Generative AI (GenAI) has rapidly become one of the most talked about technologies of the decade. For insurers, whose businesses rely on vast volumes of structured and unstructured data, the appeal is obvious. New GenAI-powered capabilities are emerging in underwriting, claims triage, fraud detection and customer engagement, promising to reshape the industry’s operating model by automating repetitive tasks, accelerating decision making and elevating service.
But alongside this excitement, a noticeable shift is happening in insurance boardrooms. The early phase of experimentation is giving way to a more pointed question: if GenAI is truly transformative, why are the financial gains so difficult to see in the combined operating ratio or on the balance sheet? This disconnect between technological promise and measurable productivity is not new. It echoes the longstanding “productivity paradox,” where large-scale technology adoption fails to deliver immediate, observable gains.
The paradox first gained attention during the IT revolution of the 1980s, when organisations struggled to translate new digital tools into demonstrable improvements. Insurance is particularly prone to this challenge, weighed down by regulatory complexity, legacy systems and a naturally risk-averse culture that slows the pace of change. Even with AI investment in the sector set to exceed $35 billion globally by 2030, more than 80% of insurers say they cannot yet quantify GenAI’s financial impact. The issue is not a lack of capability, but the difficulty of measuring value in a model defined by longtail outcomes and delayed effects.
Understanding this measurement gap is essential before approving new licences, commissioning bespoke development or scaling proofs of concept. Much of GenAI’s value materialises over longer horizons. Enhanced underwriting accuracy influences loss ratios only after several cycles. Claims automation reduces leakage and increases consistency, but reveals its full impact only once applied at scale. Many of the benefits that matter most – higher customer satisfaction, increased retention, stronger broker relationships and improved brand reputation – are critical yet inherently hard to express as immediate financial returns. Meanwhile, the hidden costs of responsible AI, including compliance, data governance, actuarial validation and cybersecurity, are often underestimated. And with the technology evolving faster than organisations can adapt, establishing stable ROI baselines becomes harder still.
Breaking the paradox requires a more nuanced way of thinking about ROI. Instead of relying solely on traditional financial metrics, insurers need frameworks capable of capturing both direct, quantifiable gains and the softer but equally meaningful indicators of long-term strategic value. Improvements in claims handling can be tracked through reductions in processing time, while operational savings become visible as automation reduces the hours required for routine tasks. Alongside these concrete outcomes sit softer benefits such as enhanced customer experience, higher net promoter scores, more productive underwriting and better risk control driven by cleaner data and fewer compliance breaches. Taken together, these form a much richer and more accurate picture of GenAI’s contribution.
As insurers roll out GenAI, many encounter the early dip in productivity associated with the J-curve. Transformation rarely unfolds in a straight line. Training, integration, workflow redesign and cultural adjustment create short-term disruption before benefits begin to accelerate. Effective leadership means recognising this pattern, preparing teams for it and actively working to minimise its depth and duration. One of the simplest ways to do this is to begin with high-impact, low-complexity use cases that deliver quick, visible wins. Data cleansing and transformation is a prime example. GenAI can rapidly clean and restructure large datasets – a task that typically consumes significant human capacity – creating better management information and improving decisions across underwriting, pricing, reserving and operations. The same applies to fraud detection, where GenAI can surface anomalies with minimal configuration. These early successes help build confidence, demonstrate tangible ROI and create momentum for broader adoption.
Once an insurer successfully resolves one of these foundational challenges, the approach can be replicated across other lines of business. Each additional win strengthens organisational confidence, provides evidence of ROI and lays the groundwork for more advanced applications. Over time, the cumulative effect is a business with cleaner data, faster decisions, smoother processes and better customer outcomes – all delivered without a proportional increase in headcount. This is how the paradox begins to break down.
However, as organisations lean into these early opportunities, it is vital they do not lose sight of the role humans still play in the insurance value chain. The biggest risk in the current wave of adoption is overreliance on AI at moments where human judgment remains essential. GenAI is powerful, but it is not 100% accurate and removing the human in the loop at the wrong touchpoints – particularly in customer facing or high risk scenarios – can introduce new errors and unintended consequences. The industry’s enthusiasm has, in some cases, led to the adoption of solutions that are not yet fully mature or not aligned to a clearly defined business problem, resulting in deployments that lack clear ROI. Without a firm understanding of what challenge a tool solves and how success will be measured, AI can quickly become an expensive experiment rather than a transformative capability. This underscores the need for strong governance and consistent guardrails. With no unified industry standards in place, the level of oversight varies significantly between insurers, and there is a growing need for clearer guidance, particularly for generative and agent based AI operating in sensitive or customer facing environments.
As ROI comes under sharper scrutiny following the recent wave of enthusiastic AI adoption, this shift is not a sign of disillusionment but a sign of maturity. The insurers that succeed in measuring and realising GenAI’s true value will be those who embrace a broader definition of productivity, prioritise long term strategic outcomes and guide their organisations confidently through the transition. GenAI will not deliver value automatically, nor will its impact unfold instantly. But with clarity, discipline and thoughtful execution, it offers insurers a powerful opportunity to reshape their operating models into more efficient, more customer centric and ultimately more competitive businesses.
