IBM wasted no time in parlaying the winning performance of its artificially intelligent Watson on the TV game show “Jeopardy!” into real business. Last week, it announced a deal to commercialize Watson’s capabilities by developing a system to help doctors and nurses identify the most likely diagnoses and treatment options. The first offering could be available in 18 to 24 months.
In addition to eyeing healthcare, IBM sees potential for Watson in insurance, where expert systems are already ubiquitous in underwriting.
In an interview with Insurance Journal, IBM’s insurance industry representative Jamie Bisker said there are several potential assignments in insurance for Watson.
Watson might help large insurance companies that are selling multiple products in multiple states or even countries deal with a myriad of regulations.
“So you could have, say, information for Indiana, Ohio, and Kentucky listed. Most of the insurance regulations are probably similar, but there are differences,” Bisker told Insurance Journal. “So a question and answering system such as Watson could be queried about, ‘I want to sell a term life policy in Kentucky, could I use the same forms that I developed in Ohio?’ A Watson type system could answer that question.”
Rather than being a threat to insurance agents and customer service representatives, an artificial intelligence (AI) wizard like Watson could become their best – and smartest– ally, helping them to not only answer their customers’ questions and but also learn about new products and sales scenarios, according to the IBM executive.
“I could actually envision an agent or an executive literally referring to the system in the presence of the consumer or the trainee, saying, ‘Hey, I don’t know this. I’m not positive about that. Let’s ask Watson. ‘Hey, Watson. What do you think?'”
Bisker thinks that fears of Watson replacing humans, while understandable, are unfounded, just as they were about the Internet.
“[T]here was a big kind of fear that there would be a disintermediation of agents when the Internet started. Clients would turn to the web and people wouldn’t need agents. The same thing happened, to some extent, with travel agents,” he said.
However, Bisker said, most people who use the Internet for insurance, use it to shop, then “ultimately… they need to talk to a person. Ultimately, a lot of the transaction happens people-to-people.”
Also, Bisker believes that while young consumers may be more comfortable than older buyers purchasing insurance online, these younger buyers will see more value in interacting with real agents as they age.
“Because even though they might know the answers they need, they like some reassurance that it’s the right thing to do. They still want to talk to that that face. They want to put a face to that product they’re buying and the services that they need to get,” Bisker said.
While Watson’s claim to fame is question-and-answer scenarios, insurance customers do not always know what questions to ask. They depend upon their agents to know not only the answers but also the questions.
In the world of artificial intelligence, there is not a lot of difference between asking the right questions and supplying the right answers, according to the IBM representative. The key is to involve the right people in building the database —including a company’s best and most experienced agents.
“[Y]ou can imagine that, when a company… would be building their dataset or their database that would be used by a question-and-answering system, they would say, ‘You know what? We need to get our best agents in here. We need Barbra from Indiana and we need Jane from Pennsylvania. We’ve got to get Joe from Texas. Tell us your stories.’ So they could literally be asked questions, ‘What was your toughest sale? How did you handle this? What are the questions people forget to ask?”
Using Watson, a company can build a knowledge set using its most experienced people and then the knowledge these people bring can be kept forever and made available to people throughout the company.
IBM says the technology to employ Watson-like AI in business already exists; the challenge now is to build the customized language and datasets for specific industries like insurance. Bisker sees this happening in insurance within three to five years.
While Watson might not be looking to replace humans, in several respects it does offer advantages over mere mortals. “It wouldn’t be tired; it wouldn’t have an emotional bias, and it couldn’t be bribed,” Bisker said.
Watson in Insurance Podcast: Listen to Insurance Journal’s complete interview with IBM’s insurance industry representative Jamie Bisker.