The following text field filters the results that follow as you type.
AI creates real value in specialty insurance when embedded directly into underwriting and claims workflows. Explores why treating AI as a business capability—not an IT initiative—can improve speed, consistency, and decision-making across the organization.
Published on May 11, 2026
6-minute read
Earlier this year, an essay by technology investor and former AI founder Matt Schumer gained wide circulation for its claim that machines can now perform much of the technical core of knowledge work (“Something Big Is Happening”). The claim was provocative, and not entirely wrong, but it stopped short of the harder question. Capability alone does not determine outcomes; can AI create durable economic value inside real businesses, under real constraints?
This question is particularly pertinent for specialty insurers. We operate in a highly regulated industry with a complex value chain and layers of legacy technology. Judgment, accountability, pricing discipline, claims governance, and trust are not peripheral considerations—they’re the business. The real question, then, is not whether AI is a can-do-it-all machine, but whether it can be embedded in ways that genuinely improve how specialty insurance works.
Viewing AI as a capability, not a system
Part of the industry’s uneven response to AI reflects a deeper misunderstanding of its form and function. AI is still too often approached as a technology project that’s owned by IT, delivered through central programs and then handed to the business.
That framing no longer fits reality. Treating AI as an IT asset may be familiar, but it increasingly acts as a structural constraint rather than a governance choice—and is fundamentally misaligned with how value is now created. AI isn’t a system to be deployed, but a capability to be used: one that underwriters, claims handlers and support functions can adopt directly to improve and redesign their own processes and output.
This shift in how AI must be approached is arriving alongside three wider forces. Model capabilities are improving in visible leaps. Broker and policyholder expectations are rising just as quickly. Regulators are increasingly focused on governance, accountability and consumer outcomes when new technologies are introduced. For a sector built on careful judgment, the combined tempo is uncomfortably brisk.
AI delivers value when it’s treated as a business capability, not an IT deployment.
Unlocking AI’s value requires intention over invention
AI only matters if it creates economic surplus. Technical capability on its own is not the test. Faster models, clever demonstrations or isolated pilots do not move the needle—unless they improve service in ways customers value or allow firms to deploy capital and expertise more effectively. In specialty insurance, that surplus shows up through improved operating efficiency and faster turnaround time.
Specialty insurance will remain a judgment led business; what’s changing is the rhythm at which that judgment must be applied. When carriers can use AI tools to help organize and surface relevant information earlier in the review process, underwriters can reduce cycle time while maintaining appropriate underwriting discipline. The real danger is not that AI occasionally generates a flawed output, but that organizations that hesitate will simply take fewer decisions, learn less from them and drift out of step with their distribution partners.
What the industry is telling us
A recent Markel LinkedIn survey, conducted in February 2026, underlines this tension: responses showed a sector divided between those moving assertively and those proceeding with caution. That caution is understandable, but increasingly difficult to justify as a long term position.
Importantly, respondents converged not on novel or speculative use cases, but on the same pressure points where operating friction and economic drag are already most visible. Specialty insurers are rightly skeptical of opaque models and unclear accountability. Yet the survey also revealed a growing recognition that partial or selective adoption is unlikely to remain viable as economic pressures intensify.
When it comes to adopting AI where it truly adds value, caution is understandable—but increasingly difficult to justify as a long-term position.
The cost of hesitation and unclear ownership
Specialty insurance’s reliance on deep human expertise carries a structural cost. Document heavy workflows, manual triage and fragmented tooling scale headcount faster than premium. Combined ratios are not neutral to inaction. In that context, delaying AI adoption does not preserve current economics—it entrenches inefficiency at a moment when rating pressure is tightening returns across multiple segments of the specialty market.
Ownership matters because the constraint is no longer intelligence alone, but integration into live, accountable operations. AI adoption stalls when responsibility is diffused across central programs, legacy systems and unclear decision rights. It accelerates when business teams own both the problem and the outcome.
When AI is treated as a business capability rather than an IT owned initiative, improvement compounds. Small gains accumulate quickly in how information is surfaced, decisions are framed and exceptions are handled. Over time, the difference shows up not as a single breakthrough, but as a durable operating advantage.
More subtly, though, this strengthens loss ratio discipline by reducing variance, ensuring similar risks are assessed in similar ways and anchoring decisions more closely to stated appetite.
How Markel is responding to AI
These realities helped shape Markel’s approach: we fund AI centrally and apply it across the insurance value chain—including underwriting, claims, operations, finance and support functions. Projects that explore these innovations are led by business owners, not IT, to keep accountability close to decision making.
Within underwriting, teams are running targeted AI rewiring sprints focused on areas where document intensity and complexity impose a material drag on performance. For example, AI-enabled tools are being used in complex underwriting lines to help teams review large volumes of material more efficiently. This allows underwriters to identify information gaps quicker and helps support well-informed underwriting discussions. For brokers and policyholders, this translates into clearer outcomes, delivered earlier and with greater consistency.
By embedding AI across underwriting, claims and operations—and keeping ownership with the business—Markel is reducing friction, improving speed and strengthening decision making.
Adaptation and adoption are economic necessities
The purpose of AI is not to replace judgment, but to apply it more evenly with less friction. Underwriters can spend more time interpreting information and less time locating it. Straightforward risks move faster, and complex ones receive the scrutiny they deserve.
Matt Schumer was right to draw attention to the scale of recent AI advances. The debate must now move from what machines can do in isolation to what organizations can do with them in practice.
In specialty insurance, that question becomes whether firms can justify the economics of operating at yesterday’s pace as the market continues forward. Those that treat AI as a curiosity will learn slowly. Those that embed it as a core business capability, governed with seriousness and owned by the people accountable for outcomes, will redefine how effectively insurance does its job.
