For decades, insurance has been defined by episodic moments, such as a policy issued, a claim filed and a loss assessed. However, a new generation of AI agents can reshape that model by shifting the industry from reactive transactions to continuous engagement built around prevention, early intervention and outcome management.
“Insurance can become more about prevention than dealing with a claim,” says Will Hyams of Howden Group Holdings, who spoke on a panel for the Microsoft AI Tour London on ‘Pioneering the future of financial services with AI’. “We could help clients to avoid that catastrophe if we can – and the real way to do that is through real-time data.”
That ambition is increasingly achievable as insurers embed agentic AI across the value chain.
Tyler Jones, chief marketing officer at Duck Creek Technologies, describes a “fundamental shift from reactive protection to proactive outcomes, where insurers gain a far more granular understanding of risk and can intervene earlier”. Intelligent core platforms like Duck Creek increasingly embed AI directly into underwriting, policy and claims workflows, allowing insurers to move from episodic transactions toward continuous insight and prevention. At the heart of this transformation are AI agents, which are “systems that support autonomous tasks, help make decisions and take action under human oversight,” as Jim DeMarco, head of insurance advisory for worldwide financial services at Microsoft, defines them.
Many insurers began their AI journeys with digital assistants that summarise documents, extract data and reduce manual effort. Those tools delivered quick wins. But, as Jeffery Williams, a managing director in Microsoft’s insurance advisory organisation, explains, this is only the starting point.
The more significant shift comes when insurers move from using AI to support individual tasks to managing intelligent agents that coordinate entire workflows. “In core operations, agents ingest unstructured submissions, medical records, images and emails, converting them into structured insights and recommended actions,” says Williams.
Agents can prepare decision briefs, track regulatory requirements, surface risks and exceptions, automate hand-offs and ensure continuity across teams. Execution, in this model, is no longer confined to a single function. It is distributed across people, processes and systems. AI evolves from a supporting tool into a digital colleague, operating alongside actuaries, underwriters and claims professionals.
Nowhere is the shift from reactive insurance to proactive prevention more significant than in actuarial functions. Today, many insurers still rely on periodic, model-driven analysis. Emerging risks may only surface during scheduled model re-runs. Williams envisions something different: an actuarial environment where emerging risk is surfaced continuously as conditions change.
“AI agents can ingest and normalise satellite imagery, hazard layers, internet-of-things sensor feeds and unstructured field reports, converting them into governed risk features with clear lineage and documentation,” he says. “Rather than manually stitching datasets together, actuaries can rely on agents to coordinate spatial analytics, scenario testing and model execution, highlighting changes in exposure or climate conditions that warrant closer review.”
This approach allows insurers to scale judgement over labour, addressing persistent talent shortages and rising complexity. It also reinforces the broader shift from reactive processing to earlier intervention and prevention.
Underwriting is one of the clearest examples of human-agent collaboration in action. AI agents can assemble submissions, evaluate risk signals and propose pricing and terms, freeing underwriters to focus on risk appetite, negotiation and final decision-making. Rather than replacing underwriting expertise, agents absorb the front-end cognitive load.
The Swiss insurtech startup Legal-i illustrates what this can look like in practice. Its agentic AI platform reasons over complex medical records to support underwriting and claims assessments. Legal-i turns PDF records into interactive patient histories and pulling key information from doctors’ reports, lab results and admin documents, Legal-i helps insurers handle cases up to 80 per cent faster, with greater accuracy and more consistent decisions.
“Agents parse broker emails, ACORD [Association for Cooperative Operations Research and Development] forms and attachments; enrich submissions with internal and third-party data; and present prioritised worklists based on risk, value and complexity,” says Williams. “For complex risks, agents synthesise engineering, financial and medical documentation into concise, auditable briefs with clear source attribution, enabling faster and more consistent decisions across teams.”
Meanwhile in claims, the balance between automation and human oversight is particularly sensitive. Agents can resolve low-complexity cases end to end, while claims managers retain control of complex, high-impact scenarios where judgment, empathy and regulatory accountability matter most. “Humans retain responsibility for outcomes, while AI expands operational capacity,” says Williams.
