Distribution / Electronics · 2 weeks
Supplier price lists parsed in seconds instead of hours
Dealers received dozens of unstructured supplier price lists a day. The system now reads any format, applies each dealer margin and returns a ready CSV.
- to produce a price list
- hours → secondsto produce a price list
- device models recognised
- 30+device models recognised
- to production
- 2 weeksto production
The problem
Every supplier sent prices as free text in a different shape: model, colour, storage and country mixed into one line, no two lists alike. Somebody had to read them, work out which device each row meant, add a margin and retype the result into a customer-facing list. That was hours of work every single day, and the errors it produced were expensive.
What we built
- Parser that recognises model, colour, storage and country of origin from raw supplier text in any layout
- Per-dealer margin rules applied to the whole list with one command
- Output as a formatted message and as a CSV file ready for a customer or a spreadsheet
- Device catalogue kept current, from iPhone 11 through the latest models and Apple Watch
What changed
- A price list that took hours is produced in seconds
- Margins applied consistently, so no more rows priced by hand and no more typos
- Same pipeline serves every dealer with their own rules
- 30+ device models recognised without manual mapping
Other work
Hospitality
Booking agent that replaced the front desk after hours
A venue was losing reservations every night and weekend. An AI agent now takes bookings around the clock and hands the front desk a clean queue in the morning.
E-commerce / Wholesale
One post, priced and delivered to every partner
A supplier was forwarding every product to dozens of partners by hand, recalculating the price for each. Now one publication reaches all of them, each at their own rate.