"Wait, seriously?" That was the first thing I said when I received a 30,000 JPY sales notification on the very first day.
From there, it reached a total of 200,000 JPY in 7 days. The only cost incurred was 35,000 JPY in advertising. I worked alone. No designers, no engineers. I never even touched a terminal (the black command-line screen).
I received a great response to the following post recently.
https://x.com/columbus_ceo/status/2084470763181199612
From that point, it hit 200,000 JPY in 7 days. It continues to sell today. All I used were Claude, ChatGPT, and Shopify.
In this article, I have organized the entire procedure I practiced and will explain it for Japanese business operators.
Step 1: Throw All Market Research to Claude
What I sold was an alcohol-related overseas product that people questioned, "Will this really sell?" It was a product with a track record in crowdfunding two years ago. However, that data is two years old. Competitors and categories have increased since then.
So, I gave everything to Claude.
Overseas Amazon reviews. Japanese crowdfunding reviews. Competitor pages. The instructions were simple.
"I'm giving you all the reviews, so please evaluate this strictly compared to the competition."
When you say "strictly," it really gives harsh feedback. An AI that only praises you is useless.
The conclusion was: "It's a two-year-old product, but there are still elements that can win." It even identified which features would resonate with the 20,000 JPY price range target audience.
Step 2: Drawing the Line Between Site Creation and AI
This is the crossroads for non-engineers. I am not a developer.
I only used the MCP integration for the desktop version of Claude (a system that connects AI to external tools). I even asked Claude how to do it.
What I achieved with MCP: Automatic login to Shopify. Product registration. Adjustment of inventory and prices.
Where MCP fell short: Editing theme sections. Pouring in images and text. I fixed these parts manually.
However, Shopify's section feature is no-code. Insert an image. Insert text. Insert a video. That's it. Since the structure and copy were generated by AI, all that remained was the task of embedding them.
Step 3: Advertising and Creatives
Designer costs were zero.
I received realistic photos and videos from the overseas manufacturer for free and used ChatGPT to generate the missing parts.
The true value of Claude here lies elsewhere. It's having it write prompts (instructional text) to make other AIs create strong creatives. Claude is overwhelmingly strong at competitor comparison, copy, and ad text. I think Code x can do it too.
It even provides operator-like advice such as daily budgets for Meta ads, learning period estimates, and "don't chase profit too hard until learning accumulates."
80% AI and 20% Human
To be honest: AI creates 80% of the basic design. The remaining 20% can only be filled with human words.
In another home gadget project, the copy that lowered the CPA (Cost Per Acquisition) didn't come from AI. It came from customer interviews.
"Reclaiming time with my wife after she and the kids have gone to sleep."
It's faster to ask customers for a line like this than to ask AI. This is the human's job.
Conclusion: Why Did It Sell 200,000 JPY in 7 Days?
Because of AI. That's half correct.
The other half is "because I was able to run a pre-order model without inventory for a product with overseas appeal."
Negotiate without MOQ (Minimum Order Quantity) or minimum guarantees. Have existing inventory sent. Close and place orders every month. If it's a fast-shipping manufacturer, clearly state "delivered in 2 weeks."
There were zero cancellations in 7 days.
Run it as a pre-order, and stock up once it starts selling. What you should watch is cash flow.
By the way, there are cases where other brands are making 37 million JPY a month, so please check them out.
https://x.com/columbus_ceo/status/2019678626908938502
Noji's Perspective: What You Should Really Learn from This Procedure
There are three points to this article.
1. AI replaces marketing. It does not replace products and contracts.
Watching 120,000 JPY in sales on the night of a fireworks festival, I didn't think "wow." I thought that without a strong product, AI will eliminate the job of EC management. AI can handle sales, advertising, and analysis. Therefore, only those who can bring in products will remain.
2. Negotiation is free. OEM costs tens of millions.
Exhibitions, LinkedIn, overseas crowdfunding. Look there first. You can get a contract in a week if you're fast, or a few months if you persist. Don't be fooled by the words "start small as a side hustle." What you should start small with is inventory, not product quality.
3. Beginners just need to connect Claude and MCP.
I haven't mastered AI. People who can use it can do much more. There is an advanced route like Tetsuo Funahara using Claude Code and CLI. But for non-engineers, an MCP connection is more than enough to reach this level.
This applies directly to small and medium-sized Japanese businesses as well. In our company's view, 80% of the victory is decided by "what to sell" before starting.
The following article is also helpful.
https://x.com/columbus_ceo/status/2083050063358746785
Our Initiatives
- Negotiation agency for overseas manufacturers
- Cross-border sales to build overseas sales teams
- In-house support
We plan to grow this site to a scale of 5 million JPY monthly turnover and 2 to 2.5 million JPY profit. I will share it again as an ongoing experiment.
https://x.com/columbus_ceo/status/2068647931918569843
https://x.com/columbus_ceo/status/2016113941701919015
Information on Benefits
If you send EC to our official LINE, we will distribute the following:
- A 36-page white paper summarizing the methods used in-house this time: "Manual for Creating a 100,000 JPY/Day EC Driven by AI"
- 2 audio commentary radios: "How to Sell 200,000 JPY Every Day Driven by AI ①②"
If you are interested in an exclusive overseas license FC, please send "Request for Interview." I will send a scheduling link.
Official LINE: https://lin.ee/YRMYbWJ
In an era where AI takes over 80%, what will you sell with the remaining 20%?
Thank you for reading.
