For many organizations, AP invoice registration is supported with many team members performing manual data entry.  Volumes, system limitations, and available information can also affect the capabilities of the AP team, which can limit the company from implementing PO and line item validations or even 3-way matching.  The promise of cost savings often drives the desire to automate the AP invoice registration process, which has accelerated technology support and service company offerings all wanting to claim their offering is superior to another.

Typically, the extent of business process throughput is typically tied to how much you want to spend.  There are many start-ups with venture capital backing that promise their AI and Machine Learning solution will automatically read invoices out of the box, and then, magic happens to validate the invoice data before registering the information in your AP system.  This dream can all be yours for the starting price of $250,000 just to engage a conversation!

For those AP teams working on more of a shoestring budget, the promise of Optical Character Recognition (OCR) to digitize the PDF invoices is/was enticing.  The issue with OCR-only solutions is that OCR just moved the issues from data entry to data correction.  OCR is an imperfect science, so depending on the quality of the PDF, an “I” could be read as a “1”, an “O” could be read as a “0”, or a “W” could be 2 “V’s”.  Unfortunately, OCR-only solutions take this imperfect data and register it inside of your financial system, therefore the correction is required to review and validate what was entered.  In most cases we’ve seen, OCR-only solutions caused more problems and did not result in the expected cost savings.

Related: Vigilant’s Intelligence Process Automation Services.

OCR engines with an integrated field capture, validation application (like ABBYY FlexiCapture or Microsoft Form Recognizer) have better success, since they usually take inputs from the financial systems to validate data.  These applications enable the training of invoice fields to extract and then use machine learning algorithms to improve the manual processing over time.  One word of caution on these applications is that they require a continuous improvement methodology embraced by the AP team to keep up with the changes that occur over the course of business (e.g. Business acquisitions, name changes, address changes, logo changes).  AP teams that have not embraced the continuous improvement aspect required with these applications have had throughput erosion, resulting in more and more invoices requiring human involvement to process.

The biggest issue for organizations trying to automate AP invoice is that the vendor data was not governed, therefore companies could have 5 different entries for AT&T or IBM.  Bad vendor data causes issues for these advanced OCR applications with machine learning because the algorism cannot always detect the correct vendor by the name and address since there are many competing entries.  If the OCR application cannot automatically detect the correct vendor, then you typically must hard-code the vendor you want the engine to pick up, which means you are no longer using machine learning or leveraging advanced technology. 

Our recommendation is to start your AP invoice automation project by cleaning up and standardizing your vendor data and creating a data governance policy to ensure vendor data (or any data at a point of origin) is validated and controlled before entering into your key processing systems where the data is used in downstream reporting.

Robotic Process Automation (RPA) offers an intriguing middle-ground, where the RPA Bot can engage the OCR engine to digitize PDF documents and extract the trained fields, then the Bot can be programmed to validate data, correct data, derive missing data, and perform 2-way and 3-ways matching.  We’ve seen 100% throughput when using standardized vendor data, OCR tools, and RPA, but in extreme environments with bad data, up to 80% was consistently achieved.

If your organization is using Oracle ERP, consider the Oracle iSupplier Portal (iSP), which gives your vendors and suppliers a customer-facing portal to upload their invoices, track submissions, check on payment statuses, and cut down on the customer inquiries that take your team away from investigations and processing throughput.  Vendor’s appreciate the self-service nature of the iSP portal, which gives real-time feedback to your vendors on the status of their invoices. 

Looking for a replacement of your expensive OnBase or OpenText Enterprise Content Management (ECM /CMS) then consider what Microsoft is doing with their SharePoint 365 offering paired with Microsoft Syntex, which can read and tag document, powered by Microsoft’s AI engines.  Microsoft has been working to make SharePoint and Teams a single collaboration and content platform that can reduce your operations expenses.

Thank you for reading.  We hope this article helps you on your AP invoice automation journey. Let us know what questions you have or drop us a line to give us feedback on our point of view.