This is a small product-intake workflow, not a promise that AI can run an ecommerce store by itself. The useful part is narrower: a product photo becomes a reviewable WooCommerce product page without a person opening WordPress, uploading the image, writing the first description, and sending the review link by hand.
The workflow handles the repetitive first pass. A person can still check the title, price, category, and claims before using it in a real store.
The generated product page
The result is a real WooCommerce product page in a temporary test store. For this demo, the workflow creates the product with the uploaded image, a short title, a short description, and a demo price. The screenshot still shows the store's default category because category mapping was not part of this prototype.
The loop closes in the same chat
The operator sends a product photo into a private admin chat. After n8n creates the WooCommerce listing, the bot replies with the product URL so the person can inspect it immediately. Telegram is just the internal intake tool here, not a public FloxoLab contact channel.
The workflow map
The n8n canvas stays simple: receive the photo, fetch the Telegram file, upload it to WordPress Media, analyze the image, create the WooCommerce product, and send a message back to the chat.
The AI and tool setup
The image analysis step describes the visible product details. The AI agent then uses that analysis to prepare the product fields and call the WooCommerce tool. In this demo, the price is fixed and the text fields are intentionally short.
Why this works for WooCommerce teams
This approach works well for WooCommerce because the store already has the pieces the workflow needs: media uploads, product creation, product status, categories, and product URLs. n8n can talk to WooCommerce through native nodes, so a small team can test a product-intake workflow without building a custom admin panel first.
For this demo I used a temporary WordPress and WooCommerce test store. That made it easy to validate the workflow quickly before thinking about a real client store.
What the AI is allowed to do
Listing text should come from the image analysis, not invented product claims.
The title and description stay compact so a person can review them quickly.
This prototype uses a fixed demo price. A real store should pull price, stock, SKU, and category from a trusted source.
No sizes, materials, shipping promises, inventory, or brand claims should be added unless they are provided by the store.
Good fits
Small shops that still create product pages manually from product photos or supplier images.
Teams onboarding new product lines that need a faster first draft before final merchandising review.
Teams that repeatedly turn image inputs into product records, review queues, or marketplace drafts.
The intake chat could be Telegram, Messenger, or Viber depending on the team's daily tools. The important part is private operator intake, not a public support chatbot.
What I would improve next
Create products as draft by default, then publish after approval.
Add a review step so the operator can approve, edit, or reject a generated product before it goes live.
Map SKU, category, price, stock, size, color, and supplier source from trusted tables instead of asking AI to infer them.
Search existing products before creating a new page, especially when supplier photos repeat across batches.
If a message has no photo, the file cannot be downloaded, or the image analysis is weak, send a clear error back to the chat.
Bottom line
This workflow is useful because it turns a product photo into a reviewable WooCommerce listing without building a custom admin interface. A person should still check the title, category, price, and product claims before using it in a real store, but the blank-page work is already done: image upload, first draft, product page, and review link are prepared automatically.
Even with human review, the workflow can save the repetitive first pass: uploading the image, opening WooCommerce, creating the product, writing a short description, and sending the page link to the person responsible for approval.