Rapid technological change represents a great deal of potential for the insurance industry, although finding the right applications will be a challenge, according to Deloitte's
AI-supported insurance could generate $4.7 billion in premiums annually by 2032, according to the report, and increased popularity of embedded insurance will top $722 billion in premiums globally by 2030.
Gen AI has many applications for insurance that are not yet being pursued, according to Karl Hersch and Sandee Suhrada, principals in Deloitte's insurance consulting practice. Hersch is a co-author of the outlook report and Suhrada contributed to the report.
Underwriting applications for Gen AI are already getting scrutiny from state insurance regulators, concerned about its potential for bias in decisions about who is covered and to what extent.
"How do you use it to help you underwrite better, more smartly, and find better risks that are more profitable without being biased?" Hersch said.
To that end, insurers should stick to specific applications for Gen AI, Suhrada explained. "You do need to collect the proof points," she said. "Opportunity exists across the value chain, in how you design the product, and how you are underwriting. Agent enablement is one of the value propositions we are working on with an insurer, and it's around agent experience."
Gen AI solutions could increase the frequency of review of policies, according to Suhrada. Moments in policyholders' lives can trigger reviews, which open up new revenue streams for carriers.
Similarly, in embedded insurance, which is often offered in the travel industry, there are opportunities to find specific elements to cover, which would also be new revenue for carriers, according to Hersch.
"Right now, if you go buy a ticket on United or Delta, you're going to get Allianz or [another insurer], you're going to get the standard stock travel insurance," he said. "But if you can use Gen AI to customize that to say, well, you're flying here, you're going to Hawaii, and you're going to do X, Y and Z, we can cover this more finitely, more precisely."
For instance, a customer going hang gliding in Hawaii could be sold a temporary additional term life policy or a special injury policy for that single event. "You pick up through the embedded buying what's actually happening, and then use the technology to customize a point of sale micro policy," Hersch said.
Not everything might be a profitable enterprise, but getting data from industries with embedded insurance could yield opportunities.
The flow of information between carriers and distributors, known as licensing and appointments (LNA), once Gen AI-enabled, could make it possible for many agents to follow the claimant-challenger model with more targeted and specific intentions before data transformation is completed. "You can actually, as your model is coming up with the right answers for you, have the challenger model checking the results or outcomes and look into the biases," Suhrada said.