Procurement digitalisation with an item catalogue built by OCR
Item catalogue seeded by OCR of invoices, with orders checked against negotiated prices.
The challenge
Supplier orders went out by phone and email, with no item catalogue, no control of negotiated net prices and no traceability of consumption per piece of equipment.
Our response
Rather than plugging in a procurement package, we built the module directly into the existing platform so that it shares the same item catalogue as the inventory. The item base was seeded by OCR extraction and structured parsing of historical supplier invoices, which made it possible to start with a real catalogue rather than an empty one. The process follows the actual split of roles: the field expresses the need, first checks stock availability, and management approves and then orders. Unknown items are created as drafts and reconciled when the invoice is received.
Key points
Item catalogue seeded by OCR and parsing of historical supplier invoices
Stock is queried before any order, which avoids redundant purchases
Automatic check of invoiced prices against negotiated net prices
Traceability of consumption per piece of equipment and supplier consolidation
A single item catalogue shared between procurement and inventory
Technical stack
- Document OCR and layout parsing (docTR / PaddleOCR)
- LLM structured extraction with a constrained JSON schema
- PostgreSQL + item reconciliation engine
- Two-level approval workflow
- Next.js 15 / TypeScript
- Open-source LLM hosted in Switzerland
Sector
Industrial & materialsLet's talk about your project
Tell us about your business challenge and we'll explore together how a tailored AI solution could support it.