Navigating the Unknowns with Chatbots and Insurance

Insurers are in the risk business, but how prepared are they for the risks presented by generative AI and chatbots in insurance?

On this episode of The Insuring Cyber Podcast, Tim Johnson, partner at law firm Browne Jacobson, discussed how insurance might respond to the opportunities and risks presented by these technologies. He leads the firm’s commercial services practice for insurance clients. He works with a number of insurers and intermediaries on policy drafting and distribution, compliance and commercial issues.

“What insurers do at the moment is they understand the particular risk, they gather as much information as possible, and they know the sector in which that risk operates,” he said. “So whilst they know some things might happen that could result in a claim coming, and they’re largely outside their control, ultimately they are known risks. Underwriters go into those policies, they write those risks, with their eyes open because they’ve got years of actuarial data for that particular sector and those kind of risks.”

With AI chatbots, that’s not necessarily the case as the technology is so new, he said.

“By throwing AI into the mix at the moment, we just don’t really know how accurate AI is,” he said. “And so for an underwriter, it’s essentially adding an unknown into the mix.”

This is leading to some hesitancy among insurers, he added, because although they’re in the risk business, they’re generally not prepared to take additional risk that they’re unable to quantify.

“I think hesitancy is a good thing for a number of reasons,” he said. “Clearly, there’s huge potential. Underwriters are naturally hesitant … because we just don’t know what we don’t know yet in terms of the potential impact and inaccuracy of AI. So that’s why I think there is hesitancy, and for those reasons, I think it’s sensible for insurers and for underwriters to be cautious and hesitant about placing too much reliance on AI.”

However, Johnson explained that he doesn’t share the perspective that insurers are behind other industries in innovating and moving forward with technology. He believes the industry will get there with AI as well.

“If we think about the innovations we’ve seen in recent years, we’ve seen the use of telematics. We’ve seen parametric contracts. We have health insurance products that track health data using smart watches and incentivize customers to make positive risk decisions,” he said. “We’ve had the use of big data in insurance for years, so generally, insurers are innovative in my view, in terms of both the design and the distribution of their products.”

There is one caveat, however.

“All of those examples I’ve just given are areas where that innovation effectively gives insurers additional certainty, gives them more information, or gives them some objective measures upon which they can assess a risk, whether it’s through health data or how an individual is driving a vehicle,” he said.

With AI, that additional element of certainty doesn’t yet exist, he said.

“Where we are at the moment with AI is different. I think AI at the moment, rather than adding clarity, it can add to uncertainty for the reasons I’ve discussed and that we just don’t know how accurate it is yet,” he said. “And so that’s why notwithstanding the sort of push forward to be as innovative as possible, in my mind, insurers aren’t using AI as much as other sectors.”

He does see the potential though, particularly in helping to draft and clarify policy wordings. In fact, he and his team did a preliminary experiment in which they asked ChatGPT to draft a high-net-worth household insurance policy within certain parameters.

“We did it just as a bit of fun, really. It was more fun than you think,” he said. “We wanted to sort of say, well, how good is it? Because, you know, you read about so much with AI and ChatGPT, and we thought, can it actually draft a contract? So we gave it some parameters and asked it to produce a contract, and it produced a decent sort of first draft. That’s where I could see it being quite useful in that, within literally seconds, it can produce a first draft document of several pages, which is a decent base upon which a professional can then complete the job.”

Johnson said while at first blush, the document did look and feel like an insurance policy with clearly identifiable insuring clauses and exclusions, there were a number of problems that would impact its legal effectiveness.

“I think where we are at the moment is it can create first drafts, which then someone who knows what they’re doing and is given access to that draft can fix it,” he said. “I think that’s sort of where we are in terms of document generation at the moment.”

Johnson expects that AI chatbots will be utilized more in insurance as underwriters become more comfortable with the technology, but he still urged caution.

“I suppose my final note would be to continue to exercise caution for all of the reasons we’ve discussed,” he said. “I don’t think just plowing forward and applying it to everything is the way. I think we’re a long way off that, but equally, just keep as close an eye on the developments as possible because it’s inevitable that it’s on the way and we’re going to see more of it.”

Check out the rest of the episode to hear what else Johnson had to say, and be sure to check back for new episodes of The Insuring Cyber Podcast, publishing every other Wednesday along with the Insuring Cyber newsletter. Thanks for listening.