Extracting value with AI invoice processing
Lisa Seguin Lisa Seguin
March 10 4 min

Extracting value with AI invoice processing

As digital transformation continues to reshape our professional and personal lives, there is an inexorable move toward digital transactions, content, and communication. Digital documents are replacing paper, but in many cases the way we handle them is largely the same. Invoice processing is one such document-centric area that still has one foot in the past and the other in the future. Whether digital or paper, the way we deal with invoices hasn’t fundamentally changed for over a century. However, change is coming.

In this article, we’ll explore the current state of invoice processing, and how AI can bridge the gulf between papers/PDFs and today’s fully digital accounting systems.

Following the paper trail

According to a 2019 report from Billentis, somewhere around 550 billion invoices are created annually, of which only 55 billion are exchanged on a paperless basis. By 2035, the number of invoices is estimated to quadruple. Invoices aren’t going away anytime soon, and many of them will still be dealt with in traditional paper formats.

Many such paper invoices are scanned before being processed in bulk, alongside invoices received in digital formats such as Adobe’s PDF. The classic steps of invoice processing must then be applied.

Figures have to be identified on the forms,comparisons with purchase orders must be made,values entered into accounting records and exceptions must be managed. It’s a time-consuming endeavor that requires a lot of labor hours and can result in errors if one number is entered incorrectly or a figure transposed with another.

In a world where fast, effortless digital transactions are standard, manual invoice processing seems almost an anachronism – a quaint relic of a bygone age. It’s not as if efforts haven’t been made to try and apply automation to the process, but until now they have yielded varying levels of success.

Early attempts at digital invoice processing

Optical character recognition (OCR) technology has historically been used to try and automatically find, gather, and process data from invoices. The intention was that the software identifies the figures on the document and then makes the calculations and evaluations.

Maybe this would have worked, for a few key pitfalls. Until relatively recently, OCR hasn’t been reliable enough to handle the myriad typefaces used in many different invoices or make sense of degraded characters in a poor-quality scan. This means errors can creep in and require calculations to be double-checked by a human supervisor, which goes against the entire point of the exercise: removing human effort from the equation.

Struggling with templates

One of the most important stumbling blocks to digital invoice processing in the past is that there is no single universal invoice format. This lack of uniformity has been difficult for software to handle. It just hasn’t been smart enough to be able to examine a document, like a human would, and discern what each column and figure relates to, without first knowing what to expect. Until now, this contextual understanding has been a step too far.

What was the remedy? The software would have to be provided with examples of an invoice template, with each area of the document marked up manually on the screen by a human user. The software would then know what each element of such a template was and deal with it accordingly.

However, each business may deal with countless other businesses, each using its own invoice template. Every one of those templates would need to be inputted into the system so the format would be recognized. This means creating an entirely new workload to support a system that, once again, is supposed to reduce human labor.

AI makes the breakthrough

Advances in AI mean that software is now smart enough to finally play a meaningful role in invoice processing. AI systems can easily and quickly be configured by administrators, using zero-code, to extract information from any document, enabling users to process a larger quantity and variety of forms in far less time.

These AI systems assist the user in extracting data from documents and, over time, their accuracy steadily improves through user feedback, saving time and mitigating errors. Good news for beleaguered accounting personnel.

Discover Document Intelligence

Element AI’s Document Intelligence makes smarter, easier invoice processing a reality. When staff are no longer weighed down by lengthy, tedious invoice processes, they’re able to work on higher value tasks that take the business forward. That means they can put their smarts to good use, too.