Most firms have some experience with Robotic Process Automation (or RPA). Some have built RPA development factories to accelerate delivery of bots. Others have only dipped their toes in the water with some pilots.
But leveraging RPA alone, without the context of a broader automation strategy, often poses challenges for organizations.
One New York-based firm is currently unable to build around 100 bots to streamline its operations area, not because they lack the talent to do so, but because the upstream data is unstructured, arriving via ad hoc emails from sales representatives.
Without cognitive tools — like natural language processing (NLP) that can transform unstructured data into predictable formats, the real benefits of RPA are yet to be unleashed.
We see best-in-class organizations viewing RPA as only one tool in the toolbox for a broader Intelligent Automation (or IA) strategy.
An Ecosystem of Automation
One large manufacturer with which we work has successfully reduced over 80 percent of its manual effort related to the Procure-To-Pay (P2P) process. A simplified illustration of the process depicts several tools, stitched together to provide a more complete solution than RPA alone.
Starting at the left side of the figure, you can see the following four steps are crucial to enabling automation, whether via API calls or bots:
- Lift text from images in the form of scanned documents
- Classify the documents
- Interpret the documents
- Review and process exceptions
We know from experience that RPA alone offers significant benefits, ranging from reduced cycle-time, decreased error-rates, increased customer satisfaction, and more.
But today’s RPA tools generally require some form of predictable and/or structured data. The inability to automatically interpret unstructured data and turn it into actionable steps can severely limit the ROI of some key processes.
The Criticality of Interpreting Unstructured Documents
For large commercial insurance companies, many key processes are initiated through ad hoc emails.
A contractual change to a policy (i.e., an endorsement) can look like this:
With a recent real world test at a top 10 insurer, Capgemini has demonstrated the ability to reduce manual processing of these types of emails by 60 percent and more.
Using three distinct technologies — NLP, natural language heuristics, and canonical data — Capgemini achieved higher-than-expected metrics on a test dataset of several hundred emails that the system had never before seen.
These metrics include the handling of key attachments (standardized PDF forms; various workbooks describing vehicles, locations, revenue, industry types and others; as well as inline text).
Well-managed RPA programs can offer many companies impressive benefits, both quantitative and qualitative. RPA programs can drive increased revenue, margin, and customer loyalty.
However, many crucial business processes begin with documents.
A holistic solution that addresses these “cognitive requirements” will provide competitive advantage to many organizations over the next 18 months.
Doug Ross helps lead Intelligent Automation at Capgemini / Sogeti in North America. Any questions or feedback? Please email firstname.lastname@example.org.