Every finance team has looked at invoice data extraction at some point. A supplier invoice arrives as a PDF, a tool reads the supplier name, the date, the total and the lines, and hands the data on. On a demo it looks finished.

In the office the hard part starts after that. Someone still has to find the purchase order, check the lines against it, confirm the goods arrived, and only then enter the invoice. The extraction saved the typing. The checking is where the hours really go in vendor invoice processing.

This article covers what extraction gets right, where it stops, and what a complete flow looks like when the data is checked against the ERP before it is entered.

What invoice data extraction does well, and where it stops

Header fields

Supplier name, invoice number, invoice date, due date, currency and total. On a clean PDF a modern tool gets these right almost every time.

Line items

Item description, quantity, unit price and line total. This is harder, because every supplier lays out the table differently, but AI readers handle it without a layout per supplier.

Languages, scans and photos

Invoices arrive in English, in the local language and sometimes in both on the same page. As a scan, a phone photo or a system PDF. The source should not matter.

It does not know who sent the invoice

A name on a PDF is not a supplier record. The invoice has to be tied to the right supplier in the ERP, with its payment terms and its open orders.

It does not know if the price is right

The extracted unit price is a number. Whether it matches the purchase order is a different question, and that is the one the controller cares about.

It does not know if the goods arrived

An invoice for 100 units when the warehouse received 80 looks perfect to an extraction tool. It only looks wrong next to the goods receipt.

It leaves the entry to a person

Extracted data in a spreadsheet or a queue still needs someone to enter it in the ERP. The typing moved. It did not disappear.

What the complete flow adds

Identify the supplier

From the sender address and the document. New suppliers work from the first invoice, with no template to set up.

Find the purchase order

The PO reference on the invoice, or the supplier and the items when the reference is missing. The open PO is read from the ERP with its lines and prices.

Check every line

Quantity, unit price and total, each against the PO and the goods receipt. This is the three-way matching that AP teams do by hand today.

Stop what does not match

A higher price, a missing receipt, a duplicate invoice number. Each stops with a plain reason, so the person who opens it knows what to decide.

Enter what matches

Only invoices that passed the checks are created in the ERP, linked to the PO and ready for approval and payment.

This is the point Avi Rafalson, DOConvert’s CEO, makes in Beyond Extraction, the next frontier in intelligent document processing. Reading the fields is the entry ticket. The value is in what you check them against.

What to look for in invoice data extraction software

A direct connection to your ERP

The tool should read the PO and the goods receipt from the ERP itself and write the invoice back. Not through a middleware platform, not through a CSV export.

No template per supplier

Ask what happens when a new supplier sends a first invoice. If the answer involves setting up a layout, the tool will not keep up with your supplier list.

Exceptions with reasons

A rejected invoice should say why. A price difference and a missing receipt need different people to look at them.

Multi-step flows

PO, goods receipt and invoice belong to one flow. A tool that handles invoices in isolation leaves the matching to you. That is how supplier invoice capture turns into a real saving rather than a faster inbox.

How it works with DOConvert

DOConvert is the AI layer for the ERP. It connects directly to Priority through the DOConvert Priority connector, which is bi-directional, and to Oracle Fusion Cloud ERP and Infor M3, with no middleware in between. Here is one invoice.

The invoice arrives by email

A PDF with 18 lines from a supplier you have worked with for years. Nobody forwards it or scans it again.

DOConvert identifies the supplier

From the email address and the document. The supplier record in the ERP comes with its open purchase orders.

The PO and the goods receipt are read from the ERP

Lines, quantities, prices and what the warehouse confirmed. The comparison starts here.

Every line is checked

Say 17 lines match the PO and the receipt. One line shows a unit price above the PO. That line stops with the reason, and the invoice waits for the buyer or the controller.

The invoice is entered

Once the open line is decided, the invoice is created in the ERP, linked to the PO and the receipt. From the AP team’s side, they stopped keying supplier invoices from email and started deciding exceptions instead.

Who it fits

• CFOs and VPs Finance who want invoices checked against the PO before anyone sees them
• Controllers and AP managers with hundreds of supplier invoices a month in several formats
• IT and IS managers on Priority ERP, Oracle Fusion Cloud ERP or Infor M3 who want a direct connection, no middleware
• Companies of 200 to 5,000 employees, and larger groups that start with one subsidiary

Questions and answers

Is invoice data extraction the same as OCR?

No. OCR turns an image into characters. Extraction finds the fields in those characters. Neither checks the data against the purchase order or enters it in the ERP.

Do I need a template for each supplier?

Not with an AI reader. A first invoice from a new supplier is read and checked the same way as one from a supplier you have worked with for years.

What happens when an invoice has no PO reference?

The supplier and the items point to the open order. If there is no matching order at all, the invoice stops with a note, which is usually the right outcome.

Is this the same as purchase invoice processing?

Purchase invoice is the UK term for a supplier invoice. The flow for purchase invoice processing is the same, read, identify, check against the PO, enter.

How long does it take to set up?

A typical DOConvert implementation takes about 10 business days, in the cloud or on your own servers.

Want to see it on your own documents? Book a Demo or Get Started