We cut invoice processing time by 80% and eliminated manual data entry errors. Full automation in 3 weeks.
Problem
The client processes 400+ incoming invoices per month from various suppliers. Staff manually downloaded PDFs from email, typed data into Excel, then entered it into the ERP — 2 people, 2.5h a day, with regular errors.
Solution
Python script monitors a dedicated email address, downloads PDF attachments, and launches the pipeline. Handles both text PDFs and scanned invoices (OCR via Tesseract).
VAT number, invoice number, date, line items and amounts extracted by regexes and NLP models. VAT validation against GUS API. Invalid invoices go to a manual review queue with highlighted fields.
Valid invoices automatically enter Fakturownia (archive) and the ERP via API. Make.com orchestrates the flow and sends a daily summary to the team email.
"A few weeks of work and we've saved hundreds of hours per year. Best investment of the year."
– Operations Manager, logistics
Start your project
Tell us about your project — we'll get back within 24h with a concrete plan.
Get in touch