Supplier invoices posted without re-keying

Problem

Around 300 supplier invoices a week arrived as PDFs in a dozen layouts, and two people re-keyed them into the accounting system. Errors surfaced weeks later as unreconciled totals, and the process lived entirely in those two people's heads.

Approach

We mapped the process as it was actually performed, then built a pipeline that extracts the fields, matches each invoice to its purchase order, and posts the clean ones automatically. Anything that fails a check, such as unreconciled totals, an unknown supplier or a duplicate number, lands in a review queue with the reason attached. The extraction model runs on the company's own server.

Stack

• Python
• PostgreSQL
• Docker