Traditional AI and GenAI: Transforming the Insurance Industry
January 9, 2024
By Paul Cataldo
Fueled by the power of artificial intelligence (AI) that now includes generative AI combined with traditional machine learning, the insurance industry is undergoing a profound shift. This shift goes beyond basic AI functionalities, diving deep into the realm of intelligent guided decision-making.
IBM once described cognitive computing as an advanced system capable of learning at scale, reasoning with purpose and engaging with humans in a natural and meaningful manner. This concept integrates various AI disciplines such as neural networks, machine learning, natural language processing, and contextual awareness, operating in a way that resembles human problem-solving. When this term was first coined, the idea of computers mimicking human reasoning and thinking seemed distant, but today we’re witnessing this become a reality.
As more applications leverage traditional and generative AI, these technologies will turbocharge their impact on society and business. The latest developments are truly astounding; we are witnessing AI advancements that resemble human abilities to read, understand and reason. What is even more exciting is that this is just the start. This moment in AI development is akin to the early 1980s when personal computers were just being adopted, foreshadowing even more transformative and profound changes ahead.
What does the power of generative AI look like? Consider Anthropic’s Claude, a next-generation AI assistant that can read and process the average novel of about 75,000 words in about one minute. With AI systems already delivering this kind of processing power, it isn’t surprising that a recent McKinsey report estimated that the combined impact of generative and traditional AI could unlock and add $17 to $25 trillion of value to the global economy.
In the insurance industry, this AI combination empowers insurers in ways never before possible. Insurers can not only underwrite risk more accurately, provide better customer service, and automate tasks, but now, with the help of generative AI, they can virtually have a conversation with their organizations’ vast, previously inaccessible knowledge bases. They can now consult decades of accumulated claims data, medical notes, and outcomes to improve decision-making in their underwriting and claims management operations.
Consider an insurer with 20, 30, or 40 years of accumulated claims data having detailed medical and adjuster’s notes, claim outcomes and associated costs. By leveraging traditional and generative AI applications, adjusters can engage with the institution’s entire knowledge base. This interaction allows them to better understand the direction a claim might take and to determine the next best actions to drive better outcomes. Likewise, this combination also empowers underwriters to make more informed decisions. Not only can they assess the predicted risk of writing a specific policy, but also gain deeper insight into the specific factors that are driving that policy’s risk up or down.
However, the effectiveness of AI in insurance hinges on access to extensive, high-quality data. In general, the more data an AI model is exposed to, the more accurate and effective it becomes. Industry data lakes, enriched with diverse data sets, are crucial for training AI models comprehensively.
At my company, we maintain an industry data lake with tens of millions of policies and claims observations. We have enriched this information with third-party data features including economic, health, geographic, and demographic information. Both generative AI large language models and traditional AI machine learning models can benefit from these vast data assets. They can learn from numerous instances, gaining extensive knowledge that improves their effectiveness across many different types of claims and policies.
Insurers that embrace these technologies today are poised to gain significant competitive advantages in the years to come. In underwriting, AI has had a demonstrative impact on key operating metrics and profitability. For example, an insurer leveraging AI decreased its loss ratio by 13%, reduced quote turnaround times by 80%, and experienced an 86% increase in direct written premium per employee.
On the claims side of the business, one of our customers reduced workers’ compensation claims costs by about 10%, which translated to more than $300 million in savings for the insurer’s clients.
The AI transformation in insurance isn’t just a technological evolution; it’s a paradigm shift in how the industry operates, and a precursor of what’s to come in 2024 and beyond. The industry is moving into an era that goes beyond automation to one where decisions are profoundly informed. As insurers continue to leverage AI’s cognitive advances, the future holds the promise of an insurance industry that operates with unprecedented intelligence, effectiveness, and efficiency, benefiting both insurers and customers alike.
Paul is a senior technology executive with over 20 years of experience, with a focus on leveraging big data to improve business processes and outcomes. His early work transforming a data science services company, DataSage, into a product company leveraging machine learning to build a customer analytics solution for Walmart, Amazon, Outpost, and other leading retailers, resulted in one of the first AI unicorns (exit value over $1 billion).
Paul also led the product marketing and management teams at IntelliVid, a video analytics company (acquired by Tyco) which developed a groundbreaking solution for loss prevention. He later served as vice president of marketing for OATSystems, an MIT spinout that pioneered IoT solutions for retail and manufacturing applications, prior to its acquisition by Checkpoint Systems. His recent work includes developing the AI business strategy for a major insurance carrier, leveraging machine learning to improve IT operations and the end user experience.
Paul holds a BA degree (cum laude) from the University of Massachusetts, and recently completed MIT Sloan School’s Artificial Intelligence: Implications for Business Strategy and Stanford University’s Machine Learning programs.