How to avoid the biggest mistake insurers are making with AI

• Some brokerage functions should not be handed off to AI
• Check AI-generated materials before presenting to clients
• Most insurance products require agent interactions

Insurers should think more about relying on AI, or they could be making a huge mistake, distribution executives say.

When making operational decisions, "you have to validate all the time that the information that we are utilizing is actually the best information," said Tom Zacharopoulos, head of distribution at Westfield Insurance, speaking in a recent webinar. 

Testing AI systems before implementation is imperative, said Nicholas Davis, senior vice president and national distribution leader at Chubb.

"Launching before testing properly is a mistake any organization could make. You can't go back," he said. "It's important to be thoughtful on change management and the culture of the organization, whether it's training and the time that takes, or having the resources and ability to answer questions with the technology."

Oversight and governance can help ensure AI data output is correct, Davis added. "You have to be mindful of how quick you move everything to a new system," he said.

Some insurance broker functions still cannot be handed off to AI, said Nabeel Tanveer, WTW's head of affinity and programs for the U.S. 

"Recently, I came across an AI-based insurance broker. There's the tedious part of being an insurance broker: policy check-in, the billing, the maintenance. You don't need an insurance license to do that duty. So AI is going to dominate that space," he said. "Getting appointments is going to be a challenge. Negotiations between the carriers and the broker will always be like this. I don't think that's ever going to be handed off to an AI agent."

Tanveer's affinity and programs department, working with companies to provide insurance to their customers, is cautious about going to clients with any material produced by AI. 

"Someone who has experience in the space can then figure out this variable is wrong, because AI is making assumptions like a human, and some of those assumptions are incorrect," he said. "Organizations are going to have to start implementing in a much more stringent way to make sure that output is actually correct."

While insurers will use AI more often for simpler products such as pet insurance, insurance brokerages, even at middle market level, will still require agent interactions, according to Tanveer. 

"From an agency efficiency standpoint, we're trying to use AI where we can to have an accelerated pace of doing policy analysis, comparing policies or finding flaws in a competitor's policy," he said. "That's where we're going to be using more and more AI tools internally to help us."

Westfield intends to proceed slowly, Zacharopoulous said. "We are in the early stages, and in the early stages, there's going to be testing, there's going to be learning, it's trust and verify," he said. "You can't just jump into the deep end of the pool."