In a panel at the recent Insurtech Insights USA event, Steven Watson, Chief Technology Officer at AXA XL, joined Tom Wilde (Indico CEO) to discuss the transition from the data era to the decision era in insurance. The conversation highlighted the evolution of AI and its impact on the industry, touching on lessons learned from past waves of AI adoption and the current state of generative AI. In this article, we’ll be exploring the key points from their discussion within the greater context of the insurance industry as a whole.
The Evolution of AI in the Insurance Industry
Steven Watson kicked off the discussion by reflecting on the 2018 wave of AI, marked by the rise of deep learning and a heightened focus on data strategy. Watson emphasized that despite the excitement around AI, the real challenge was not just in data science but in data engineering. “There’s a lot of talk about AI, but at the end of the day, it comes down to your data. That was true of the last wave. It’s true of this wave. Unless you’re looking to do very generic tasks, you need to make it relevant to you, to your company, to your users.” He was stressing the importance of having a robust data strategy to make AI meaningful for the enterprise.
This sentiment is highly relevant for Indico, a company that specializes in applying machine learning and AI to unstructured data specifically. We find that an emphasis on data strategy and the successful integration of structured and unstructured data is crucial for leveraging AI effectively.
The Role of Data Strategy in Insurance
Watson highlighted the transition from traditional big data technologies to more accessible, cloud-based solutions. This shift allowed companies to reduce the initial investment in data processing and storage, making it easier to adopt new technologies. However, he pointed out that the fundamental principles of data strategy remain unchanged—arguing that companies need to know where their data is, how to determine if the data is good, and how to ask insightful questions about that data in general.
Related content: From Data to Decisions: Lessons Learned from the first wave of AI
Lessons Learned and Future Challenges for Insurers
Reflecting on the lessons learned from the previous AI wave, Watson noted that a significant challenge was the time spent on finding and preparing data for AI initiatives. He mentioned that this issue persists today, particularly with the growing complexity of integrating unstructured data into AI models—the problem that Indico’s intelligent document processing (IDP) tools were built to solve. “The last wave was really heavily focused on ‘How do we use structured data to make predictions?’ And now with generative AI, there’s a lot bigger focus on unstructured data,” Watson stated.
Related content: The role of AI and ML in document processing with Chris Payne, EY Partner
Governance and Regulation in AI for Insurance
The panel also touched on the evolving landscape of governance and regulatory requirements in AI. Watson emphasized that governance often lags behind technological advancements, creating a complex environment for compliance. “Governance is always playing catch up with technology,” he remarked, highlighting the increasing number of controls and the complexity of compliance in the AI era.
We’re moving from the age of deterministic models, where an input produces the same output every time, to non-deterministic models where the same input can produce multiple outcomes. The challenge now is explaining to regulators how the data, prompt, and instructions you provide an AI model with can produce those various different outputs. This issue is especially pressing in insurance—or, as Wilde calls it, “The business of decisions.”
Ensuring that AI models are transparent, explainable, and compliant with industry regulations will be crucial for gaining trust and widespread adoption. We hope that continued dialogues like this one will contribute to forward momentum in both regulatory clarity and an awareness of key compliance issues.
Moving from Data to Decisions in Insurance
As the discussion progressed, Wilde steered the conversation towards the concept of moving from a data-centric to a decision-centric approach in insurance. He noted that insurance companies are fundamentally in the business of making decisions—whether it’s underwriting risks, pricing policies, or adjudicating claims. “At its core, an insurance company creates competitive advantage by its decision-making ability,” Wilde asserted.
Watson echoed this sentiment, emphasizing the potential of generative AI to enhance decision-making processes. He pointed out that while people are excited about using AI to automate mundane tasks, the most valuable application of the technology is in supporting decision-makers with better insights and reducing their cognitive load. “If we can use gen AI to assist the underwriters, I think that’s a great way to introduce it and help them understand where it can help them with the things they don’t like [doing],” Watson said.
Addressing Trust and Security Concerns in AI for Insurers
Wilde and Watson also discussed the critical issues of trust and security in deploying AI technologies. Watson highlighted the need to balance security with the benefits of AI, suggesting that companies should explore bringing AI capabilities to their data internally rather than sending data out to external platforms. “So, when the data is in our network, it is within our security perimeter… So one of the things that is interesting that we ought to be looking at is, ‘How do we bring these capabilities to the data that is in our network versus sending the data out of our network to someone else?’” he advised, underscoring the importance of maintaining control over data.
This approach is crucial for Indico, where we must ensure that our AI solutions are not only effective but also secure and trustworthy. Building robust security measures and transparent data practices should take first priority in application of new technology, as they are key to gaining the confidence of insurance companies and regulators alike.
Final Thoughts: AI and Machine Learning in Insurance
Throughout the panel, both Steven Watson and Tom Wilde provided valuable perspectives on the evolution of the insurance industry’s ability to inform decision making with data at the highest levels. Their insights highlight the ongoing challenges and opportunities companies face regarding AI, particularly generative AI, to enhance their decision-making processes. For Indico, these discussions underscore the importance of a robust data strategy, effective governance, and a focus on delivering decision-centric AI solutions. Overall, as the industry continues to evolve, Indico is well-positioned to help insurance companies navigate the complexities of AI and achieve greater efficiency.
Subscribe to our LinkedIn newsletter.
Frequently asked questions
- What are some concrete examples of how generative AI is currently being used to support decision-making in insurance companies? One concrete example of how generative AI is being used to support decision-making in insurance companies is in underwriting. Generative AI models can analyze vast amounts of unstructured data, such as medical records, social media posts, and customer feedback, to provide underwriters with more comprehensive insights. This allows for more accurate risk assessment and personalized policy pricing. Additionally, generative AI is being used to automate claims processing by analyzing claim documents, detecting fraud, and predicting claim outcomes, which helps expedite the claims settlement process.
- How are insurance companies addressing the technical challenges of integrating generative AI with their existing legacy systems? Insurance companies are addressing the technical challenges of integrating generative AI with their existing legacy systems through a phased approach. They start by identifying specific areas where AI can provide the most immediate benefits, such as claims processing or customer service. Companies then develop APIs and middleware to enable communication between AI models and legacy systems. This integration is often supported by cloud-based solutions, which provide the necessary scalability and computational power. Additionally, insurers invest in training their IT staff to manage and maintain these hybrid systems, ensuring smooth operation and minimal disruption to existing workflows.
- What specific measures are being implemented to ensure the privacy and security of sensitive data when utilizing AI technologies within insurance companies? To ensure the privacy and security of sensitive data when utilizing AI technologies, insurance companies are implementing several measures. One key approach is to keep AI processing within the company’s internal network, reducing the risk of data breaches associated with external platforms. Advanced encryption methods are used to protect data both in transit and at rest. Companies also establish strict access controls and regular audits to ensure that only authorized personnel can access sensitive information. Furthermore, compliance with industry standards and regulations, such as GDPR and HIPAA, is maintained to protect personal data and ensure transparency in data handling practices. By prioritizing robust security measures and transparent data practices, insurance companies aim to build trust with their customers and regulators.
