Building an AI Product Identifier for a Luxury Reseller

Sometimes the best AI projects come from the most unexpected places.

A family friend runs iLuxury, a luxury goods resale business. Their workflow was entirely manual: receive consignment items, photograph each one with a sticky note showing client info and price, then manually type everything into a spreadsheet — brand name, model, material, description, customer info — for inventory tracking and tax export documentation.

For a business processing dozens of items daily, this was slow, error-prone, and tedious. The perfect candidate for AI automation.

The Problem

Every consignment item arrives with a photo showing the item alongside a handwritten sticky note with:

• Client name and ID
• Price (in HKD)
• Date received

The reseller needs to:

1. Identify the brand (Chanel, Louis Vuitton, Hermès, etc.)
2. Identify the specific model
3. Describe the material and condition
4. Extract the client info from the sticky note
5. Log everything into a Google Sheet for inventory and tax export

Doing this manually for each item takes 3–5 minutes. Multiply by 50+ items per week, and you've got a full-time data entry job.

Version 1: The n8n Chatbot

The first version was an n8n workflow with a chat interface.

How it works:

1. User sends a photo via the chat interface (one image per message)
2. An AI Agent powered by OpenAI's vision model analyzes the image
3. The agent extracts: brand, model, material, description, client info, and price from both the item and the sticky note
4. Results are automatically appended to a Google Sheet

Simple, effective, and it works. But it had limitations — one image at a time, chat-based (not ideal for batch processing), and required n8n to be running.

Version 2: The Lovable Web App

For a more polished experience, I built LuxeVintage Image Analyzer using Lovable — a full web application with drag-and-drop batch uploads.

Key features:

• Batch upload: Drag and drop up to 15 images at once
• AI-powered analysis: Each image is analyzed for brand, model, material, color, and price
• Editable results: Click any cell to correct before exporting
• Google Sheets export: Create a new tab in any Google Sheet and export the full table
• CSV/TSV export: Copy directly for spreadsheet paste

The table extracts: Date, Product ID, Customer Name, Price (HKD), and Brand Item Description. It correctly identified a "Chanel J12 Watch" at HKD 5,000 and a handbag at HKD 2,280 — all from photos of items with sticky notes.

Why Two Versions?

Each version serves a different use case:

The n8n bot is great for the shop floor — snap a photo, get instant identification. The web app is better for end-of-day batch processing when they need to log 20–30 items at once.

Real-World Impact

What used to take 3–5 minutes per item now takes seconds. For a batch of 15 items, the total processing time dropped from ~60 minutes of manual data entry to about 2 minutes of drag-drop-export.

More importantly, the accuracy improved. AI vision models are remarkably good at identifying luxury brands and models — often catching details that manual entry would miss or abbreviate.

Lessons Learned

• Start with the simplest version: The n8n bot took an afternoon to build and validated the concept before investing in a web app
• AI vision is production-ready for product identification: OpenAI's vision models handle luxury goods identification with high accuracy
• Lovable is ideal for data-tool UIs: The drag-and-drop interface, editable tables, and Google Sheets integration came together quickly
• Build for the user's workflow: The reseller doesn't care about the AI — they care about faster data entry and accurate export docs
• Two tools > one perfect tool: Rather than trying to build one app that does everything, two focused tools cover more use cases

Bonus: The iLuxury Website

Beyond the back-office tools, I also built the customer-facing website for iLuxury using Lovable — a full e-commerce storefront with product catalog, designer filtering, customer testimonials, and a consignment submission flow.

Try it live:

The website, the n8n chatbot, and the LuxeVintage analyzer form a complete technology stack for the business — from customer-facing sales to back-office inventory processing to tax export documentation. All built with AI-assisted vibe coding.