Generative AI in Insurance: Use Cases, Potential Benefits, and Risks

The (Generative) Elephant in the Room

ChatGPT launched on November 20, 2022, and was the fastest product to reach 100 million users in history, doing so in just under two months. Nearly a year later, the hype surrounding ChatGPT, and Generative AI more broadly has only accelerated. Nearly every industry imaginable is scrambling to explore, invest in, and understand the potential ramifications of this groundbreaking technology. Supported by billions of dollars in funding, the pace of innovation is staggering. Just last week, OpenAI announced the ability to create custom-GPT’s, which could be its App Store moment and spur an even greater rise in popularity akin to the early days of the iPhone.

Naturally, we’re paying attention to how this will impact the insurance industry. Nearly every conference, email newsletter, networking event, and investor summit has turned its attention to GenAI and the enthusiasm surrounding its potential doesn’t seem to be slowing down.

While exploring the potential of Generative AI is more than likely worth the time investment for carriers and startups alike, our unpopular opinion is that the near-term benefits are overhyped. We believe it will take years to see any dividend from a GenAI investment and carriers will be better off in 2024 if they continue to focus on solving core problems such as reducing premium leakage and nondisclosure. 

A Brief History of Generative AI

In this article, we will discuss the potential use cases for Generative AI in insurance. But first, let’s take a quick look at where this recent phenomenon came from. Generative AI, a subset of artificial intelligence, has recently exploded in popularity due to the virality of ChatGPT. It has shown incredible potential in various industries and is the fastest app to ever reach 100 million users. 

It should be noted, however, that this technology is not brand-new. Instances of Generative AI date back to the 1960s, in the early days of chatbots. In 2014, with the introduction of generative adversarial networks, or GANs, Generative AI took a major leap forward. GANs are a type of machine learning algorithm that enables generative AI to create seemingly authentic images, videos, and audio. Two additional advancements made the recent explosion in popularity possible – transformers and the large language models they enabled. 

TechTarget does a great job explaining these advancements. “Transformers are a type of machine learning that made it possible for researchers to train ever-larger models without having to label all of the data in advance. New models could thus be trained on billions of pages of text, resulting in answers with more depth. In addition, transformers unlocked a new notion called attention that enabled models to track the connections between words across pages, chapters, and books rather than just in individual sentences. And not just words: Transformers could also use their ability to track connections to analyze code, proteins, chemicals, and DNA.” 

This led to what is better known today as Large Language Models, or LLMs. LLMs are the base of apps like ChatGPT. We’re still in the very early days of building out the app layer and the interfaces we will use in the future. While the generative AI investment bubble continues to inflate, fueling the next generation of applications built for end-users, we wanted to take a look at the potential impact on the insurance industry. 

The Promise of Generative AI in Insurance

Generative AI has the potential to transform various aspects of the insurance industry, bringing efficiency and an improved customer experience. Here are some examples of how generative AI might be applied:

1. Summarizing Insurance Documents:
   • Generative AI can be employed to analyze and summarize complex insurance documents, policies, and contracts, such as an Attending Physician Statement (APS).
   • By extracting key information and providing concise summaries, it can enhance readability for both customers and internal staff, facilitating quicker decision-making.
2. Analyzing and Triaging Claims:
   • Generative AI can assist in the initial analysis and triage of insurance claims.
   • Automated systems can review claim documents, extract relevant details, and categorize claims based on complexity or urgency, streamlining the claim processing workflow.
3. Automated Content Creation:
   • Generative AI can potentially streamline the creation of complex insurance documents, policy agreements, and even marketing materials. By analyzing existing data, the technology might be able to generate coherent and tailored content, reducing manual work and improving turnaround times.
4. Streamlining Quotes:
   • Generative AI can optimize the process of generating insurance quotes.
   • By analyzing historical data, market trends, and customer information, it can provide accurate and competitive quotes quickly, improving the efficiency of the underwriting process.
5. Automating Claims and Settlement Process:
   • Generative AI can play a crucial role in automating parts of the claims and settlement process.
   • Automated systems can assess claims against predefined criteria, flagging anomalies or potential fraud. They can also facilitate faster and more accurate settlement calculations.
6. Replying to Customer Queries:
   • Generative AI-powered chatbots can be integrated into customer service systems.
   • These chatbots can provide instant and accurate responses to customer queries by leveraging a vast knowledge base, ensuring consistency and efficiency in customer interactions.
   • Generative AI-driven chatbots and virtual assistants can provide instant responses to customer inquiries. This allows for 24/7 support and frees up human agents for more complex tasks.
7. Enabling GPT Chatbot Assistants:
   • GPT-powered chatbots can serve as virtual assistants, guiding customers through various processes, such as policy selection, claim submission, or status tracking.
   • They can understand natural language queries and provide personalized assistance, enhancing the overall customer experience.

