Insurers work to stock data to feed AI

Register now

Artificial intelligence is one of the most talked-about technology topics in insurance right now, as Insurtechs and incumbent carriers alike reveal excitement for the technology's potential in underwriting, customer service and claims.

However, the relatively new technology does have some shortcomings, experts say. A lack of historical data, insurers’ legacy systems and AI’s inability to decipher natural language beyond basic speech has stymied the technology’s growth. And for all of AI's industry-wide adoption, carriers appear to only trust machine learning to carry out the same tasks as web portals or mobile apps.

“It’s just a new channel,” said Jeff Goldberg, SVP of research and consulting at Novarica. “Anything [carriers] put into other channels that customers are comfortable doing, and submitting information for, they have included in AI chatbots and virtual assistants.”

The potential is there for AI to eventually make more complicated decisions on workers’ compensation and large commercial property claims, Goldberg says. But for now, insurance AI is in its infancy and not something companies go out on a limb to publicize.

“Every time you open up a new channel for interaction, it’s not just about the channel. It’s all the back-end systems supporting it," said Goldberg. “That’s not a limitation of chatbots. It’s just a reality of how the AI fits on top of your data, APIs and web services.”

AI capabilities are being adopted by carriers in several ways across the industry. GEICO launched its virtual assistant “Kate” for its mobile app in January. Competitors Liberty Mutual, Allstate and Progressive have also launched voice-enabled applications on Amazon Echo or Google Home in the last 10 months. Meanwhile, Aetna, among AI’s early adopters in insurance, rolled out its virtual assistant, “Ann,” in 2012 to help customers navigate its website. RBC Insurance, part of Royal Bank of Canada, released an AI chatbot named Arbie in 2011 for the same purpose.

Of course, a slew of startups are all in on AI as well. Uniphore, a speech-recognition solutions company based in India, was last week named the winner of MetLife LumenLab’s inaugural startup accelerator. Insurtechs such as Claimbot, Avaamo, Insurify and Next Insurance have also surfaced over the last year with their own AI platforms. The latter two partnered with Facebook to sell coverage using its messenger app in March.

“Next insurance is early on in its journey to use AI for better decision-making,” said Sofya Pogreb, COO of Next, a managing general agent for small business insurance founded in 2016. “The company is 15% there.”

Next’s biggest challenge so far is finding enough claims data to supercharge its AI, Pogreb says. Unlike banking, insurance lacks a central repository of information for companies to tap into. The MGA that distributes small-business coverage for Munich Re, Nationwide and Liberty Mutual is now working with vendors like LexisNexis, as well as insurance partners, to gain access to customer information.

“It takes a long time for a company to build out its data set,” said Pogreb. “On average, carriers get one claim for every 1600 customers. So if you want 1 million claims, you need to sell 1.6 billion policies.”

Based in Palo Alto, California, Next sells 70% of its policies on mobile. It says it is gauging customer comfort in buying coverage through Facebook Messenger as well, which has had minimal impact to date, according to Pogreb.

Facebook is working to fix that. The social networking site released a new framework for developers on May 15 to improve the conversational abilities of chatbots. The goal is to teach machines to handle more natural conversation instead of short commands, TechRadar reports.

As it stands, 90% of online policy purchases using Next’s AI require zero human interaction. However call center agents are sometimes brought in to answer additional questions or confirm a bot’s legitimacy, it says.

“People do not speak in succinct sentences,” said Ben Case, lead cognitive solutions architect of IPsoft, an insutech offering a cognitive platform called “Amelia,” which automates IT functions. “We have to improve beyond natural language processing to understanding broader data points. Finding additional meta-data you can use to almost get proactive about the situation.”

AI’s future in insurance is not as limited as its current state. The consensus is the technology can one day ride shotgun to agents in distribution, be the first step in managing claims or first notice of loss and interpret data for underwriting. The opportunities are endless, insurers, developers and analysts agree.

“AI’s potential impact on insurance is humongous,” said Goldberg. “As long as there is a process in place for advancement, AI chatbots’ current limitations are not a problem necessarily.”

For reprint and licensing requests for this article, click here.