Classic invoice recognition needs to be trained: for every supplier and every document type, read zones are defined and maintained – where the invoice number is, where the total is. With ten suppliers that is workable. With a thousand it is not: every new invoice layout starts from scratch, and maintaining the templates becomes a job in its own right.
Modern recognition works without document training. It reads a document and interprets it – working out for itself that a particular number is the invoice number and a particular amount is the grand total, even on documents the system is seeing for the first time. This works right down to a photographed receipt or a handwritten invoice.
How you notice the difference day to day: roll-out takes weeks rather than months, new suppliers need no setup, and variable scan and photo quality does not throw the process off course.
How our recognition works →Please note: this article reflects the legal position and administrative guidance as at the date stated and is no substitute for tax or legal advice.