Can Generative AI be Used in Insurance Underwriting?

The consensus is that yes, it can. We’re a little more skeptical. In life insurance, for example, carriers use machine learning for risk assessment in addition to a largely rules-based underwriting engine. This helps remove bias, allows for explainability, and appeases strict regulators who want to ensure a fair underwriting environment. P&C underwriting has a little more leeway, but most data scientists will be quick to tell you that there is an important distinction between machine learning and true artificial intelligence.

• AI is a broad term that refers to systems or machines that mimic human intelligence. AI deals with making machines simulate human cognitive abilities and actions without human assistance. AI can sense, reason, act, and adapt.

While we’re getting closer to AI in some respects, we don’t feel that regulators will allow true AI to be used in underwriting any time soon, if ever.

Despite our feelings, here are some ways we’re hearing that it could potentially help in underwriting:

• Personalized Policies: With the capacity to process large volumes of customer data, generative AI might be able to create highly customized insurance policies that cater to individual needs. This level of personalization enhances customer satisfaction and engagement.
• Risk Modeling and Scenario Planning: Generative AI can simulate numerous scenarios based on historical and real-time data, assisting insurers in developing more accurate risk models. This allows for proactive adjustments to underwriting strategies, enabling insurers to stay ahead of emerging risks.
• Fraud Detection and Prevention: By analyzing patterns and detecting anomalies in large datasets, generative AI can enhance fraud detection and prevention. This technology can quickly spot irregularities in claims or behavior, helping insurers mitigate financial losses.
• Dynamic Pricing Models: Generative AI can contribute to the development of dynamic pricing models that adjust premiums based on evolving risk factors and market conditions. This flexibility allows insurers to adapt their pricing strategies in real-time, improving competitiveness and profitability.

Why Generative AI Benefits for Insurance Might Take Time

The deployment of generative AI in insurance, while holding the promise of increased efficiency and streamlined processes, is not without its downsides. One significant concern lies in the potential for mistakes to be compounded, leading to substantial and costly errors. For instance, in tasks such as summarizing an Attending Physician’s Statement (APS), even a minor error can have far-reaching and expensive consequences, resulting in millions in costs. Unlike human errors, which may be more forgivable in the eyes of policyholders and carriers alike, mistakes made by AI systems can face heightened scrutiny. We’re seeing a similar dynamic play out in the autonomous vehicle space. Striking the right balance between automation and human oversight is crucial to mitigating the risks associated with costly errors while leveraging the benefits of advanced technology in insurance operations.

While generative AI remains in vogue, our position is that it will take years for its widespread adoption in insurance, especially underwriting. Here are some other reasons behind the potential delay:

• Data Availability and Quality: Generative AI requires large amounts of high-quality data to generate accurate insights. Many insurers are still in the process of digitizing their data and improving its quality. Additionally, generative AI and LLMs have a long way to go before the application layer is vertically specific. 
• Complexity of Underwriting: Insurance underwriting involves intricate decision-making that considers a multitude of factors. Developing generative AI models that accurately replicate this decision-making process is a complex task. Carriers today are still very focused on explainability when dealing with advanced machine learning and often default to rule-based approaches to remain compliant. We struggle to see how carriers will make the leap to black box genAI systems anytime soon. 
• Regulatory Challenges: The insurance industry is heavily regulated, and deploying generative AI models will require thorough testing and regulatory approval.
• Complex Implementation: Integrating generative AI into existing insurance workflows can be intricate and resource-intensive. Proper training, data preparation, and system integration are necessary for effective implementation.
• Ethical and Fairness Concerns: Generative AI models must be carefully monitored to ensure they don’t inadvertently introduce biases or unfair practices into the underwriting process.

Our Suggestion

While exploring the potential of generative AI is likely worth the time investment, we believe the near-term benefits are overhyped. The biggest question carriers are trying to answer today is, “How do I create better digital experiences and sell more policies online without increasing risk?” 

With that said, we believe that carriers should focus on solutions that address these issues head-on and will have an immediate and long-lasting impact on profitability. For example, ForMotiv enables carriers to increase Placement and Bind Rates while reducing Premium Leakage and Nondisclosure. We do this by understanding the intent of the end user to drive the next-best action tailored to the individual. Leveraging real-time behavioral intent data, carriers are creating dynamic experiences that adapt to the purchase intent and risk profile of each individual user.

No magic black box, only explanation-based, out-of-the-box solutions purpose-built for insurance. Reach out if you’d like to see how some of the biggest carriers in the US are leveraging ForMotiv Solutions.