As insurers consider what to do with generative AI technology, the first step may be figuring out what applications have practical value.
"It could be claims, processing in marketing – customer service is a very low hanging fruit – fraud detection, and underwriting eventually," said Andrew Schwartz, analyst, P&C insurance, at Celent. "There are a lot of areas that are candidates for utilizing these tools to one's benefit. It's first important to identify where we think this can actually make an impact for an organization. They don't have unlimited resources. Focus on actually targeting a place where we want to start out and we think there can be a huge impact."
Gen Re, the P&C, life and health reinsurer that is part of Berkshire Hathaway, has been considering
"We're emphasizing self-determination, while making clear the guardrails," he said. Schmid sees Gen AI as a "decision support tool" for insurance, in underwriting and claims, rather than a way to automate decision making.
"It's really all about workflow and inserting components that are AI assisted," he said. "For instance, as underwriting submissions come in an email with attachments, it's important that we have a tool that extracts that information and classifies that information in preparation for decision making. AI will play a role in various components of this workflow, which means the workflow has to be highly modular."
Low-code workflow tools, paired with Gen AI, allow creation of modular applications and modular workflows that can integrate well with other operational tools, while retaining flexibility, according to Schmid.
"You can easily reverse certain decisions that you have made, because this is a technology that is still evolving at a rather rapid pace," he said. "You may make a decision for a certain AI system for automation of a certain step that you might want to change later on. The aspect of reversibility which supports exploration, is quite important in this, in such a fast changing technology landscape."
Productivity and operational efficiency is one of several possibilities for Gen AI, states Rima Safari, a partner in the insurance practice at PwC. She points to claims operations in particular.
"Claims is the perfect example, where there's so much data being ingested," she said. "Then you need to classify and prioritize the claims and triage them and then you also have to make the adjudication decision, of 'Is this person eligible for the claim or not? And what would be the payment for it?' That's a great area where there's a lot of manual activity involved. But there's a lot of unstructured data involved and you could transform claims operations with Gen AI."
In customer service, Gen AI cannot fully replace human interaction in insurance, Safari said. "Gen AI just makes that human's job a lot easier so that they can focus on the customer," she said. "They can spend more time handling the customer and doing critical thinking around the decisions that need to be made, as opposed to collecting information, manually transcribing the call and entering it into the system."
When insurers consider where to use Gen AI, they should look for functions that do not require much integration of Gen AI into their system, or just very simple integrations, according to Safari. The complexity of using Gen AI with end-to-end automation for functions like claims or underwriting makes it a bad place to start, she added.
Safari and colleagues studying Gen AI identified functions such as summarizing data or translating information in different languages, because they contain patterns where Gen AI's language characteristics can be useful. "It's much better to build on these reusable patterns as opposed to starting from scratch, and doing each use case at a time," Safari said.
Gen AI is designed to draw from large language models (LLMs) that use natural language processing, itself a part of AI, but it goes beyond what previous types of AI can do, as Schwartz of Celent explained.
"This technology transforms our interaction with machines where we're able to use language as a prompt to generate innovative outputs," he said. "That demonstrates a sophisticated understanding of manipulation of the human language, and paves the way for more intuitive, seemingly human-like interactions."
Anything Gen AI can do for insurers' internal functions or customer services is, like many other operational functions, going to depend on good data.
"You need to have a modern data ecosystem that's leveraging your own internal data and external data, leveraging data sharing platforms and data marketplaces," said Cindy MacFarlane, global insurance, AI and data leader at Deloitte. "There's just so much opportunity there. But it'll be a little while before that happens."
