AI in Life & Health underwriting and claims: Why human oversight matters

Artificial intelligence (AI) is rapidly transforming Life & Health insurance underwriting and claims. AI helps insurers improve productivity, accelerate underwriting decisions, support claims assessment and analyse increasingly complex medical and risk information. However, successful AI adoption depends not only on technology, but also on how people interact with AI and how AI-supported workflows are designed.

In this latest Swiss Re Life & Health Insights article, we explore why a human-centric approach to AI is essential for insurers seeking to unlock the full value of AI-supported underwriting and claims workflows. Drawing on behavioural science research, practical insurance examples and Swiss Re solutions, the article examines how organisations strengthen decision-making while managing the risks associated with AI adoption.

Insurers should focus on four priorities when implementing AI in underwriting and claims.

1. Navigating AI inaccuracies

While AI outputs are often highly accurate, AI can still generate convincing but incorrect information. Research cited in the article shows that people frequently accept inaccurate AI recommendations without identifying errors — a phenomenon known as automation bias. Human-centric workflow design helps underwriting and claims professionals verify AI outputs, identify mistakes and retain responsibility for decisions.

2. Ensuring unbiased decisions

AI systems learn from historical data and existing processes, which means they can inherit or amplify bias. Insurers can identify potential sources of bias, validate AI systems before deployment, monitor performance after launch and apply appropriate governance throughout the AI lifecycle.

3. Maintaining human expertise

As AI increasingly supports tasks such as information extraction and summarisation, organisations need to ensure employees continue to build and maintain critical underwriting and claims expertise. Insurers can preserve strategic knowledge and skills through ongoing learning, active human involvement and thoughtful workflow design.

4. Managing emerging risks

Some AI risks only become visible after deployment. Changes in customer behaviour, medical practice, fraud patterns, regulation or underwriting guidelines can reduce AI effectiveness over time.  Governance, monitoring and adaptation can help organisations manage these evolving risks and maintain performance.

Successful AI adoption depends not only on technology but also on how people use, challenge and oversee AI. Drawing on behavioural science, the article explores how human-AI interactions can be designed to support judgement, oversight and accountability.

Rather than focusing solely on frictionless automation, organisations can benefit from introducing appropriate challenge and oversight into AI-supported processes. Effective human-AI collaboration can be strengthened through approaches such  as structured challenge, devil's advocate roles and other forms of "productive friction".

Alongside practical insurance examples, including MagnumXP Underwriting Assistant and Life Guide Scout, the article provides questions that underwriting and claims teams can use to evaluate AI-supported workflows. These questions help organisations assess AI accuracy, identify potential bias, maintain expertise and establish effective governance mechanisms.

As AI adoption accelerates across the insurance industry, organisations that succeed will focus on both technology and people. A responsible AI approach enables insurers to combine the speed and analytical capabilities of AI with human judgement, accountability and expertise.