Insurers test advanced AI in claims processing, mindful of regulation
The insurance industry has long embraced various forms of artificial intelligence to bolster claims handling, fraud detection and pricing. A pivotal time has arrived, however, as generative AI and large language models like ChatGPT are changing insurance processes, particularly in claims. From enhancing accuracy and speed to improving customer experiences, AI’s integration into claims processing has enabled carriers to address inefficiencies and inaccuracies seen in the more traditional, manual claims handling, industry experts and practitioners say.
“Insurers have been using forms of AI in their business for several years now. They’ve been using models in claims to support fraud detection. They’ve been using models in pricing and they’ve been using models in marketing,” said Chris Raimondo, EY Americas insurance technology leader. “I think where the inflection point came is the introduction of generative AI, and its ability to really produce new content that was of high quality in a human-like manner, whether that’s conversational, whether that’s written or even images. And I think that’s been a game changer for the sector, because insurance is a content-rich industry. It’s filled with documentation. It’s filled with fragmented manual processes, and it’s filled with data as well.”
Some insurance companies are testing traditional forms of AI.
At Builders Insurance, “we’re about a year and a half into our journey with AI… But we really wanted to dip our toe into AI and see what to do,” said Kenneth Bunn, vice president of claims, who spoke at a recent webinar hosted by Gradient AI. “So we looked at several different areas, such as first notice of loss, direct assignment, and optical character recognition in document processing, and we’re excited about the opportunities for straight-through processing. When we really stepped back and looked at what we were going to be able to do in the quickest amount of time most efficiently, we landed on a risk-ranking model, as well as a reserve adequacy model, for work comp.”
Gradient AI’s software-as-a-service platform offers artificial intelligence tools that predict underwriting and claim risks, and reduce time spent in claims processing and quoting. Signal Mutual has adapted Gradient AI’s solutions into its claims framework, as well, for a number of different functions, such as triage.
“This has really allowed us to focus appropriate teams and claim handling protocols, depending on the level of risk,” said Ann Latimer, Signal Mutual’s senior vice president, head of claims operations. “We’re also starting to use it to assist in setting reserves, but especially in the question of when there’s a large discrepancy between the adjusters’ view of the reserve and the model’s view of the reserve.”
As executives embrace the potential of generative AI, they are also mindful of the data challenges that come with its implementation.
“This technology is completely dependent on data. AI technology, generative AI or other types of AI – it’s only going to be as good as the quality of the data that’s actually being input through the models,” said Raimondo. “Insurers still have a lot of data challenges in terms of being able to source the data that they need, confirm the accuracy of that data, and also the comprehensiveness of that data. They’re going to have to do a lot of work in parallel to getting comfortable with the technology to also get their data in a better state to really be able to take advantage of it.”
Raimondo said the introduction of AI technologies into the claims process adds an additional layer of complexity from a regulatory standpoint. State commissioners have issued guidelines for insurers’ use of AI to protect consumers and ensure fair and transparent operations. While these regulations are essential for safeguarding the interests of policyholders, they can sometimes make using AI in insurance challenging.
“It’s a highly regulated industry,” Raimondo said. “The regulation is fragmented as well, because it is state-based through departments of insurance. There’s some hesitation to get too far ahead of the curve, because of the regulatory nature of the industry and the potential for bias and reputational risk that they don’t want to introduce into their companies. So that’s why a human being involved throughout this is very critical in the early days of utilizing the tech.”
Travelers uses its Code of Business Conduct and Ethics to guide its use of AI and use of data, for instance by using stronger governance controls on third-party data used to enhance Travelers’ internal data.
“We’ve been using AI within claims for 15 years. We have an entire suite of machine learning models that are used all across the claim handling process that help with triage and assignment, fraud and identification and early severity detection,” said Erik Roen, senior vice president and CIO of claim technology and claim business intelligence and analytics at Travelers.
Roen also highlighted Travelers’ Responsible Artificial Intelligence Framework, which serves as part of Travelers’ governance framework.
The company tries to follow three basic rules or laws, he said.
“First and foremost, we want to make sure we pay what we owe on every single claim to fulfill that promise,” Roen said. “Second, we want to provide a great experience to that customer, that claimant. And then the third law is that we want to make sure we handle it as efficiently as possible, internally, as long as we don’t make that experience worse for the customer and we’re not not paying what we owe. Those three laws, in that order, really are what guide us in everything that we do, and how we’re changing the claim process, whether we use AI or generative AI or not.”
In the future, Raimondo predicts that generative AI will become an even more powerful tool in use cases beyond customer-centric roles.
“Think about all the different knowledge workers in an insurance carrier that are specialized,” he said. “You have underwriters, claims adjusters, actuaries… and I think that’s where you’re going to see that CoPilot or co-bot introduction, where instead of having human assistants pulling together volumes of data, these knowledge workers are going to use CoPilot [an automated messaging and sales prospecting tool from CoPilot AI] to assist them with their their tasks, whether that’s underwriting or adjudicating a claim. “And you’ll see this pivot from them not only just capturing content, but also helping with the synthesis and interpretation of the content as well… Generative AI is going to be able to do that for the knowledge worker to make them more productive.”
