Sebastian de GoeijShort answer
OCR, optical character recognition, recognises text in an image or PDF and turns it into characters. For invoices it is the first step: the computer can read what is there. Knowing which number is the total, which line carries VAT and which GL account applies is a second step. Classic scan-and-recognise tools use templates per supplier; AI recognition uses context and history.
Key takeaways
- OCR recognises characters in a scan, photo or PDF and makes text of them.
- Field recognition, such as date, total and VAT, is a separate step on top of OCR.
- Template-based recognition breaks with a new layout or supplier.
- UBL and Peppol invoices already contain structured data, so OCR is not needed.
- Autoboeker recognises invoices with AI without templates and books from context and history.
What is OCR?
OCR analyses an image, finds shapes that look like letters and digits and converts them into text. A scanned PDF or a photo of a receipt is just a picture to a computer until OCR runs.
OCR does not understand what it reads: it knows ‘121.00’ is there, not that it is the total including VAT. That needs an interpretation layer.
How does OCR work in invoice processing?
OCR starts a chain:
- 1The invoice arrives as PDF, scan or photo, for example via email, the app or WhatsApp.
- 2OCR converts the image into text with positions on the page.
- 3Recognition decides which text is which field: supplier, date, number, amounts, VAT.
- 4A booking proposal links GL account, VAT code and possibly cost centre.
- 5After review or based on confidence, the booking goes to the accounting package.
Where does classic OCR fall short?
Classic scan-and-recognise tools use templates per supplier. That works until the layout changes, a new supplier appears or an invoice deviates slightly.
Scan quality also matters: skewed photos, low resolution, stamps and handwriting. And the booking proposal does not come from OCR but from knowing the administration.
How does Autoboeker recognise invoices?
Autoboeker recognises supplier, date and invoice number, amounts and line information, VAT, GL account, cost centre and optionally project, without templates, looking at context and history instead of fixed positions. Above the firm's confidence threshold it books automatically, and every booking shows a confidence score and explanation.
Documents arrive via the app, client portal, a linked mailbox, WhatsApp or cloud storage through Connect. Structured e-invoices such as UBL and Peppol already contain the data; they need a content check, not image recognition.
FAQ
Questions about this topic.
Optical character recognition: software that recognises text in an image or PDF.
No. OCR reads characters; invoice recognition decides which text is which field.
Template-based systems need setup per layout. Template-free AI recognition does not.
No. UBL and Peppol invoices already contain structured data.
No. It recognises invoices with AI without templates, using the administration's context and history.
