Insurer Viewpoint: Why Insurance Must Move Beyond AI Pilots to Real-World Adoption

The past two years of testing, pilots, and experimentation have given most insurers a far clearer view of where AI can and cannot deliver value. Insurers, MGAs, brokers, and agencies have spent countless hours testing tools, evaluating vendors, and exploring use cases across underwriting, claims, and customer service.

Meanwhile, the importance of realizing AI’s promise has only become more urgent, as customer expectations have been steadily rising and climate change and escalating costs have pressured loss ratios–all while the industry sits on more data than ever.

However, while individual pilots may have been successful and the technological frontier seemingly advances daily, there are still relatively few examples of AI delivering meaningful financial impact or sustained operational change. In fact, 60% of insurers remain stuck in the exploration or proof-of-concept stage of AI adoption, according to Capgemini’s World Property and Casualty Insurance Report, 2026.

There are exceptions, though. A handful of organizations have seen meaningful results, moving beyond treating AI as an innovation project to embedding it directly into underwriting, claims, and customer service workflows. The difference between these organizations and the majority of the market comes down to a small set of foundational factors that determine whether AI moves from pilot to production.

What It Takes for AI to Deliver at Scale

In my experience, successful AI deployment depends on four key factors: clear problem definition, strong data foundations, system connectivity, and operational trust. Failing to account for any of these creates real barriers to impact and scale.

Too often, organizations begin with a desire to do something with AI without committing to an ambitious, well-defined business problem. Ideas tend to come bottom-up, such as a small use case, a low-risk proof of concept, or a productivity hack. According to the same research, 42% of top global P/C insurers set no KPIs to measure AI success, leaving programs reliant on individual champions rather than institutional capability. Without alignment from the start, pilots generate activity without producing meaningful outcomes, like a chatbot that can only answer basic questions, or a document extractor that exists outside any real workflow. Even a successful initial pilot will lose steam without continued commitment and shared objectives.

Data quality presents another common obstacle. Pilots often perform well in controlled environments but struggle when exposed to real production data. Again, the research backs this up; 74% of insurers point to data quality and cross-functional accessibility as barriers to scaling AI. If the underlying information is fragmented, inconsistent, or incomplete, the output and results will be as well.

Similarly, to unlock full potential, AI must be integrated across your core technology systems. Just as teams operate across several systems, AI must be able to access all these tools as well. By integrating AI into policy administration systems, claims workflows, servicing platforms, and operational processes, your AI agents will have the tools to analyze, recommend, and act in real time. The report also reveals that organizations leading AI adoption are more than twice as likely to have integrated data and cloud infrastructure enterprise-wide (86% vs. 35%) and three times more likely to have built shared, explainable AI infrastructure enabling cross-functional scaling (74% vs. 24%).

Finally, in addition to these technological and data considerations, operational trust is equally critical and often underestimated. Adjusters, underwriters, service representatives, and agents are the ones who will use these AI-powered workflows or work alongside them. Launching pilots in isolation, often on a siloed innovation team, can result in disjointed workflows, resistant adoption, or, worse, mistrust. By co-designing with operational teams from day one, you build buy-in, drive adoption, and create a meaningful feedback loop to improve your AI products over time.

In regulated environments, trust requires that AI systems must be explainable, which means human-in-the-loop checkpoints where it matters, a strong analytics framework to guard against bias, and ongoing monitoring of both the systems and the processes they support. This is not unique to distribution; it applies equally at the carrier level, where the stakes of an unexplainable decision are just as high.

Without these foundations in place, most pilots will remain novelties rather than core business capabilities.

What Good Looks Like at Scale

The benefits of embedding AI directly into operational workflows extend far beyond efficiency gains. Leading organizations are realizing additional benefits such as improved customer experience and risk decisioning by integrating AI directly into core workflows across claims, customer service, underwriting support, and other operational functions. Unlocking these broader benefits requires reimaging workflows for the AI era, not merely automating standalone tasks.

Looking closer to home, at Hippo our own experience mirrors these wider industry trends, having embedded AI into claims and customer service processes to increase responsiveness and expand operational capacity. Our AI-powered customer service capabilities now handle routine interactions across policy servicing and billing, while our digital first notice of loss workflow captures and organizes claims information so adjusters can focus on higher-value activities. We expect more than 70% of claims to be filed digitally, and our current claims staffing model could support a 30-35% increase in claims volume.

These outcomes illustrate what production-scale AI looks like in practice. Success comes from starting with clearly defined business objectives, building trust with operational teams, creating strong data foundations, and connecting AI to the systems where work actually happens. When those elements come together, AI shifts from a series of disconnected pilots into an operational capability that can reshape how insurers serve customers, manage risk, and scale their businesses.

Ramsay is the chief product and AI officer at Hippo.

Topics InsurTech Data Driven Artificial Intelligence Carriers