Service
AI-written product descriptions that a person approves
Writing a description for every product is the task nobody has time for, and unchecked AI text quietly makes mistakes. I set up generation with model failover and quotas, an automatic quality check, and a one-tap approval step in Telegram.
Discuss a project →What you get
- A description for every model, in your tone, from your product data
- Automatic check plus /approve, /reject or /regen in a Telegram bot
- Retries, model failover and quota handling so the queue never stalls
- Approved texts flow straight into the store and marketplace feeds
Where I have already done this
Client project · e-commercefish-syncStock from a shop's warehouse program turns into product cards on its website, Rozetka and Facebook by itself.53 models found in 85 SKU81 variants linked to their parentsAutomation · in daily useCommerce automationA manager sends a product video to a Telegram bot, and the system does everything else.6 jobs from one upload85 commits, March–September 2026
Questions
- How many descriptions a day?
- It depends on the AI quota you choose. In one shop it runs at about 22 a day and the queue simply continues the next day.
- Can AI text go live without a check?
- Not in my setups: every text passes an automatic check and a person approves it first.