Willem Paling: From Messy Middles to Autonomous Agents and the Race for Trust at Scale

AI in insurance has crossed a threshold. The question is no longer whether insurers can experiment with artificial intelligence. The question is whether they can govern it, scale it and trust it fast enough to remain relevant.

In this episode of Scouting for Growth, Sabine VanderLinden speaks with Willem Paling, Executive Manager of AI and Analytics at IAG, about what happens when a major insurer moves AI from side project to core operating infrastructure.

Willem takes us inside IAG’s agentic AI pivot: a shift from isolated proof-of-concepts to established AI products, embedded governance and human-AI operating models built for regulated insurance environments. IAG has launched more AI models in the past two years than in the previous six combined, created a network of 150 Gen AI activators across the business and deployed CASI, an AI claims assistant supporting real customer conversations.

But this conversation is not about speed for speed’s sake. It is about the discipline required to make AI in insurance useful, explainable, scalable and safe.

At the centre of the discussion is the “messy middle” of insurance: the friction zone between intake and decision. This is where claims, underwriting and service teams manage PDFs, handwritten notes, engineering reports, medical packets, emails and fragmented data. It is also where AI is creating measurable value.

Not by replacing human judgement. By giving experts better-structured context, faster.

For insurers, the implications are immediate. Claims teams can move from manual triage to AI-supported case assembly. Underwriters can spend less time collecting information and more time applying judgement. Service teams can interpret, summarise and escalate customer issues with greater consistency. Leaders can monitor AI models, risk, drift and performance as operational assets, not experiments.

For insurtech startups, the opportunity is clear. The next wave of insurance AI will not be won by generic tools. It will be won by ventures that understand insurance workflows, regulatory accountability, data provenance, explainability and embedded AI governance.

For regulators, the conversation is also shifting. Autonomous AI systems need confidence-building layers: audit trails, override pathways, reasoning traces, escalation rules and measurable accountability. Autonomy is not switched on. It is earned.

For boards, the strategic question is becoming more urgent. AI agents will increasingly help customers discover, compare and purchase insurance. When that happens, insurers will not only compete for human attention. They will compete for machine interpretation.

If an insurance policy cannot be read, understood, compared and recommended by an AI agent, it may not simply rank lower. It may disappear from consideration entirely.

That is why machine-readable insurance is no longer a technical detail. It is a distribution strategy.