Document capture and coding by artificial intelligence

What data the software reads from an invoice

SaldeoSMART reads a document from a PDF, a scan, a phone photo, an e-mail or from KSeF, Poland's National e-Invoicing System, and pulls out the data: the counterparty, the number, the dates, the amounts, the VAT rates and, with the right configuration, line items and barcodes too. On that basis artificial intelligence suggests the category, the dimensions and the document type.

The accountant checks the suggestion, fixes the exception and approves. Nobody retypes invoices any more.

It looks slightly different in a firm and in a company

In an accounting firm documents arrive from clients in batches: through the portal, by e-mail, from the mobile app. After capture the accountant sees a list of items to check instead of a pile of scans. The more documents from the same supplier, the fewer the corrections, because the system learns the pattern.

In a company it is mostly about purchase invoices. They come in by e-mail and from KSeF, get a suggested coding, and the person who signs off the spend sees them for approval with the category and dimension already filled in.

How many credits capture uses

• An invoice from a scan or a PDF: 1 credit per page, not per document. That is why our calculator asks for the average number of pages.
• Coding an invoice from KSeF by AI: 1 credit per document. Simply downloading it from KSeF costs nothing.
• A bank statement: 1 credit per three transactions read.
• Reading line items and barcodes is charged extra. It is an option for companies that post goods, not just totals.

What is in which plan

In the plans for accounting firms

In the plans for companies

Source: SaldeoSMART price list, October 2026 (the same one our calculator uses).

Invoices that need correcting the first time

Capture copes with most invoices, but not with all of them. A document the system sees for the first time (an unusual layout, a poor scan, a foreign supplier with its own template) needs a correction. After a few documents of the same kind, the corrections stop. We have already met most unusual invoice layouts at other clients.

Capture quality starts at the scanner. A crooked photo taken in a hurry costs more time than it saves, so we agree scanning rules with the team during the implementation.

What drives the credit bill

During the implementation we set up categories and dimensions for your chart of accounts, rules for the AI, counterparty recognition and posting schemes. The key decision is which documents go through the AI and which go straight to a scheme. That decision has the biggest effect on the monthly credit bill.