Artificial intelligence has quickly become one of the biggest priorities on the agenda for insurance leaders. Whether it's accelerating claims processing, supporting underwriting decisions, improving customer service or helping to detect fraud, insurers can already see where AI has the potential to create value.
The challenge is that most organisations are struggling to scale their AI use cases.
Despite growing investment and no shortage of ambition, moving AI from promising pilots to organisation-wide capability is still difficult. The reason is rarely the technology itself. More often, it's what sits beneath it: fragmented data, complex legacy environments, governance requirements and the need to explain every decision with confidence.
Our latest Insurance Custom Software Development Report 2026 paints a similar picture. Half of the insurance organisations surveyed identified managing data privacy across multiple systems as a major security and compliance challenge, while another 50% cited integration issues with legacy systems that affect compliance. Nearly half (45%) also raised concerns about how sensitive third-party providers handle data.
These findings point to a reality many insurance CIOs will recognise. AI may be the headline, but data quality, governance and regulatory accountability are often the factors that determine whether an initiative scales or stalls.
Put simply, insurers don't have an AI problem. They have a trust problem.
The focus on trust is increasingly visible beyond the insurance sector itself. The FCA's ongoing Mills Review is exploring the long-term impact of AI across retail financial services, examining issues such as accountability, transparency and consumer outcomes. As AI becomes more embedded in decision-making, trust is no longer a compliance consideration. It's becoming a strategic requirement.
Many insurers are already seeing positive outcomes from AI-powered document processing, underwriting support and claims automation; the challenge begins when organisations try to expand those successes beyond a pilot environment.
A claims automation solution might perform brilliantly in a controlled proof of concept. But when it needs to pull information from multiple policy administration systems, interact with decades-old platforms, comply with regulatory requirements and maintain a clear audit trail, the complexity increases dramatically. What initially looked like an AI challenge quickly becomes a data and governance challenge.
AI is exposing weaknesses that already existed. It's shining a brighter spotlight on fragmented data estates, inconsistent processes and technology architectures that were never designed for an AI-first world.
Read what our Insurance Technology Director, Mark Hankin, believes it means to be AI-ready, and what foundations need to be in place before AI can truly scale.
Historically, governance has often been viewed as something that slows innovation. In insurance, the opposite is becoming true.
As AI becomes embedded in underwriting, claims handling and customer services, insurers need confidence that decisions can be understood, challenged and explained.
This is why explainability, auditability and human oversight are becoming central to successful AI programmes.
Governance shouldn’t be treated as a compliance exercise that happens after deployment. It should be built into your platforms, workflows and operating models from the outset. That allows you to move faster because trust has already been designed into the process.
“The constraint is rarely the model alone; it is whether leaders have confidence in the data, controls and accountability around it.”
Mark Hankin
AI can only scale when organisations trust the data, governance and operational frameworks that support it.
One of the most valuable insights from our latest report is that AI readiness has little to do with having access to AI. Most insurers can access the same models, platforms and tools.
The real differentiator is whether those tools can operate effectively within an organisation's existing environment.
As our report concludes, AI initiatives often falter not because the technology underperforms, but because legacy architecture, fragmented data and unclear ownership prevent intelligent systems from operating effectively at scale.
A practical readiness agenda starts with four questions: which decisions are suitable for AI; which data sources are authoritative; who owns model and outcome risk; and what evidence is required before deployment. Answering these questions turns trust from an aspiration into an engineering and operating discipline.
The insurance industry has never lacked ambition. The opportunity presented by AI is enormous, and many organisations are already proving how it can improve efficiency, decision-making and customer outcomes.
Success depends on investing in the foundations in the right sequence: prioritising material use cases, establishing accountable data ownership, integrating with the core estate, embedding controls and preparing operations to monitor performance over time.
Investing in foundations, connecting data, modern architecture, strong governance and operational trust are key to success.
In insurance, trust has always underpinned the product. Increasingly, it must also be engineered into the data, platforms, controls and operating practices behind AI.
Because in insurance, trust has always been the product. Now it's becoming the infrastructure behind AI as well.
Read our latest guide, Closing the gap between AI ambition and execution