There is a moment in most technology cycles when the question shifts from what can this do to what does this change? With AI, that moment has arrived.
AI is embedded across workflows, reading claims files, shaping underwriting judgement, and filtering cyber threats, while increasingly sitting behind outcomes that carry real financial, operational and human consequences.
The challenge now is responsible adoption. That sounds straightforward, but in practice it is anything but. Capability is moving quickly. Governance, less so. Many organisations are already relying on systems they cannot fully interrogate, using outputs they do not always challenge, within frameworks that are still evolving. The result is a subtle but important shift in how risk is created, distributed,and, at times, amplified.
For insurers, this is uncomfortable territory. The industry is used to dealing with uncertainty, but typically with the benefit of history, data and precedent. Here, those anchors are weaker. Exposure is already building across multiple lines, while questions of liability, accountability, and insurability are still being worked through.
But this is not an argument for caution at the expense of progress. AI is already delivering real value, improving speed, consistency and insight across a range of insurance functions. Realising that value relies on strong governance, clear points of human judgement and a measured approach to building trust in the outputs.
This edition of the Willis Research Network's Risk and Resilience Review focuses on that gap, looking at how AI is being deployed in practice and where risks are beginning to surface.
The selection of articles reflects that intent. Taken together, they build a more complete picture of responsible adoption, highlighting a set of interlocking challenges that cut across people,processes and systems.
For the Willis Research Network, this is an area of ongoing focus. AI itself is not new. What is changing is how it is being used and with it, the shape of exposure, much of which is only now starting to come into view. The task now is to move beyond general discussion and into something more grounded: identifying where risk accumulates, how it travels, and what effective adoption looks like in practice. Not just controlling downside, but understanding how human judgement and AI systems work together to deliver better, more resilient outcomes.
AI risk, liability and insurability in 2026
As AI becomes part of day-to-day decision-making, questions of liability are starting to move from theory into practice. This article looks at how responsibility is beginning to take shape, particularly where organisations rely on AI outputs across underwriting, operations, and client interactions. It also highlights where existing insurance structures may come under pressure as these use cases scale.
Assessing trust in AI systems
Much of the focus in assessing AI performance centres on individual metrics, but trust in AI systems depends on how multiple dimensions interact in practice. This piece broadens the discussion to trustworthiness, exploring how issues such as reliability, fairness and governance translate into real exposure. It provides a more practical lens on how these risks can be understood and, increasingly, assessed.
AI-powered threat detection: A new era in cyber risk management
Cyber risk is evolving in step with AI. On one hand, AI is strengthening detection capabilities, enabling faster identification and response to threats. On the other, it is lowering the barrier to entry for more sophisticated attacks. This article explores that dual dynamic, showing how AI is reshaping both sides of the cyber landscape. For insurers, the challenge is keeping pace with a risk that is becoming more adaptive, more scalable, and harder to benchmark using traditional approaches.
Scaling AI in claims and resilience advisory
Some of the most immediate impacts of AI are being seen in core insurance processes. This piece looks at how AI is being applied in claims handling and resilience advisory, improving efficiency while also changing how decisions are made. It also touches on the practical challenges of scaling these capabilities, particularly around oversight and consistency.
The anatomy of an AI leader
Cyber risk is evolving in step with AI.
Not all organisations are progressing at the same pace with AI and the reasons go beyond access to technology. This article explores the characteristics that distinguish leading adopters, from data infrastructure and governance to culture and execution. It makes a simple but important point: successful adoption is less about tools and more about how they are embedded into the organisation. For risk leaders, that has direct implications for operational resilience and long-term competitiveness.
Autonomous vehicles: The road to trust
Autonomous vehicles offer a clear illustration of how AI redistributes risk. As control shifts from human drivers to systems, responsibility becomes more complex, spanning manufacturers, software providers, and operators. This article examines how trust is built in that environment, through testing, validation, and governance. For insurers, it highlights the challenge of adapting to a world where liability is shared, evolving, and closely tied to technological performance.
The space between fiction and failure: Human agency in the age of AI
Much of the public conversation around AI focuses on extreme scenarios, but most risks emerge in more routine settings. This piece brings the focus back to everyday use, examining how human judgement interacts with AI systems in practice. It shows how over-reliance, misinterpretation, or weak oversight can create failure pathways that are difficult to detect early. The central message is that resilience depends not just on the system, but on how people engage with it.
Not all organisations are progressing at the same pace with AI and the reasons go beyond access to technology.
