AI, Automation, Machine Learning, Business Process, Accounts Payable

Data Extraction from Unstructured Invoices using Cognitive Automation

The accounts payable (AP) process is immensely important in any organization since it involves nearly all of a company's payments outside of payroll. Regardless of the company's size, the mission of accounts payable is to pay the company's bills and invoices that are legitimate and accurate.

The AP process covers the complete cycle from vendor master maintenance through procurement and invoice processing to the resulting payment processing and period closing activities.


When an invoice is received by AP, following steps are taken :

  1. Invoice recording
  2. Vendor creation (if it is a new vendor)
  3. Validation and matching between purchase order raised, goods receipt note (if applicable) and invoice received from vendor.

Once, the above cycle is completed successfully, payments to suppliers are done and recorded in general ledger.

To safeguard a company's cash and other assets, the AP process should have internal controls to:

  • prevent paying a fraudulent invoice
  • prevent paying an inaccurate invoice
  • prevent paying a vendor invoice twice
  • be certain that all vendor invoices are accounted for
  • be certain that all the vendor invoices are paid in time

Business leaders are struggling to save operations cost and time on their processes while improving scalability and reducing errors.

According to, it is found that on an average AP process cost per invoice processed is 7.75 USD.

Below graph shows AP Process cost per invoice processed (Cost Vs Performers)


And according to APQC research, labour cost consumes 62% of the total AP costs.

Also, since the AP processes are largely manual, sometimes timely payments are not made resulting in penalties.

As the companies receive invoices from different sources and channels, they are in different structures, formats and languages. Therefore, companies do need a automation solution which can understand unstructured invoices and extract data from them without manual intervention to save cost, time and human errors. This problem can be easily solved using machine learning which is a part of Cognitive Automation (CA) technology.

The study defines Cognitive Automation as the “simulation of human thought processes” that allow systems to make “autonomous decisions” and automatically adapt to “changing conditions and evolving business rules and dynamics.”

Cognitive Automation provides a format/language agnostic way to automate AP processes. In comparison to RPA or traditional OCR based rule engines, it is more scalable and delivers better results in terms of costs saving and processing times.

At Ikarus, we have developed a Cognitive Automation framework which can automate your AP processes around unstructured text data. Here's a demo video on smart data extraction using Ikarus technology:

Click here to download the case study demonstrating the value of Cognitive Automation in AP Process