Canon leverages AI for document processing to save 6,000 hours

Canon USA works with a large network of vendors. Its finance department struggled to receive and process as many as 5,000 vendor invoices a month—most of them filled with a complex matrix of information like date ranges, machine serial numbers, meter readings, fuel fees, and other charges. On top of that, many of the invoices were created on paper documents, and much of the processing involved invoices for relatively small dollar amounts—sometimes less than $150. 

“We had four full-time employees dedicated to processing all of this paperwork, and sometimes there were human errors,” says Thomas Earvolino, Director of Financial Systems for Canon USA. “We had a basic RPA workflow in place, but it relied on RegEx models that were rigid, hard to maintain, and had little or no learning capabilities or flexibility when it came time to add new vendors. There were a lot of exceptions to processed invoices, which required many interventions by our employees. It was all pretty mundane but essential work. And at best, the automation was delivering 35% to 40% accuracy.” 

Accelerating invoice processing with AI 

Following conversations with UiPath, Canon USA decided to try a different route, using UiPath Document Understanding with its powerful artificial intelligence (AI) functionality along with UiPath Action Center, which enables teams to easily manage exceptions. 

“We were recommended to work with Greenlight Consulting,” Earvolino says. Greenlight is a UiPath Diamond Partner and was named UiPath 2022 Partner of the Year. “Greenlight had a clear understanding of our challenges and what we were trying to achieve, and within four months we had a working model that was ready for production.” 

The Greenlight and Canon USA teams collaborated to replace the RegEx-based extraction with customized machine learning (ML) models for invoice processing. To help handle exceptions in a timely manner, Action Center was used. The robot was programmed to manage all Oracle 11i SAP screens, including popups which, in the older system, prevent seamless end-to-end processing. Processed invoices were sent to an Oracle 11i system. 

Automation brings greater accuracy and efficiency and far less need for human interaction 

Earvolino says UiPath Document Understanding has delivered dramatic improvements to his department’s invoice processing operations. 

In less than nine months after deploying the UiPath solution, we processed about 40,000 invoices, or about 4,500 monthly. We initially had a goal of processing 75% without human intervention, but achieved about 90% straight-through processing during that time period.”

Thomas Earvolino, Director of Financial Systems, Canon USA 

The much more efficient and accurate processing delivered by the AI capabilities of Document Understanding allowed Canon to repurpose the staff to other value-added activities. 

Earvolino says occasionally during busy periods, the department may have to assign existing staff to handle very high workloads. But in general, there is far less human intervention needed than with the legacy system. Employees using Action Center typically need just one minute or less to fix invoice issues. 

“We estimate that we’re saving about 6,000 hours annually in employee time through the elimination of manual interventions. We also had a three-to-four-month backlog of invoices prior to this solution,” he says. “That backlog was eliminated. It’s a heavy weight lifted off our shoulders," Earvolino adds. “We were able to accomplish more than we thought was possible in such a short time frame and we’re grateful to have had a dedicated implementation partner like Greenlight. We wouldn't be here without them.”

Another key feature of UiPath Document Understanding is that AI can keep learning as it processes more and more data. However, when there are significant changes affecting the ML modeling – for example, most vendors are now adding fuel surcharges in the wake of high gas and diesel prices – Canon USA can work with their implementation partner, so the ML models accommodate the new information.  

UiPath Document Understanding is very scalable and can be an immense help in increasing efficiency, improving the accuracy of invoice handling, and putting employees to work on more value-added tasks.”

Thomas Earvolino, Director of Financial Systems, Canon USA 

Canon is currently looking for other areas where they think AI and Document Understanding could be beneficial. 

I am impressed with the efficiency and accuracy AI has brought to our operations. It's clear that the power of AI has enabled us to streamline our processes, freeing up time and resources for other critical tasks. Looking forward, we plan to expand our use of AI to further optimize our workflows and enhance our overall business performance.

Thomas Earvolino, Director of Financial Systems, Canon USA