In this special guest feature, Eric Bussy, Worldwide Corporate Marketing and Product Management Director at Esker, discusses the importance of machine learning for enabling back office processes. Eric is responsible for the development of strategic products, services and solutions. He joined Esker in 2002 as Director of Marketing Communications, and in 2005, extended his responsibilities to include product management.
Many businesses have yet to understand the full potential of machine learning and cognitive computing for back office processes. The benefits range far beyond simply increased productivity and faster supply chains. Today’s advanced algorithms have the computational power and speed to gather, analyze and manage vast repositories of data. As the Internet of Things (IoT) transitions from theoretical to practical application, information harvested at every touch point in business operations will provide opportunities for insights into areas no traditional analysis has yet explored. The amount of data is too vast for humans to adequately mine and leverage. The key to gaining those insights—and keeping pace with competitors—is machine learning.
Where once we associated automation with standard manufacturing processes, it’s knowledge workers who stand the most to gain from advances in machine learning. From finance to healthcare to retail, automation will soon be critical to document-heavy industries. A recent McKinsey report said predictable manual processes, especially data processing, are among the activities that most lend themselves to automation. The potential to streamline the collection and processing of data is nearly unlimited. And it starts with back office document management.
For the purposes of discussing document processing automation, it’s more appropriate to refer to machine learning as auto-learning. Like almost all emerging tech, it’s cloud technology that enables the practical use of advances in auto-learning. Prior to the cloud, document processing automation took a very basic approach. Systems would build an ever-growing knowledge database around users’ habits. Over time, this rudimentary automation could make changes that the user would have made automatically, similar to autocorrect systems in word processing programs.
But this static data extraction method has limits. OCR is highly dependent on document quality and the use of specific fonts; handwriting is nearly impossible for it to convert. It recognizes characters falsely, sometimes failing to extract them at all. These limitations require manual correction, which puts strict boundaries on the amount of increased efficiency for office processes such as sales order processing.
Auto-learning takes this compute intelligence to the next level by completing tasks the user once did manually rather than just correcting data that was entered by hand. Like the first generation of automation, these cognitive processes take time to learn from data that users input repeatedly, such as customer orders. The system detects recurring patterns to make sense of data that didn’t make sense to it before, making note of specific keywords commonly associated with each field. The system automatically recognizes the keywords and searches for the appropriate related data to fill the field. Unlike previous iterations of document process automation, the goal of auto-learning systems is to extract certain data elements without any user intervention.
It’s flexible data extraction rather than static, and the system becomes more “intelligent” and efficient over time. In essence, it learns from its mistakes just like humans. Unlike humans, however, once that mistake is corrected, it doesn’t make it again. As the system learns, it more accurately extrapolates document content from each modification. For example, if a customer service representative (CSR) corrects the value of an inferred field, an auto-learning system refines the header data according to the data in a set of areas in the document, or from the data source itself. With every correction, the system learns more and becomes more accurate. As computational algorithms become more advanced, the system learns quicker and provides deeper operational insights.
In its 2016 Hype Cycle for Emerging Technologies, analyst firm Gartner placed machine learning at the pinnacle of its expectation peak, and predicts that within two to five years it will have reached widespread mainstream adoption. In fact, smart machine technologies are predicted to give rise to the most disruptive class of technologies over the next decade. Organizations will be able to adapt faster and predict trends more accurately. The speed of business is going to increase to a point where those who fail to implement machine learning early enough will undoubtedly be left behind.
Sign up for the free insideBIGDATA newsletter.