The challenge
The customer, a large multinational equipment distributor, wanted to streamline invoice processing through digitisation by extracting account information from invoices into its custom-made ERP system.
The work was done manually and took several minutes per invoice. As the number of daily invoices grew, a human-centric process was no longer feasible, and the push to process faster led to more and more human errors.
Our solution
We delivered a web service that lets users upload an invoice and receive the extracted information in a predefined structure.
The service runs on a pipeline that combines document image processing, optical character recognition and classification of the invoice type.
- Processing and enhancement of document scans
- Information extraction based on optical character recognition (OCR) with post-processing
- Identification of information blocks
- Invoice type classification
- Packaging and export of the extracted information
The outcome
The service allowed the customer to process invoices without any human intervention. Invoice processing became 25 times faster and the customer saved 850 man-hours per month.
The error rate fell by half, since most errors had been human errors.
