RPA meets cognitive capture
A “good” invoice is characterized by matching the purchase order amount with the shipping receipt and approved terms. Good invoices can be approved for payment without a human office worker intervention. Invoices in a somewhat different format from a new vendor or the result of redesign are “trained” to be automatically handled even if the field information is in different locations.
Those automated invoice processing systems have become worth more than $500 million. The challenge is to handle more exception items—such as those with partial shipments or damaged shipments and those that have taken a discount when they should not have or that are otherwise out of compliance with their contract. Another challenge are those invoices that are in a different currency or use a different language. They are the areas that consume white-collar clerical resources. Some can be handled through more rules sets and workflow, but others need to use more advanced cognition capabilities.
Cognitive capture is key
Advanced capture solutions are integrating tightly with the business application. They are validating and making decisions based on a combination of captured inputs and business process rules. As RPA gets expanded into that area, it gains power and capability. Pega, a vendor of BPM with RPA, has recognized that, but Pega is not strong when it comes to capture from paper and PDF image files. IBM with Datacap has formed an alliance with Automation Anywhere, and when you add elements of Watson, IBM can have a formidable offering.
Much electronic communication within business has leveraged standardized PDF attachments to eliminate the cost of mail. RPA users, who have eliminated the low-hanging fruit of repetitive key entry, are starting to want to implement more varied inputs and have the systems decide on the disposition. It is the same problem of classification and extraction that the transaction capture industry has been facing, and thus cognitive capture is key.
Kofax with Kapow RPA has been targeting new onboarding applications—particularly in the banking area, but does not seem to have integrated it with its KTA transactional capture solution.
In today’s e-world, the companies that have a sustainable competitive advantage are those that can understand the variety of “big data” inputs that they receive and react to them as quickly as possible— preferably faster than the competition.
We have believed for some time that the key to the value of input understanding and conversion technologies including RPA is in understanding and conversion as close to the point of impact as possible. Particularly interesting is customer engagement management and digital engagement management. To better engage with customers, it is important to understand and route interactions to the person who can best handle the interaction.
Chat has evolved into a technology where we often are uncertain whether we are interacting with a person or a “bot.” Those technologies are using NLP and other AI techniques to understand what the customer wants and attempt to satisfy it quickly and efficiently. It often requires access and reaction to the business systems with extraction of data. RPA does that. But it might require customers to pick up their phone and scan a document that needs to be interpreted—capture does that. It may also require language conversion or reference to a certain contract or correspondence that is in the realm of cognitive capture where advances are being made.
The worlds of RPA and cognitive capture must merge. Each area needs the other. To advance, RPA is in need of more intelligence, which cognitive capture can provide. We have estimated Cognitive Capture 2.0 services as a $30 billion services market by 2022. Conversion technologies including OCR, OMR, voice recognition, photographic understanding and extraction, as well as video understanding and extraction are key and will enable more effective RPA processes. Cognitive capture will help advance RPA to the next level.