AI in Insurance: Moving Beyond the Hype
We know that insurers have big ambitions around AI. Roots Automation’s Stateof AI Adoption in Insurance 2025 report found that 70% of insurance leaders are actively testing or exploring AI solutions. AI is widely seen as key to improving both operational performance and financial outcomes.
But progress is slow. The same report shows that fewer than 22% have advanced their AI projects beyond testing into full production. Most insurers are stuck in pilot mode,with limited real-world impact.
This gap between ambition and execution is a common theme across the industry. Many are still experimenting with generic tools and off-the-shelf solutions that sit outside core operations. As Giles Mazars, Group Chief AI Officer at vector8, explains, while insurers are still “talking a lot about generic AI,” the real shift is happening around “AI on very specific tasks that are really embedded into your own processes, not something on the side.”
So, what’s keeping AIfrom scaling? And how can insurers ensure their bold AI ideas translate into valuable impact across underwriting, claims, sales and beyond?
What’s holding AI back in insurance isn’t a lack of interest or investment. It’s the way it’s being applied. Too often, it’s treated as a side project instead of a strategic enabler.
Many insurers still approach AI as a standalone initiative. It runs alongside the business, not within it. There’s no clear link between AI spend and the outcomes that matter most in underwriting, claims or customer service. As Mazars says, AI should “serve the business strategy,” not operate independently from it.
That disconnect shows up in the tools being used. While there are strong insurance-specific AI products available, most solve for one part of the value chain in isolation.They often require users to switch systems or log in to a separate interface,creating a fragmented experience that adds friction instead of removing it.
At the same time,generic large language models can struggle in complex, regulated environments. They aren't built to handle the detail or nuance of insurance-specific tasks. Specialised models — smaller, focused and fine-tuned — can often deliver better performance and lower cost.
Even when the right tools are in place, adoption can still stall. If AI isn’t embedded into day-to-day processes, it becomes just another disconnected system that slows people down instead of helping them move faster.
Insurers don’t need more proof-of-concept tools. They need AI that supports real-world challenges. The most effective AI transformation projects focus on everyday tasks that are ready for improvement. Critically, this means these processes must be connected and integrate seamlessly into existing, often fragmented,legacy systems.
Document extraction is one of the clearest examples. It pulls key information from unstructured sources like invoices, forms or supporting documents. The technology has matured. It no longer relies on fixed templates and can now interpret layout, context and meaning with far greater accuracy.
In motor claims,it helps process complex invoices with multiple line items. In health insurance, it can separate out different reimbursable components. In underwriting, it can read supporting documents and surface the details underwriters need to move faster and make better decisions. These are different workflows, but the underlying capability is the same.
While impactful, these examples offer just a glimpse of what’s happening across lines of business and organisation. They're a small fraction of the use cases being developed and applied in response to the nuanced challenges of the sector.
At vector8, we’ve built our approach around modularity and reuse. We’ve designed ablueprint that is structured around agents, skills,accelerators and a model gateway.
Skills handle specific tasks using defined business logic. Accelerators are reusable components that can be configured for different processes. And the model gateway routes each task to the best-fit model and lets you switch or update models without touching the rest of the system. This setup helps teams stay flexible in a landscape where models are improving rapidly, minimising future risk by investing in tools that are sustainable and adaptable.
And crucially, these tools are designed to support people, not replace them. “Automating everything or nothing is not a good idea,” Mazars said. “The best idea is to augment the people doing the job so they just need to validate.”
This is where AI starts to deliver lasting value. Focused, modular and built to reflect how work actually happens.
There’s increasing pressure on insurance leaders to show visible progress with AI. But without the right foundations, that often means launching pilots, rolling out generic tools or building point solutions that don’t scale. It’s progress on paper, not in practice.
Many projects never make it past the test phase. Some deliver small wins but stay siloed. Others are built around technology rather than the real work of underwriting or claims. The result is a lot of activity, but not much that changes how the business operates.
This is a missed opportunity. According to the ACORD Insurance Digital Maturity Study, robust AI integration could reduce expenses by up to 14.6 percent for P&C insurers, unlocking an estimated 480 billion dollars in annual savings across the industry. For life insurers, the value could exceed 300 billion dollar seach year.
Beyond cost, the benefits of a well-integrated AI approach are wide-reaching. Done right, AI can:
● Improve broker and customer experiences by making interfaces faster and more intuitive
● Save time through real-time processing and automated triage
● Reduce manual work by automating first-pass document reviews
● Lower error rates, especially in pricing and data capture
● Enhance risk assessment with automated premium verification and visual analysis
● Strengthen fraud detection
These are not theoretical gains. They come from targeted, well-integrated solutions that are designed to flex, scale and fit insurance as it exists today.
The gap between AIambition and impact is still wide across the industry. But it doesn’t have to stay that way.
The insurers seeing real value aren’t the ones chasing the latest trend. They’re embedding AI directly into core underwriting, claims, and sales processes and building in away that reflects their unique ecosystems, from motor and health to property and public liability.
Those leading the pack are making the leap from isolated projects to repeatable and scalable AI tools. They are aligning with strategy. Solving real problems. And creating a foundation that can evolve with the technology.
AI can deliver on its promise, but only if it’s built for the reality of insurance.
We work with insurers to build solutions that support their business strategy while tackling immediate challenges. Whether you're starting out or scaling up your AI ambitions, we can help you take the next step.
Talk to our team to find out how.
Most insurerstreat AI as a side project rather than integrating it into core operations or business strategy. That disconnect leads to limited adoption, duplication and slow progress.
Tasks like document extraction, verifying premium information, risk assessment, and real-time triage consistently deliver impact when embedded in workflows and reused across departments.
Start with business strategy, not tech. Build modular, flexible components that can evolve as models change. Focus on adoption, reuse and integration from day one.
Uncover fresh perspectives with our handpicked blog posts on AI advancements.
.png)
