Invoice Extraction is not AP Automation

Published on   2023-11-04

There are so many old veteran vendors and new IDP startups out there, and the one thing almost all of them have in common is:

We process Invoices!

Ok, fine, we believe you. Everyone can process invoices. AP Automation solved! Or is it?

What is AP Automation?

AP Automation automates the Accounts Payable part of the Purchase-to-Pay workflow that exists in every company. The AP part of that process deals with the incoming invoice. But extracting the data from it, and thus saving the manual work of keying this data into the ERP system, is only a small part of that workflow.

An invoice can be paid only after many things happen:

  • The invoice data is entered into the ERP system.

  • The invoice data is verified.

  • The vendor is identified.

  • The invoiced items match what was ordered and delivered.

  • The line items are assigned to cost centers and accounts.

  • The invoice is approved.

Let's take a look at these steps.

Data Entry

All relevant invoice data needs to be entered as a new object in the ERP system. The IDP solution typically uses machine learning to extract the raw data as printed on the invoice and maps it to the ERP system. Already at this step, a data extraction service alone is not sufficient. The IDP solution also needs connectors to the common ERP systems so it doesn't have to store the extracted data in an XML file and someone has to create an import script from the ERP system to load that XML data. Ideally, the AP Automation solution interacts directly with the ERP system.

Data verification

Before the data is submitted to the ERP system, the AP Automation solution must check that the invoice data is correct. First, it should check that the data itself is consistent:

  • The amounts add all up (subtotal, total, line items, etc.).

  • The dates make sense (invoice date is in the past but after the order date etc.).

Then it should reconcile the data with the ERP master data:

  • Is the extracted PO Number in the system?

  • Do we need the tax amounts?

  • Are tax IDs printed (required in the EU) and are they correct?

The Vendor

It is of course not good enough for an IDP product to extract the vendor name and address as printed on the document. The ERP system needs to know which exact vendor in the master data it is. So the AP Automation solution needs to map the vendor data from the invoice to a record in the vendor master data. All that the ERP system cares about, is the vendor record's ID. Due to OCR errors and other discrepancies between invoice and master data, this is not a trivial match or simple lookup. AP Automation products with a fault-tolerant data matching technology excel here.

If the vendor doesn't exist in the ERP system, the AP Automation solution should trigger a workflow in the ERP system to create a new vendor and put the invoice on hold until that happens.

If the vendor data has changed, e.g. the vendor has a new bank account, the AP Automation solution should detect that and trigger workflows to update the master data.

Order and Delivery

One of the most time-consuming manual tasks is not the data entry, but the mapping of the invoice line items to the purchase order or delivery line items.

Vendor invoices use line item descriptions from the vendor, not from the buyer's material list. The vendor doesn't know the ID of the buyer's material, so that is never printed. And lastly, often the price and the quantity differ. Maybe the price increased since the order, or a contractual discount was not applied. Maybe there was only a partial delivery so the quantity deviates.

The accountant needs to map line items even if almost all their data differs (different descriptions, unit price, quantity, total price). This is where an AP Automation solution with fuzzy matching and support for multi-PO invoices, partial deliveries, goods receipt matching, configurable price tolerances, etc. can shine and save a lot of manual work.


Once every line item is mapped to a Purchase Order or delivery line, they all need to be coded. Accounts and cost centers need to be assigned. That can be trivial if you have only 10 different accounts, but highly complex if you have general ledger codes that are composed of 10 hierarchy levels each with 100s of entries, and if the invoice has 200 line items.

Intelligent AP Automation software can automatically suggest accounts and cost centers based on the manual assignment history or machine learning.


Last but not least, when the invoice is properly stored and mapped in the ERP system, every company has different approval requirements. Sometimes that depends on the total amount, and higher amounts have to be approved by higher-ups. Sometimes multiple people need to approve and that can even depend on the accounts assigned to the line items. Good AP Automation solutions have a highly configurable approval workflow.

Other features to look out for

  • eInvoicing support

  • Highly efficient user interfaces

  • Self-improving data extraction through machine learning

  • Support for QR code invoices

  • SaaS or onPremise solution needed?

  • Exception and rejection management

  • Supplier communication to reconcile issues

  • Supplier self-service portals


Automatically extracting invoices is great and saves time but it is only a small part of the AP workflow, most of which can also be automated. But that requires much more sophisticated solutions, not just data extraction services. Some vendors, who offer highly configurable complete AP solutions, in no particular order:

  • Basware

  • Corcentric

  • Kofax

  • Coupa

  • AvidXChange

  • SAP with OpenText (built into the ERP system)

Share on   Facebook   Twitter   LinkedIn   Email   Whatsapp   Telegram
This post was published in
Intelligent Automation