Cognitive Computing and Unstructured Data – a Boon for Insurance?

As the intelligent computing market continues to mature, one key lesson learned is that it’s about more than cost takeout. By improving auditability and accuracy, Robotic Process Automation (RPA) solutions are enhancing governance and oversight of third-party relationships, a critical need in financial services. RPA is also helping industries such as retail gain a competitive edge by helping to analyse and predict consumer behavior and tailor goods and services accordingly.

In terms of what’s on the horizon, the linkage of RPA capabilities to cognitive “thinking” tools has the potential to further change the game. Consider: RPA solutions are really good at doing very specific things, but they require highly structured input and need to be trained explicitly. When an RPA tool encounters an exception to the rule it’s been taught, it’s stymied. As a result, RPA offers little help to insurance companies struggling to deal with mountains of backlogged data stored in claims, account updates and invoices. The reason is that the relevant data on these myriad forms is unstructured and variable – a policy number can be in one box in one form and in a different box in another. As a result, the paperwork doesn’t lend itself to specific, consistent and repeatable rules.

Enter the emerging field of cognitive computing, which has the capability to analyse variation, recognize patterns and “learn” from experience. Much like you or I scan a newspaper article for a few key words and can identify whether it’s from the sports or business page, a cognitive computing application can detect patterns and reach a decision.

Apply this capability to the tasks facing insurers and you’re on to something: cognitive solutions on the back end analyse the unstructured data contained in disparate forms, invoices and policy documents stored in legacy systems and extract the relevant bits. RPA solutions on the front end process that data 24×7 and transfer information to digital platforms.

The potential is there to conquer the mountains of legacy data that have bedeviled insurers for years.