As question answering (QA) systems like Watson mature, let's consider how providing access to insurance knowledge in an automated fashion can change the insurance landscape. A key element of this is the fact that the world operates in an unstructured way. People have been trained over the past 50 years by the information systems we use to structure our data. We fill in blanks, color in the circle, use zip codes, routing numbers, and all the rest to get our computers to serve us better. Basically, this has worked and provided us with the wide spectrum of tools we use today.
However, the future is happening. Information technology has progressed and continues to challenge our assumptions about how it can work. Today unstructured information and data can be used by information systems to perform an increasing number of tasks. The insurance industry needs to keep a collective open mind about how these types of intelligent systems can and will do. Watson's performance when playing Jeopardy! demonstrates a more general range of knowledge, but the natural-language (i.e., everyday language) aspect combined with a specific set of information like claims, underwriting, or products, will be game altering for carriers.
For instance, the trend toward greater use of self-service to control costs for claims and policy information updating (servicing) has been throttled by the quality of the interactions. A Watson-like QA system could support a given self-service interaction regardless of the mechanism used (face-to-face, voice, IM, text, hard copy, Web site, etc.) because it would provide a consistent source of information and knowledge.
For a "live" interaction, the QA system may have been consulted by the claims adjuster or contact center rep before or during the meeting—it would be one of those positive techniques where the contact person says, "Mr. Customer, I don't know the answer to that, but I know how to get you an answer right now…" The use of a QA mechanism would provide the ability to augment a person's knowledge (and that has training/education implications) while at the same time serving the customer more quickly and consistently.
Keep in mind that Watson represents an early version of the QA systems for which we are projecting use cases. In each "case," a QA system would have to have loaded a broad set of data and information, and possibly even have access to more real time data such as Google or a news feed. Another way QA systems will impact the insurance industry will be the demand for the creation and maintenance of the knowledge stores themselves for areas such as specialty underwriting, risk management, geographically specific underwriting, or high-value claims. The carriers that have the expertise in a given area can maintain the knowledge and sell it as well, but at the risk of not having much else to sell as time goes on.
So as to the question of where a Watson type system could be used in the insurance industry goes back to the problems of getting accurate and verifiable answers to questions that are control points in a business process. One such area is regulatory and legal compliance. A QA system could be loaded with the text and figures of regulations across geographic boundaries. In the U.S., that would be various states. For carriers that are more globally oriented, their coverages may cross national boundaries as well.
An enterprise risk management (ERM) executive may need to know how proposed regulations impact the company from a risk and pricing perspective. Using a QA system, she can ask the questions that will prioritize where she directs her team to focus their efforts. This may be especially useful and efficient as more "principles-based" insurance regulations are enacted around the world. Even in the existing regulatory scheme, the ability to pose queries to a question answering system and get answers with a very high confidence level could potentially save insurers millions of dollars.
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.