Implementing GenAI while mitigating risk

Insurers are embedding GenAI into customer portals and mobile applications to simplify an otherwise complex buying process, significantly elevating the customer experience. In particular, the evolution of agentic AI in the form of GenAI-enabled chatbots will empower consumers to move effortlessly through the insurance purchasing process.

These advanced chatbots exhibit deep awareness of customer intentions and engagement patterns and can support them with context-sensitive information.

GenAI also brings greater speed to new business underwriting and first notice of loss claim processes. Digital ingestion solutions leveraging Large Language Models (LLMs) enhance the precision of document analysis and data mapping. On top of these productivity gains, GenAI is starting to enhance the decision-making processes in underwriting and claims through more timely underwriting risk insights and support for the claim evaluation process.

Types of risks associated with GenAI usage

Despite these undeniably powerful and game-changing use cases, GenAI, like any other emerging technology, presents insurance companies with just as many challenges to overcome as new opportunities to explore, particularly concerning risk.

There are many dimensions of risk insurers will need to address with the appropriate policies and governance when using GenAI.

• Intellectual property rights: The training content LLMs use is often broadly sourced, which raises concerns of potential copyright infringement.
• Accessibility: Technology that relies heavily on GenAI carries possible accessibility violations for disabled customers.
• Data accuracy and quality: GenAI-enabled agents can generate inaccurate or out-of-date outputs.
• Privacy and security: Should LLMs expose personally identifiable information, it would raise privacy concerns.
• Bias and discrimination: LLMs lack emotional intelligence and are not automatically aware of certain cultural sensitivities or moral and legal implications. If a GenAI-powered system denied an insurance claim to a marginalized individual or family based on biased algorithms there could be legal and reputational implications.
• Violating AI regulations: GenAI solutions must consider the various regulatory bodies shaping the AI space today, like the EU’s AI Act and the potential state regulations in the works across the US.  
  

AI governance has become a top priority for all insurers and ideally, it is coordinated across other governance committees.

AI governance should also align with the enterprise AI strategy, which has a foundation of robust risk management principles. Common pillars of AI governance include:

• Compliance: External and internal audits.
• Security: Privacy protection, access controls framework and AI-enabled security architecture.
• Responsible AI: HR polices and partner and supplier interaction policies.
• Data quality: Integrated LLM data architect standards, including information management policies.
• Decision authority: Standards for accountability and approvals, usage authority and escalation procedures.
• Engagement and adoption: Education, cost management framework, usage benchmarking and consumption metrics monitoring.

Taking action

GenAI is still relevantly new, which means there are a host of unknowns even the largest enterprises have yet to discover.

Building a test-and-learn environment is essential to uncover potential security and GenAI performance risks. Insurers must define and operationalize risk management and governance policies in conjunction with – not independent of – the stages of GenAI adoption and roll-out such that they can dynamically adjust and refine as the technology advances and insurers increase the pace and sophistication of engineering these solutions.

Furthermore, insurers should partner with engineering firms that bring first-hand AI/GenAI development experience combined with deep insurance business skills to help build solutions that deliver measurable business value while adhering to crucial risk management, compliance and governance principles.

Gail McGiffin is the Managing Director and Global Insurance Advisory Lead at EPAM, where she helps insurers deliver large-scale transformation programs focused on growth, efficiency and digital innovation. With over 35 years of experience in insurance leadership, she specializes in underwriting modernization, data-driven platforms and operational efficiencies.

See also:

• AI versus human touch in workers’ compensation claims management
• Insurers need a better way to act on portfolio risk
• 7 Ways financial professionals are using AI to streamline workflows