We sat down with Tom Wilger, VP of Artificial Intelligence here at BP3, to discuss the ins and outs of how IDP is being used by organizations to simplify their invoice process.
Sarah: Tom, let’s start with the ‘big picture’. Many technology analysts are suggesting the adoption of IDP solutions has hit an inflection with annual growth exceeding 50%. Can you tell us a little bit about IDP and why so many organizations are investing in the technology?
Tom: At its core, IDP is about automating how complex documents are processed - typically eliminating tedious and time-consuming manual work - and then feeding the output into downstream applications.
To do this at scale, we draw on a range of core technologies including optical character recognition, computer vision, machine learning and deep learning models, and natural language processing. From a maturity perspective, we have reached a point where IDP solutions can efficiently process all types of documents.
Although, as a tech geek, I’m excited about the technology, I’ll admit that isn’t what is driving the uptick in adoption. Simply put, organizations are investing in IDP solutions because they are capturing tangible benefits in terms of reducing costs, improving operational efficiency and effectiveness, and increasing productivity.
Sarah: Well, we may have to do another interview to dive deeper into the technology itself. For now, let’s focus on how it is being used. It strikes me that the potential applications for IDP solutions could be quite broad. After all, despite being in the Digital Age, we still see unstructured documents, often paper-based, at the heart of so many ‘mission critical’ activities. Are there certain areas where you are seeing the most rapid adoption?
Tom: Absolutely. Some of the most common use cases where organizations are realizing tangible benefits from IDP include invoice processing (which is an area where BP3 is doing some particularly interesting work with our clients), order forms, insurance claims, patient onboarding, and proof of delivery. However, as you mentioned there are really a wide range of potential use cases.
Sarah: You mentioned BP3 is helping its clients automate invoice processing. We know that many organizations struggle with processing thousands of invoices each month. Can you talk more about their challenges and why they are looking at automation rather than adding headcount?
Tom: Our experience suggests Accounts Payable departments in many organizations are struggling with similar challenges - especially those that receive a high mix of paper invoices and a diverse range of formats. Because their process is often tedious and manual, they struggle to control processing costs and cycle times - often alienating their suppliers and incurring financial penalties. Unfortunately, in today’s labor market, finding, hiring, and training new employees can be one of the biggest challenges that businesses face. In fact, employees are a company’s biggest expense. It’s why so many organizations turn to technology to simplify the underlying process.
Sarah: So we know that hiring new employees to process invoices might not be feasible for many businesses, but what’s another significant advantage to investing in automation software?
Tom: To be honest, the advantages of automation really speak for themselves. It isn’t unusual for organizations to realize 5X lower processing cost per invoice fueled by 2.5X improvement in both the rate of ‘touchless invoices’ and the ‘single invoice processing time’. At same time, clients are also experiencing accuracy improvements with a 50% reduction in invoice exceptions.
Sarah: We’ve talked a little bit about invoice processing and the need for automation, but can you explain what Intelligent Invoice Processing is?
Tom: Intelligent Invoice Processing is a solution that we’ve put together for our customers that is specifically designed for processing invoices at scale. From a technology perspective, we’ve curated best in breed, machine learning models for optical character recognition and invoice named-entity recognition. We then configure the solution for each of our customers by adding specific logic to the data extraction process that follows how their accounts payable team deals with the messiness of invoice data. Finally, because AI is never perfect, we backstop the machine learning models with our own Document Review team. So, whenever there’s a problem extracting a field because it's smudged, or maybe an address is missing a zip code, our Document Review team acts as the human in the loop to correct the problem.
Sarah: In other words, the client gives us information and we give them back clean data?
Tom: Exactly. Let me give you an example. We have a client for whom we are processing about 5000 invoices a day. I don’t know how many vendors they have, but I suspect it’s in the thousands. Most of these vendors are in the U.S., so their remittance addresses have a similar format. Now, when we extract addresses for THIS customer, they need that data broken out by street address, city, state and zip. That’s not a problem for us because our solution allows us to add customer-specific logic to the extraction process. At first, we were only processing invoices from U.S. based vendors, but one day we started getting Canadian invoices. Since Canadian addresses are formatted differently, these invoices were all getting sent to our Document Review team who would then manually extract the address data and send the invoice on its way. The customer didn’t know our Document Review team was doing this extra work, and why should they? Now, we eventually figured out what was going on, and we modified the custom logic to parse Canadian addresses. But, from the customer’s perspective, the beauty of this solution is that it just works.
Sarah: It seems like there’s always going to be a constant influx of invoices to process within organizations. Can you describe a typical BP3 engagement and what sets BP3 apart from its competitors?
Tom: Sure. Getting started with invoice processing is pretty straight forward. We typically start the engagement by finding out what fields need to be extracted from the invoices. It’s also important to understand if the client’s accounts payable department needs any sort of validation logic on any of the fields. Once we deploy our IDP solution to an isolated AWS account, and configure it with the client’s specific requirements, we’re ready to start testing the extraction process. Once that’s done, we typically work with the client’s I/T department to integrate the extracted invoice data into their accounts payable system. From start to finish, the whole process usually takes 4 to 6 weeks.
Sarah: Going off what you just said, what industries can benefit from this solution?
Tom: I’d say customers in manufacturing, healthcare, utilities, transportation, and retail. In short, any company that wants to improve their invoice process - it’s a low-hanging fruit that many businesses would be prudent to grab.
Sarah: Well thanks for sitting down with me and sharing more about how companies can simplify their invoice process through automation and with the help of BP3.
If you'd like to learn more about Intelligent Document Processing (IDP) and how it can help streamline your invoice process (or any document for that matter), check out our other videos on YouTube or get in touch with one of our experts.