How to read PDF invoices with OCR.

OCR is the boring part of invoice automation. It's also the part everyone gets wrong first.

What OCR reads well

Structured invoice layouts, printed text, consistent vendors.

What OCR reads badly

Handwriting, faxed scans, image-heavy layouts without an OCR pass.

What to extract

Vendor, reference, amount, VAT, due date. Anything else is optional.