Intelligent document processing

What it means

Intelligent document processing is the automatic reading and extraction of data from invoices, contracts, statements or orders using OCR and artificial intelligence that also understands non-standard layouts. Unlike simple OCR, it recognises not just characters but what the data mean.

The system can find the supplier, VAT number, dates, amounts and line items on an invoice even though every supplier's document looks different. It then passes the data to the accounting software for posting, greatly speeding up document processing.

IDP combines text recognition, machine learning and increasingly generative AI. It is often paired with RPA, which automates further steps such as posting, matching to purchase orders or routing for approval.

How it works in practice

In accounting, IDP removes the need to retype purchase invoices. The accountant checks and approves proposed data instead of entering it manually, which with large document volumes saves a great deal of time and eliminates keying errors.

The system learns from corrections: if the accountant fixes a misread field or posting, it proposes the right one for later documents from the same supplier. Accuracy therefore improves over time but drops for new suppliers and poor-quality scans.

With mandatory electronic invoicing on the way, some documents will arrive already structured and IDP will not be needed for them. It will remain useful for paper and PDF documents, contracts and other papers.

What to watch out for

IDP results must be checked. A misread amount, VAT rate or supply date leads to incorrect posting and returns, and generative AI carries the risk of hallucination, where the system fills in data not present in the document.

Documents often contain personal data and trade secrets. With cloud IDP, check where data are processed, whether the provider uses them to train models and whether a data processing agreement is in place.