If you listen to how insurance leaders talk about AI, you will think underwriting is the center of the universe. Risk scoring. Pricing precision. Portfolio optimization.

These are important, sophisticated, and heavily marketed. However, they are not where insurers win or lose trust, margin, or long-term relevance. This happens somewhere far less glamorous: claims.

Over the years, I worked with insurers across property and casualty, specialty lines, and reinsurance as they pursued digital transformation. The pattern is consistent. Underwriting attracts innovation budgets and executive attention. Claims, where most costs are incurred and most customer relationships are tested, get incremental automation at best.

However, this imbalance has become a strategic liability. If you are a CEO or board member of an insurance company, claims determine three things that matter more than any actuarial model: loss ratio volatility, regulatory exposure, and customer trust. AI changes claims fundamentally, but only if leaders stop treating it as back-office plumbing.

The scale of the issue is not subtle. Just a couple of years ago, claims expenses typically accounted for 60–70% of premiums in property/casualty insurance, according to the U.S. Department of the Treasury’s September 2024 annual report on the insurance industry. Natural catastrophes are increasing both in frequency and severity, driving record losses year after year. According to Swiss RE, global insured losses from natural catastrophes exceeded $100 billion in multiple recent years, with secondary perils, like floods, wildfires, convective storms, now driving most of the growth.

However, claims operations in many firms still rely on fragmented workflows, manual triage, and overburdened adjusters juggling hundreds of cases. The hidden problem is not inefficiency. It is variability. Claim outcomes often depend less on policy language than on who handles the claim, when it arrives, and how overwhelmed the system is at that moment. This inconsistency quietly erodes trust with customers, regulators, and reinsurers alike.

However, AI addresses this not by replacing adjusters, but by restoring consistency at scale. In the claims environments I’ve worked with, AI triage systems analyze incoming claims across structured and unstructured data — photos, videos, text descriptions, historical loss patterns — to prioritize cases by complexity and risk. Straightforward claims move quickly. Complex or suspicious claims receive early expert attention. Adjusters stop firefighting and start adjudicating. The result is faster settlement, but more importantly, less leakage.

Non-health insurance fraud costs insurers an estimated $40 billion annually in the United States, according to FBI and Coalition Against Insurance Fraud data. Traditional fraud detection looks for known patterns. AI identifies anomalies, like subtle deviations humans miss because they normalize noise. In one engagement I had, anomaly detection flagged a cluster of claims that appeared legitimate individually but revealed coordinated behavior when viewed systemically. No accusation. Just prioritization. Investigators focused on where it mattered. This distinction matters legally and culturally to find the true root cause.

Likewise, AI also transforms catastrophe response, an area where insurers quietly lose goodwill when they need it most. After major events, claims volumes spike, data quality deteriorates, and customers are emotionally vulnerable. Vision models can assess damage severity from images within minutes. Language models can guide claimants through next steps clearly and empathetically. Predictive models can forecast resource bottlenecks days ahead. I’ve seen insurers deploy these tools selectively and then hesitate to scale them. Not because they didn’t work, but because they exposed uncomfortable truths about staffing models, vendor dependencies, and outdated performance metrics, shining the light on problems no one wants to discuss.

Consequently, claims operations often resist change because they are built around heroics. That’s the struggle. AI replaces heroics with systems, and this change feels threatening in cultures that reward endurance over outcomes.

There is also a regulatory dimension executives underestimate. Regulators care deeply about fairness, explainability, and consistency in claims handling. AI, when governed properly, improves all three. It creates auditable decision trails, flags bias, and reduces arbitrary variance. The alternative, manual discretion at scale, is far harder to defend.

So why does underwriting still dominate the AI narrative? Because underwriting feels strategic and controllable. Claims feel messy, emotional, and exposed. But that discomfort is precisely why claims represent the greatest leverage point.

As a result, the insurers that invest seriously in AI claims transformation gain advantages competitors struggle to copy: lower loss adjustment expenses, more predictable loss ratios, faster catastrophe recovery, and reputations for fairness when it matters most.

This is not about becoming digital. It is about becoming reliable under stress. Consequently, the executive change required is subtle. Stop asking how AI improves pricing precision. Start asking how it stabilizes outcomes when conditions are the worst.

In the next decade, insurance leaders will be judged less by how intelligently they price risk and more by how consistently they honor it. So, the firms that recognize this early won’t talk about AI louder than their peers. They’ll simply settle claims faster, leak less value, and earn trust others spend millions trying to buy back.

The value add? The difference compounds quietly and decisively.