As insurers tap artificial intelligence to analyze risk and validate claims, they’re also deploying more of that technology in customer-facing applications. Chatbots, algorithms that enable intelligent conversations with humans via text or voice, are now being used to assist both customers and agents in many areas.

Chatbots can enable new efficiencies by assisting thousands of queries simultaneously, and as new technologies make bots more intelligent, they’re bound to offer more applications in the insurance industry in the coming years. Yet while chatbots can offer many benefits, insurers must also ensure they’re being supported with the right intelligence.

Enabling “Human-like” Conversations

A chatbot is simply an algorithm that can create a natural conversation with a human to answer questions and provide information, said Sharad Sachdev, managing director and analytics lead in Accenture’s Insurance Strategy Practice.

While insurers have been using IVR (Interactive Voice Response) at call centers since the ’90s, a convergence of new technologies are enabling far more complex conversations. Chatbots are increasingly being used in the insurance industry to support agents, and in customer-facing applications such as onboarding new clients and processing claims, said Sachdev.

Sharad Sachdev, managing director, Accenture

“Pretty much every insurance company is coming up with some version of a virtual assistant for claims or applications. And everyone is naming their bot something, it’s the new trend,” Sachdev said.

As these chatbots grow more sophisticated, companies and consumers are becoming more comfortable using them.

A survey by SnatchBot of 6,000 consumers found that more than 80 percent of users were satisfied with a chatbot experience. Researchers also say such messaging is growing as a medium for communications between corporations and consumers.

On the customer-facing end, one of the biggest advantages of a chatbot is that they are available 24 hours a day and don’t require human intervention. They can also “speak” with thousands of users and answer thousands of questions at the same time, improving the efficiency of communications.

In a complex environment like the insurance industry, there are a number of opportunities from a policy and claims standpoint to implement chatbots, said Alex Sun, President and CEO of Mitchell International.

“They are being used to give quotes on policies, provide account and billing information, and in the case of some insurtechs, process simple claims,” Sun said.

Complex Conversations with Customers and Agents

Chatbots are already being adopted by most big insurers. Allstate launched its virtual assistant ABIe (Allstate Business Insurance Expert) in 2015 to more efficiently answer questions from its 12,000 agents.

Mike Barton, division President of Allstate Business Insurance said at the time that ABIe was the company’s “precursor to cognitive computing on a shoestring.” The avatar-driven interface takes a rigid approach to understanding words and phrases and can accurately respond to questions in seconds.

Last year, GEICO launched “Kate,” a smartphone app that can communicate with customers through voice recognition and text. Kate is available around the clock to policyholders and enables a high level of self-service that can help with policy needs.

Pete Meoli, GEICO mobile and digital experience director, said that the technology has altered the way consumers interact with mobile devices.

GEICO’s chatbot virtual assistant has been programmed to communicate with policy holders on a deep level and in a natural tone. “We wanted her to be friendly with a natural interaction. She is always learning from our customers and will be an integral part of enhancing their experiences with GEICO,” Meoli said.

Many startups are also bringing new chatbot solutions to market. Maya, an artificial intelligence bot created by Lemonade, can accept applications, secure a policy for consumers in as little as 90 seconds and process claims in only three minutes.

Sriram Chakravarthy, Chief Technology Officer and co-founder of Avaamo, said conversational bots represent the “last-mile automation” for customer service.

“We believe to truly deliver the promise of conversational AI, fundamentally new technology has to be built to perform multi-turn conversations and execute judgement intensive tasks just like humans.” — Sriram Chakravarthy, Chief Technology Officer and co-founder, Avaamo

“We believe to truly deliver the promise of conversational AI, fundamentally new technology has to be built to perform multi-turn conversations and execute judgement intensive tasks just like humans,” Chakravarthy said.

Avaamo currently has more than two dozen customers in the insurance industry who are using conversational bots for various use-cases ranging from customer service to internal helpdesk. Bots are also being adopted for underwriting assistance, agent advisory services and for on-boarding assistance for HR teams.

Chatbot Success Hinges on Underlying Intelligence

While chatbots hold great promise for the industry, they are not without challenges. Poorly-designed bots that malfunction or don’t meet customers’ expectations can backfire and result in brand damage.

Insurers are advised to thoroughly test their bots before deploying and to track bot activity and incorporate feedback loops with customers. Insurers must also ensure their chatbots are compliant with regulatory requirements and have a high level of data protection, Chakravarthy said.

Despite the advances in pronunciation, listening skills and comprehension, accuracy can still be a problem with some bots, Chakravarthy said.

One best practice is to design the bot from the ground up to meet the end goals the insurer is seeking to accomplish. “The intent of the bot is not to be the next Siri or Alexa but to be specific to the domain and help solve customer challenges,” Chakravarthy said.

It’s worth noting that the more intelligent they try to make the chatbot, the “more risky it is,” Sachdev said. Many of Accenture’s clients are starting chatbot deployments with more simple conversations then moving to more complex ones once they have refined the technology.

The most successful chatbots are built not just on the natural language layer, but a high level of intelligence underneath to field questions and find information. Chatbots that force users to “spoon feed” information don’t perform well, Sachdev said.

Despite the advances and the more “human-like” conversational abilities of these algorithms, customers don’t want long conversations with automated applications.

“Machines will never have that level of emotionality in their response. The last thing you want is a 300-word response from a machine. You start with simpler situations,” Sachdev said.

Finally, insurers shouldn’t force consumers into using chatbots, and they should only be used as one optional means of communication in conjunction with mobile apps, websites or human agents.

“Every customer behaves differently, this should be one of the channels being offered and the customer should decide which is best for them,” Sachdev said. &