"I work for Amazon now," fewer sellers = more share, meet me in Tokyo Oct 13-14 - 80/20 of Ecom 8/14/26

garyI was in Shanghai this week visiting family, and I crashed the Amazon Seller Meetup I used to run before I moved to Japan in 2019.

I gave a short talk on AI best practices, then spent 3 hours just talking to sellers — from Australia, China, Europe and the US. Earlier that same day I sat down with my sourcing partner, who has 20 years on the ground in China.

The mood surprised me. Here's the honest read.

"I work for Amazon now"

Most of the sellers I talked to are not accelerating. They're staying put. Coasting.

One of them runs an 8-figure brand in the home cleaning category. He's been selling in the UK and Europe for 4 years and just launched the US in June. He's doing well.

And this is what he told me:

"I'm basically working for Amazon. I give everything to Amazon. They're charging me so much these days."

Then he added, "No wonder Jeff Bezos is so rich."

He's also getting pitched Amazon's Maverick program in the UK. Another program. Another fee.

My read: it's one more way to pull money out of sellers, and the benefits look dubious to me.

TAKEAWAY: If it feels like your margin is quietly moving from your P&L over to Amazon's, you're not imagining it. Fees are going to keep climbing. The only real question is what you do with the part you still control.

So that's the bad news. Here's what almost nobody said out loud.

Three things are quietly breaking in favor of the sellers who are still standing.

1. On tariffs, China is competitive again

This one isn't my read. It's from my sourcing partner in China, who has been doing this for 20 years.

I asked her how sourcing is going right now. Her answer: China's tariff position should be competitive again — not because China's rate is low, but because everybody else's went up too.

India faces high tariffs going into the US. Southeast Asia is high — Vietnam, Malaysia, all of them. Mexico and Canada into the US are high as well.

So when you compare China against the alternatives instead of against 2019, the math looks different than most sellers assume.

She also flagged a change in how buyers are ordering. Start quantities are smaller than they used to be. And buyers are skipping steps — they used to review samples and work through the process one stage at a time, and now a lot of them just want to place the order fast.

TAKEAWAY: Stop pricing your 2027 costs off the worst tariff headline you read back in April. Pull your real landed costs today, then compare China against your actual alternatives, not against the old normal.

2. Fewer sellers means more share for you

A lot of people quit over the past 2 years. Both conversations I had that day landed on the same point, and neither person was in the room with the other.

The first is a seller who runs a large fitness brand in the US and Europe. I asked how business was going. He said it's stable, then put it plainly:

"I think the market's gone down... so I've got more market share."

The market went down. His share went up. That's the trade.

My sourcing partner said the same thing from the factory side. Chinese competitors moving in hit a lot of sellers hard, and plenty of them pulled out and stopped selling. Her read on who's left: the ones who've been at it a long time, and they're holding a pretty big share.

So the field thinned out faster than the market did. Here's how I said it to my partner at the end of the night: there's less sellers, but the ones that survive, they're doing okay.

TAKEAWAY: A thinning field only helps you if you're pushing. If you coast into it, you thin out with everybody else.

3. AI is where the edge is now

This is the big one, and it's why I'm optimistic despite everything above.

Repeatable workflows beat clever prompts.

One attendee runs a trading platform. He has no coding background at all.

He described his loop: he feeds an agent the API documentation, the agent writes its own skill, he tests it, then polishes it. He ends up with a skill about 500 lines long.

Then he automates a weekly check to see if the documentation changed, and updates the skill when it does. In his words: "So I make a loop."

Why bother? Because without the skill, as he put it: "If I tell the agent, 'connect and get the data every day,' he has to learn every day how to do it, and he makes a lot of mistakes."

Small teams are running things that used to need an ERP.

My business partner came down from Beijing for this. He spent 8 years at Google and now builds AI agents for companies.

He walked through one client — an ecommerce brand doing about $12 million ARR with a couple hundred SKUs, suppliers and fulfillment centers all over the world. The whole thing was being tracked in Google Sheets.