AI agents can also coordinate across care networks, benefits and support services, reserving human interaction for moments that demand empathy. They personalise communications in language, format and timing, while preserving transparency and consent.
“In insurance, trust is not an abstract principle,” says Jones. “It is an operational asset that must be earned transaction by transaction. Innovation and governance are not competing priorities. They succeed only when they are embedded together inside the systems insurers use every day. That’s why the next generation of insurance platforms are designed with AI, auditability and regulatory transparency built directly into operational workflows.”
Real-world deployments already demonstrate impact. Sedgwick’s Sidekick Agent, developed in collaboration with Microsoft, has improved claims processing efficiency by more than 30 per cent through real-time guidance and decision support. Meanwhile, one major insurer deployed more than 80 AI models in claims, cutting complex-case liability assessment time by 23 days and significantly reducing customer complaints, according to McKinsey & Company’s The future of AI in the insurance industry report.
“Beyond individual claims, agents can detect life events – entering the workforce, buying a home or approaching retirement – and dynamically assemble modular coverage, service workflows and communications tailored to customer needs and channel preferences,” says Williams. “Insurance evolves into a continuous, adaptive relationship, rather than a series of isolated transactions.”
Despite the promise, scaling AI remains a challenge. Only a small minority of insurers (seven per cent) have successfully scaled AI initiatives across their organisations, according to Boston Consulting Group’s 2025 research, Insurance Leads in AI Adoption. Financial pressures – Swiss Re reports industry losses exceeding $100 billion annually for multiple consecutive years – compound the urgency, even as siloed teams and talent shortages slow progress.
For Williams, success depends on treating innovation and governance as inseparable. “Insurance is fundamentally a trust business,” he says. “That trust extends well beyond underwriting and claims decisions to how customer data is protected, governed and used responsibly.”
“Insurance evolves into a continuous, adaptive relationship, rather than a series of isolated transactions,” says Williams
Microsoft’s approach starts with a security-first foundation across Microsoft 365, Azure and Dynamics, designed to meet global standards for privacy, compliance and auditability. Governance capabilities allow insurers to define clearly what AI can do, on whose behalf and under what authority. The Microsoft Responsible AI framework emphasises fairness, reliability, safety, privacy, inclusiveness, transparency and accountability. This translates into practical guardrails such as explainability in AI-assisted decisions and bias detection.
Adoption, Williams stresses, should be phased and pragmatic: “Most organisations begin with AI as an assistant before progressing to more autonomous agent scenarios. Insurers should build an enterprise AI control plane, prove a small set of reusable agent patterns, modernise their data backbone and define explicit human accountability.”
Companies such as Cognizant are supporting this transition with structured frameworks and prebuilt assets; its Agent Foundry offering, for example, provides reusable tools and governance accelerators designed to reduce implementation time and help insurers meet regulatory and compliance requirements as they scale agentic AI.
For those that get it right, the prize is significant.
“Three to five years from now, frontier insurers will look fundamentally different, not because they adopted more technology but because they will have redesigned how insurance operates,” says Williams. “AI agents, embedded with identity, governance and auditability, will orchestrate underwriting, claims, service and finance as a coordinated fabric.”
In that future, insurance no longer revolves around isolated moments of loss. Digital colleagues operate quietly in the background, allowing human experts to focus on judgment, empathy and trust.
Partner perspective
Tyler Jones of Duck Creek shares how to support insurance firms with embracing new operating models through agentic AI
“With Duck Creek Intelligence powered by Microsoft, we help insurers apply AI deliberately, with the steadiness required of a trust-based industry. We’ve learned that real transformation doesn’t come from chasing the next breakthrough, but from applying change calmly where insurance carries real consequence.
“For one multi-line insurer, AI is now embedded directly into core policy, billing and claims workflows to reduce friction, speed up decision-making, and improve customer outcomes. By applying AI-driven automation within governed, production environments, the carrier significantly shortened product launch cycles and accelerated claims decisions – all while maintaining regulatory confidence and operational oversight. Our role is to absorb disruption, not accelerate it, ensuring insurers can modernise with clarity, resilience and confidence.”
Discover more insights in the Spring 2026 issue of Technology Record. Don’t miss out – subscribe for free today and get future issues delivered straight to your inbox.