Back in my early eCommerce days, product tagging meant endless Excel sheets, manual filters, and asking around: “Is this pasta Italian or just made in Italy?” We spent weeks aligning on logic — and still missed the nuances.
Fast forward to today. What used to take teams and time, we can now build in days — powered by AI, a simple Python script, and an API key. And here’s the best part: you don’t need to be a coder to do this. You just need curiosity, a structured file, and a willingness to let AI assist — not take over.
The goal
To classify 8,500+ products with tags that go beyond the obvious — adding depth for merchandising, personalization and customer discovery.
How we did it — without letting data leave our system
- A master Excel file with product name, category, ingredients, etc.
- A Python script (assisted by GPT) running on my machine
- Calls made securely via the OpenAI API
- No third-party platform, no copy-paste — just controlled, local automation
We started with rule-based tags (Cuisine, Dietary, Usage, Festive). Then we added an AI engine for contextual and cultural awareness, layered tags like “Premium + Festive + Japanese,” and understanding the story behind the SKU.
What changed
Before: tags were flat, repetitive and rule-bound. Now: they’re dynamic, relevant, and smarter with every prompt. We even built in async processing (10 at a time), auto-save every 10 rows, and logging with error recovery.
What this unlocks
- Smarter filters for merchandising
- Personalized experiences
- Multilingual tag expansion
- Campaign and assortment planning
- A reusable engine for future datasets
Looking back, this would’ve needed a team of merchandisers and analysts. Today, it’s AI + one person + one machine. And the most empowering part? The data never left the system. The control stayed where it should — on our side.