He hooked an agent into the team's Discord. Now it checks everything every hour. The founder can just ask, "there was a shipment due in Canada, is it there? Can I launch the campaign?" And it answers.

His line stuck with me:

"In the past, you'd have to get some kind of ERP system... and probably pay Accenture or some big tech consulting company all this money to put it all together. And now we can cobble it together with AI agents."

Your job is to be the orchestrator, not the worker.

Here's how I explained it to my sourcing partner over dinner. Think about a law firm. The lawyer has the highest salary. You're not going to have the lawyer go dig through the inbox.

You have an assistant pull the information. Another one check the numbers. Another one do the lookup. The lawyer runs all of it.

That's how I use AI now. Cheap fast models for the grunt work, the expensive model only for the thinking. I'll run 4 sub-agents at once on different pieces. That's token economization, and it's the difference between a $200 bill and a $5 one.

Her response was the sharpest thing anybody said to me all day: with AI, your own ability has to go up.

Now the guardrails, because this is where it goes wrong.

One of the students in my cohort gave an AI agent access to his ad campaigns. The agent negated a keyword. That single move wiped out 8 years of history he could not get back.

My partner's rule on security is the right one:

"Start with zero trust, and then as you build stuff up, slowly trust it more and more."

And the failure most people don't see coming, again from my sourcing partner: you told it to do the work, but if you never go look at the result, it's the same as not doing it at all.

I'm guilty of this. I'll spin up a bunch of agents, and half the time I don't go back and check what they produced. Sometimes they stall out and just sit there.

Read access first. The agent recommends. You approve anything you can't undo. And you actually read the output.

TAKEAWAY: AI isn't about doing the same work faster. It's about running a leaner team — inventory, PPC, listings — without adding headcount. Build the loop, keep your hand on the trigger for anything irreversible, and check the work.

A quick word on listings

Sellers in the room still had questions about using AI for listings and images. Honestly, that problem is mostly handled now.

One attendee generates his listing images in ChatGPT. He says it takes him "probably four hours per image" to get the one he wants — "but the perfect image does come out."

The real edge isn't making the image anymore. It's how fast you find out which image wins.

TAKEAWAY: One-shot generation is not the advantage. Learning faster is. If you're not A/B testing your creative, the AI just helps you produce losing images more efficiently.

The part you can't get on Zoom

Here's the thing. I could have given that presentation online. Any of us could.

What I could not have done online was the 3 hours after it.

I sat with a PPE brand owner whose family owns the factory. Work gloves, cooling towels, cooling vests. He launched his US business in June and sells wholesale to distributors and dealers.

He's got factory-direct pricing and genuinely doesn't know whether to go on Amazon himself or keep letting his dealers do it. You don't get that level of detail about someone's actual problem over a webinar chat.

I compared notes with an ex-Google data scientist on what AI agents can really run inside a business today, versus what people claim on LinkedIn.

And I met a seller doing $30 million a year who started on eBay in Australia back in 2004. Australia. A market I would have told you was too small to build anything real on. He built a $30 million brand out of it.

I would not have believed that if he hadn't sat there and told me himself.

Late in the night I told my partner what I think is coming:

By next year there's a divide between people using AI and not using AI. If you're not using AI, you're going to be on the wrong side of the divide.

TAKEAWAY: The information is free and it's everywhere. What isn't free is the person across the table who's 2 years ahead of you and willing to tell you what actually happened.

7 Figure Seller Japan Mastermind — Tokyo, October 13-14

That night in Shanghai is exactly why I'm running this.

Two days, in person, in Tokyo. For Amazon and ecommerce sellers who want to build AI agents and real workflows into their business.

Here's what you'll walk out with:

Early bird tickets open soon. The waitlist gets first access and the lowest price.

Join the waitlist here: https://subscribe-forms.beehiiv.com/db56d73b-33c3-4d32-a23c-643a025341e6

In case you missed it:

See you next week,

Gary