or two clever sellers pushing harder. It has to be the entire foundation being re-laid by something.

That something is AI.

Let's Start With an Era

Cross-border e-commerce is actually an old business. But you have to understand — it has never stood still.

The earliest cross-border trade belonged to the print-media era — magazines, trade shows, mailing a thick product catalog overseas. An order placed might take weeks to arrive.

Then came the PC, and with it the PC e-commerce era. Then the phone became an extension of ourselves no one could live without, and we entered the mobile e-commerce era.

And inside mobile commerce, the playbook kept evolving. First came search-based e-commerce — you type a keyword, I give you the product. Then social commerce — friends recommending, communities building buzz around products. Then interest-driven e-commerce — you scroll through videos and, before you know it, you've placed an order.

But starting in 2024, pay attention: a new era quietly arrived.

The era of AI-driven e-commerce.

What does "AI-driven" actually look like? Let me paint a few pictures you may have already seen but not paid attention to.

A store's customer service replies to your English message at two in the morning in seconds — and the tone, the politeness, the professionalism are better than the human agent who works the day shift. You ask for a discount, and it can even decide, based on your order history, whether you deserve one.

An operations manager used to spend half the day manually adjusting prices, rewriting titles, checking inventory. Now an AI assistant handles all of it. It's faster than a person, and it never gets sleepy.

A designer used to max out at twenty product images a week. Pair them with an AI design assistant, and it's two hundred a day.

Do you see the pattern?

These roles used to be filled by people. How well they were filled depended on how capable the people you hired were. Now AI has raised the "ceiling" of these roles — and raised the "floor" too. At the same time, it has wiped out the value that came purely from brute speed and manual grunt work.

The Real Difficulty Isn't the Tools

At this point, someone might say: easy fix — I'll just buy the tools, right?

Not that simple.

Let me give it to you straight: in this era, tools are not what's scarce. What's scarce is people who know how to use them.

Cross-border e-commerce was never a standardized industry to begin with. It straddles international trade, e-commerce operations, foreign-language communication, cross-border logistics, data insight — that many disciplines at once. A competent operator used to need to understand products, platform rules, English, and cross-cultural communication.

Now there's one more thing — knowing how to weave AI into your own workflow.

Here's an example. Same task: overseas social-media marketing.

Operator A uses the playbook from the last five years: open accounts, post, buy keywords, watch conversions.

Operator B uses the new playbook: let AI monitor discussions across the entire web, telling you in real time what overseas buyers are talking about this week, which keyword's search volume just spiked, which video has the highest completion rate, which user personas are worth targeting.

Whose ad budget goes further?

You already know the answer.

But here's the question — where does Operator B learn this skill?

Schools barely teach it. The SKU management, SEO optimization, and cross-border payment flows in the textbooks will get you through the door, but when you hit the AI layer, most curricula are blank.

This is the biggest mismatch in the industry right now: the tools are sprinting while the talent is crawling.

The Overlooked "Other Side"

I could talk all afternoon about the benefits of AI. But everything has a flip side.

Every technological revolution liberates one group of people while eliminating another. The steam engine did it. Electricity did it. The internet did it. AI will be no exception.

But AI also brings a few troubles all its own.

First, algorithms carry bias. You might assume they're impartial, emotionless — but they learn from past data. Whatever bias and stereotyping is buried in that data is what gets output. Hand a biased AI the job of pricing, product selection, or user profiling, and mistakes are only a matter of time.

Second, data crosses customs — but the privacy wall isn't so easy to cross. Your user data, transaction data, behavioral data shuttles back and forth between countries, and whose hands it lands in and what they do with it is often a gray area.

Third, who owns what AI creates? If AI painted the product image and AI wrote the copy, whose copyright is it? If someone else uses the same AI and the same prompts to generate nearly identical output, is that plagiarism? The whole industry is still arguing about this.

Fourth, the risk of monopoly. When everyone relies on the same one or two foundation models, the "bedrock" of the industry is effectively held in the hands of a few companies. The day they change their pricing, change their rules, or shut down their API, you have no leverage at all.

See — none of this is alarmism.

Technology itself is neutral. But the moment technology scales into an industry, it surfaces every tacit understanding that industry used to run on and forces each one to be re-examined.

Cross-border e-commerce is being surfaced and re-examined right now.

So, What Do We Do?

I've talked about this with quite a few friends in cross-border. The line I hear most is: don't panic, and don't lie flat （a Chinese internet term for disengaging and giving up ambition）.

Panic is seeing AI do your job ten times better and deciding you have no way forward. That's wrong. AI is strong at "repetition" and "speed." But judgment, taste, and genuine understanding of the customer — on these, AI is still far behind. A real person who knows how to use AI will always beat someone who only knows how to use AI.

Lying flat is deciding this has nothing to do with you and carrying on the old way. That's also wrong. The pace of change in this era is faster than you imagine. If you stand still today, tomorrow you won't be eliminated — you'll be bypassed. Customers bypass you, platforms bypass you, competitors bypass you.

My own read: over the next three to five years, cross-border e-commerce will reshuffle once. What gets shuffled out won't be big sellers or small sellers — it'll be the people who haven't made AI an extension of their own hands and mind.

How, specifically? I don't have a standard answer — anyone who claims to is lying to you. But there are a few things I think everyone in this industry should seriously consider.

First, start using it. Don't just read the news and the articles (mine included). Download, register, get your hands on it. Wire AI into your daily work. Even if it makes a mess at first, you need to figure out exactly how it's making that mess.

Second, learn proactively. The basic principles of AI, data thinking, how to write a prompt — these are no longer a programmer's monopoly. They're common sense every practitioner will need. Schools teach slowly; you have to move fast.

Third, and the hardest — get clear on what value you're actually creating. The thing you do — at its core, which layer of value are you creating for people? Can AI replace that layer? If it can, where do you plan to move? This question is worth asking yourself once a year.

That day, after I finished staring at that 16.85 trillion number, I stared at the ceiling for a long time.

An industry climbed from ten trillion to nearly seventeen trillion in four years. The next four — who knows how fast it'll run. But the way it runs will definitely not be the same.

Change keeps on changing. Only the people who change with it still have a game to play.