Earlier this month, ABBYY participated in the Intelligent Automation Week in Chicago in which I had the opportunity to moderate one of the round table discussions on content intelligence and robotic process automation (RPA). I held discussions with over 40 people on various approaches to automating content-based processes with RPA. Topics discussed ranged from the difference between basic RPA and the growing need for advancing automation around processing unstructured documents and data, to an understanding of the role of artificial intelligence (AI) in processing unstructured content.
What was clear from the conversations is that organizations in the midst of piloting or scaling up their RPA operations are often finding processes they want to automate that involve unstructured content. While RPA is well suited for the automation of repetitive structured task, it falls short when it comes to extracting actionable information from unstructured documents and data. The path to implementing a zero-touch process requires the ability to understand content so that a robot can perform the required task.
Here are 5 key takeaways from the conference:
Demand for processing unstructured content
The challenge associated with processing unstructured content was top of mind for many at the conference. Initially, RPA software robots were considered unintelligent and lacking the agility required to handle the advanced skill, and non-standardized interactions like identifying a document and locating and extracting data from within that document.
As the RPA market has matured so too has the RPA adoption by various organizations. This helped focus attention on the need to advance automation by applying intelligent character recognition (ICR), and more cognitive skills for identifying unstructured documents, locating and extracting data, and understanding the context of text, with the goal of automating beyond simple rules-based processes.
Use cases for document automation spans the enterprise
The conversations we had with organizations provided real examples of where they saw intelligent document automation critical to their RPA initiatives. It is not surprising that a lot of these discussions revolved around finance related processes like invoices, purchase orders, and sales orders. This is often a business area where many RPA customers start with their RPA initiative. It is also a good place to start as the business benefits can be quite substantial in terms of driving down costs and increasing productivity significantly.
However, in key industries like banking, there are many examples where robots will require an added layer of intelligence for identifying a document and extracting relevant data so the process can proceed without human intervention.
According to many attendees, a large percentage of processes that organizations try to automate involves content coming from outside the organizations where they have zero control over the structure and format of the data. For example, orders coming from thousands of different vendors might be automated for your large customers, but the smaller ones, which often represents a larger percentage, send in orders in the form of a PDF, image, or even an email with text.
It’s more than OCR
RPA users are finally starting to understand that automating a document process is not just about using optical character recognition (OCR). OCR is important to recognizing data contained in a form or unstructured document, but that is just the first stage in digitization as it relates to documents, images, and text.
The real value customers seek is within the next level of intelligence that can be applied to documents and images, where structured and unstructured documents are identified, and text contained in those documents is extracted at a high level of confidence and fed directly back into the RPA process. Turning unstructured content into structured and actionable information enables the robot to carry on with its task.
The Meet Up with RPA and AI
Today, users of RPA are beginning to identify real world use cases that cannot be solved with RPA alone. Those use cases are ones that most companies struggle with when it comes to extracting information from unstructured documents, images, and text. This class of cognitive automation requires technologies like machine learning and text analytics to identify and extract relevant data, as well as understand the information. Go a step further, it is about applying machine learning to learn a set of unstructured documents that could have thousands of variations, and doing so in an automated fashion where the system learns and improves over time.
Cognitive automation enables robots to process emails, financial statements, contracts, or other types of unstructured documents and hold data that needs to be extracted and processed. To truly tackle this next set of content will require the technology to advance and become easier to design and deploy.
Education will continue
Over the past 3 years, I have attended many RPA events, but the level of conversations we are having today are far more advanced than 3 years ago. However, we still have a long way to go in terms of education. The underlying technology is important to the conversation, but what is equally important is the business problem that we are trying to solve. What organizations will likely find is that many projects that are ripe for the next phase of intelligent automation that can already be solved today.
At ABBYY, we have been working with organizations across the globe for years to help them turn unstructured content into structured information that can be connected into systems and processes. Many of the use cases RPA customers present to us are ones we have been solving for years where the technology is proven.
By Bill Galusha, Director, Product Marketing, RPA & Data Capture