From license plate to the right part, without portal-hopping

The context

An auto parts shop with a counter and deliveries, working with dozens of suppliers. Each supplier has its own B2B portal with its own login, and the same part exists in several brands, at different prices, scattered across those portals.

The problem

The customer arrives with a license plate and a rough description ("the front bearing", a photo, a noise). Getting to the right reference takes experience, time, and patience: identify the vehicle, navigate technical catalogs, then compare price and stock portal by portal. All while the customer waits.

What we are building

A copilot for the counter: it takes the plate and the request exactly as the customer says it, identifies the vehicle and the part, and returns a single table with brand variants, price, and stock across suppliers. The operator confirms and orders; the decision stays theirs.

• 01From license plate to exact vehicle identification
• 02The natural-language request is interpreted into a part category and reference
• 03Technical catalogs and supplier portals queried in parallel
• 04One table: brand, reference, price, and stock, ready to compare
• 05First phase in development, with the main supplier integrated

The goal

• 01Minutes per request instead of tens of minutes
• 02Less dependence on a single person's experience
• 03Price and stock comparison always complete, never sampled
• 04A faster counter, customers waiting less