Will ChatGPT supercharge insurance AI adoption?

The upstart ChatGPT heralded an advent of conversational-AI platforms that can passably converse with humans based on a wide range of inputs. In addition to ChatGPT, which is made by the Microsoft-backed nonprofit OpenAI, other big tech companies are getting into the game with competing projects from Google (Bard) and Facebook (LLaMA). 

The rise of AI to kitchen-table prominence raises a question: Are insurance companies, which have been transforming digitally for years, ready to invest further in large language models and turn their precious customer relationships over to a chatbot?

Forrester Principal Analyst Indranil Bandyopadhyay says that a big AI revolution in insurance isn’t going to happen overnight.

“I don’t see the majority of the insurance industry going and jumping into these emerging technologies. It will take some time,” Bandyopadhyay says. “As AI continues to advance rapidly, I think the differentiation between humans and machines will become less clear in some areas. However, there are some fundamental differences, emotional intelligence and empathy is something which I don’t see in machines, even in the medium-term. … Humans will always ask the questions, and machines will help answer it, to a certain extent.”

How it works?

The Travelers Institute hosted a webinar, Making Sense of Emerging AI Capabilities like ChatGPT, March 1. Speaking on the topic, Girish Modgil, PhD., VP of automation and artificial intelligence accelerator, enterprise data & analytics at Travelers, said: “These large language models have been around for a few years. ChatGPT… does provide extremely articulate answers to questions on a variety of topics. … But although the answers seem persuasive and seemingly accurate, they can be incorrect and because of the data that is used to train these models: They have inherent bias.”

But Mano Mannoochahr, SVP, chief data & analytics officer at Travelers, added that while large language models have been around for years, the generative aspects of technology like ChatGPT are unique. 

“This is just the beginning for tools like ChatGPT and tools that generate images, they will get integrated into other productivity tools that we use in some shape or form,” he said. “It will have the potential to transform many parts of how we work over the course of the next decade.”

At this point, though, he cautioned, “if you’re looking for specific facts and what is true and false, those things are problematic with this technology right now.” 

Even OpenAI CEO Sam Altman “said that it would be a mistake to be relying on the tool for anything important,” Modgil added.

Insurtech examples
Potential AI incursions into the insurance industry aren’t limited to ChatGPT. Agency Revolution, an insurance marketing software provider for agents, recently launched a social media app that uses ChatGPT and Vestorly, a content creation platform, to deliver content for agents using AI-generated posts. 

Jason Walker, president of Agency Revolution, says agents are able to use the app to create social captions and help with first drafts of content. Walker says agents are also using ChatGPT for writing sales scripts.

“We want to complement the way that agents have built their business,” Walker says. “A lot of chatbots were introduced just a couple of years ago, but the consensus was, they’re not quite relevant yet to what I’m trying to do with my agency, which is just be human before the human gets on the line.”

Roots Automation began shopping around its ‘digital coworkers’ to the insurance industry last year. Despite conjuring images of a robot in the cubicle next door, the product actually is simply an AI software that can perform a number of insurance tasks, as a solution to talent scarcity and an alternative to robotic process automation.

The company has worked with Berkshire Hathaway, Progressive, Liberty Mutual, amongst other insurance companies to learn and build dictionaries around insurance language. 

“Working with all these big insurance carriers we built this really interesting knowledge base that becomes the AI,” says Chaz Perera, CEO and co-founder of Roots Automation, in an interview with Digital Insurance from February. “It doesn’t just stop there because what the ‘digital coworkers’ are doing is every time they come across something that they’ve never seen before, they will actually ask an underwriter at an insurance company, that they think is the expert, on this particular problem. How do I solve this? So, they’re getting smarter.”

Impact on jobs
Several attendees to the Travelers webinar expressed concern about the impact of AI on their job security. Mannoochahr responded: “That’s an extremely strong reaction to this new tech. We can’t rush to remove people from something that’s an extremely human experience. The expert – or the person with the domain knowledge – will still need to verify the output of these tools.”

It’s repetitive tasks that show promise for AI, Modgil added: “We wouldn’t want someone looking through 20,000 images of houses after tornadoes, right? I think it is really applying and using technology to make our lives better, not worse.”

Because ChatGPT is free in addition to having a pay tier, experts foresee the technology being used in productivity tools helping people to write things like emails and job descriptions. However, it is a black box, which means its answers are not auditable and its inputs are opaque to readers. 

In contrast, Jeff Tyler, head of product and data science engineer at Two Sigma IQ, which provides some AI capabilities to the insurance industry, told Digital Insurance that TSIQ puts a large emphasis on explainability. 

“Being able to link back to the source data, being able to show how we figured something out or how we extracted something and build confidence with our users. I want to be able to tell someone how I got that answer,” Tyler says. “An AI system doesn’t have accountability to truth so there is some verification validity to what we need to do with the outcomes. It’s extremely powerful, but it’s a tool. It’s not a replacement for everything a human can make a decision on, it will get better, and we’ve seen that.”

Perhaps an insurance-specific version of a generative-AI chatbot could emerge, Mannoochahr posited.

“I’m sure there’s a flavor of this that you’re going to see from carriers in some shape or form,” he said. “There is endless potential from a technology perspective.”