The challenge
An industrial supplies distributor receives framework agreements from its construction customers: Excel sheets with long lists of items, usually without prices. To quote, the sales team had to identify each line, find which supplier offered it cheapest per unit and propose a selling price.
It was slow, manual work, repeated for every customer and every yearly review, even though many items come up again and again.
The solution
We built a system that reads the customer’s Excel file in its original order and matches each line with a master catalogue of items. It first looks for safe matches, using names it already knows or barcodes. Anything it can’t resolve that way goes to an AI model, which compares candidates and checks the measurements.
For each item it brings in the unit cost from the cheapest supplier and the price that customer paid last time. The salesperson gets their own file back with a review sheet: safe matches are filled in, doubtful ones are flagged, and they only set or confirm prices.
On closing, the system returns the list in the customer’s format and order and keeps what it learned for next time.
Technologies
- Python
- PostgreSQL
- Claude (Anthropic)
- n8n
- Excel
- Supplier APIs
The result
In tests with real lists, the system resolves the items it already knows on its own and leaves the salesperson what really needs their judgement: new items and prices. A person always sets the selling price.
Because every list feeds the catalogue, the manual work shrinks from one cycle to the next, especially with repeat customers.