Blog

How AI is handling the ugly problem of tilted invoices

Victor Melo
Victor Melo
Machine Learning Engineer

AI technology is revolutionizing the world, and accounting is no exception. Find out how AI is handling the ugly problem of titled invoice scans.

June 9, 2022

4 min read

How AI is handling the ugly problem of tilted invoices

In most cases, invoice processing solutions are designed to process documents that are 100% straightened when scanned in. 

You can then use RPA paired with optical character recognition technology (OCR) to extract the data from your invoice to send it to your accounting software. However, when the OCR technology receives an invoice in a tilted position, this approach becomes problematic. Tilted invoices can lead to line items ending up in unexpected places, and this will ultimately lead the OCR system to fail when extracting information from the invoices. 

Why a tilted invoice is more than just ugly

Let’s cut to the chase. Without AI algorithms, OCR can’t read or understand line items on a tilted invoice as a human would. This means that if you are using this technology, it can oftentimes group line items (or omit items) when it shouldn’t. This can lead to huge financial problems for your company.

In other words, your OCR-based invoice processing solution could be letting you down — and your business could be paying for it!

Example: Invoice ingestion can get a little messy

For example, let’s say you have an invoice from XYZ Company with one invoice number, but three different line items: Item 1, Item 2, and Item 3. 

If your system can’t read the line items properly, then your accounting software will not ingest the expenses the way they were intended. Your accounting software could pair Item 1 with Item 2, which leads to grouping and coding incorrect expenses. This can be the catalyst for all sorts of problems in the long run, including duplicate payments or unhappy vendors who only receive a portion of your payment. Here is a list of some problems caused by misread invoices:

  • Incorrect payments - pay too much or not enough
  • Incorrect invoices posted to your general ledger
  • Inaccurate monthly reporting
  • Costly hours correcting the issue IF it’s found before or after the books are closed

These system issues are a bookkeeping and invoicing nightmare for CFOs, managers, and accountants alike.  If left undetected, this can cause both financial and legal implications within your company.  

How AI is fixing tilted invoices

With an AI accounting system like the one we have with Vic.ai, our technology has been developed to correct invoice errors before they happen. Common problems our system can avoid include fixing tilts and ensuring invoice coding and posting accuracy.  

So, how exactly does our system do this? Let’s talk through how the algorithm works when fixing tilts.

Tilted invoice graphic

  1. Finds horizontal lines

The first step is to find horizontal lines. Our algorithm searches scanned images for horizontal lines and recognizes that they are lines because they are consistent in width and orientation. This information is then used to determine whether or not an invoice is sideways relative to its original position.

  1. Recognizes the tilt

Next, the algorithm will recognize that since the horizontal line is slanted, there is a tilt in the document. It will recognize this tilt and account for the discrepancy from the horizontal line.

For instance, our Vic.ai technology can quickly scan a document to see if there is a tilt on the page. It will then “autocorrect” the position internally to prepare the system for the correct reading of the document. 

  1. Accurately reads data

After the algorithm has found the horizontal lines and recognized the tilt on the page, it will read the data with accuracy. This safeguards your company from financial crisis, vendor frustration, or other problems related to billing or bookkeeping. 

Become a finance pioneer. Sign up for Vic.ai's newsletter.

Case study: Chess & tilting issues

Before becoming an engineer at Vic.ai, one project I worked on during university was developing a user-friendly Chess platform.  One common problem I ran into time and time again was tilting issues. These tilting issues lead to errors in gameplay, not because of player error, but because of the system itself. This was something I had to troubleshoot to ensure that the system could read both horizontal and vertical lines on the playing board to ensure optimal playability for the end-user.  You can see a similar example to this when LiveChess2FEN used letters and numbers to represent different board positions during a game of chess. 

Playing Chess

As with digitizing chess, the same is true with invoices. Without invoice tilt re-orientation, your system can cause error. No fault to you or your vendors invoices, but because of error issues within your own legacy system. If your OCR or ICR software doesn’t recognize a problem from the tilted invoice, it has no way to fix it.

What if your invoices were right every time?

AI can detect and fix what's wrong with an image, no matter how it's angled. This means that if you are an invoice processor, you can quickly take a scan of an invoice “on-the-go” without having to worry about the angle of the invoice. Whether it be from a micro-tilt or a sloppy scan-in,  selecting a tried-and-true accounting AI software will give you peace of mind about every single invoice scanned in.  

In a world that hasn’t gone paperless (yet), we want to help you alleviate any machine or human errors that can pop up with invoice processing. That’s why our accounting AI software is here to correct mistakes — before they even have the chance to happen. 

Streamline invoice processing the right way with Vic.ai. Let’s avoid invoice processing errors — together. Download our e-guide to see how you can start revolutionizing your financial operations today. 

Autonomous AP processing: 10 ways AI streamlines AP

BE A FINANCE PIONEER

Sign up to receive the latest news, updates and valuable accounting and finance resources conveniently in your inbox.