AI Advancements Transform Insurance: Insights from A.M. Best Briefing

In "This Time It's Different for AI in Insurance," an article by Richard Banks for A.M. Best, Edin Imsirovic, director at A.M. Best, explains why today's advancements in artificial intelligence (AI) differ from previous cycles of hype. Speaking at A.M. Best's Europe Insurance Market and Methodology Briefings in London, Mr. Imsirovic identified improvements in computing power and data availability as the catalysts driving today's AI advancements, noting that these innovations have created a "critical mass" necessary for AI to flourish in insurance.

According to Mr. Imsirovic, generative AI stands out among recent AI developments, allowing insurers to transform unstructured data into actionable insights. He traces generative AI's foundation to a 2017 paper, "Attention Is All You Need," which introduced transformers, a key component in today's large language models. The impact became clear in 2022 with ChatGPT's launch, illustrating the rapid progression from theoretical frameworks to practical applications. "The AI is getting much better and much more useful and much more sophisticated," Mr. Imsirovic explained. For example, running a million tokens on ChatGPT once cost over $40, a figure that has since dropped to just over $4.

When discussing applications within insurance, Mr. Imsirovic pointed to customer service, claims processing, and underwriting as areas ripe for transformation. For instance, generative AI can summarize claims information, helping adjusters concentrate on complex cases. Customer service, often hindered by basic chatbots in the past, now benefits from AI-driven conversational interfaces capable of providing clear, personalized responses. These advancements make it easier for customers to understand policy details, which historically required a human touch due to the intricacies of insurance language.

Yet, Mr. Imsirovic emphasized that most AI in insurance is still rooted in nongenerative, traditional machine learning applications. These established methods have found practical use in simple lines of business, such as personal auto or small commercial insurance, where tasks can be automated without navigating extensive policy complexity. However, for insurers to fully capitalize on AI's potential—whether through generative models or traditional machine learning—he stressed the necessity of an efficient information technology infrastructure. Insurers lacking systems that allow data to flow smoothly across platforms may struggle to unlock the full benefits of AI, as fragmented data systems can hinder the accuracy and applicability of AI-driven insights.

Copyright © 2024 by A.M. Best Rating Services, Inc. and/or its affiliates. ALL RIGHTS RESERVED.