April 9, 2026
Two stories are reshaping the competitive landscape of insurance right now. One is about what becomes possible when you build a carrier from scratch around AI — and regulators actually approve it. The other is a slow-motion talent exodus that most of the industry is still treating as a hiring problem. Read them together, and you start to see the actual shape of the disruption coming — and what it demands of every carrier and distribution platform in the market.
The insurance industry has heard the phrase AI-powered so many times it has almost lost all meaning. AI-powered claims. AI-powered underwriting. AI-powered quote flows. For most carriers, this has meant layering machine learning on top of legacy systems while keeping the organizational structure — and the regulatory identity — fundamentally intact.
That changed in early 2026, when Corgi became the first AI-native, full-stack insurance carrier to receive regulatory approval to operate as a licensed carrier. Not a managing general agent. Not a fronting arrangement. A carrier — designed from day zero around AI, not retrofitted with it.
This is not a distinction of branding. It is a distinction of architecture. And for distribution platforms like Walnut, it is a signal worth paying close attention to.
What makes the AI-native carrier model structurally different is the absence of legacy drag. No broker-heavy distribution models, no manual workflows, no annual policy cycles baked into systems built in the 1980s. McKinsey has warned that when incumbents retrofit AI onto legacy systems, they risk creating "tomorrow's legacy" — complexity layered on complexity. The AI-native carrier sidesteps this entirely. The governance framework, the data architecture, the decision logic — all of it is designed from scratch for a world where AI is the operating model, not a feature.
Its product suite signals something else: AI-native carriers are not just built with AI, they are built for the AI economy — offering coverage products like AI liability that legacy carriers are still figuring out how to underwrite. According to Gallagher Re's Q4 2025 Global InsurTech Report, 77.9% of all insurtech funding in Q4 2025 flowed to AI-centered companies, with $3.35 billion deployed across AI-focused deals for the full year. Capital is not hedging here — it is making a directional bet.
For Walnut, this development has a specific implication. As more AI-native carriers enter the market with faster quoting, adaptive coverage, and real-time pricing, the distribution layer becomes even more critical — and more differentiated. Platforms that can move at the speed of AI-native carriers, and connect seamlessly with their infrastructure, will capture distribution relationships that slower, form-heavy flows will lose. The architecture of the carrier side is converging toward speed and API-first design. The question is whether the distribution side keeps pace.
Now the harder one — and the one that deserves far more serious strategic attention than it's getting.
The numbers are stark and well-documented: the U.S. Bureau of Labor Statistics projects approximately 400,000 insurance professionals will leave the workforce through retirement and attrition by 2026. In London alone, more than a quarter of underwriters are now over 50. Following COVID, experienced professionals left early while internship and graduate programs were simultaneously frozen — creating a knowledge gap at both ends of the talent pipeline.
The standard industry response has been to frame this as a staffing problem: recruit faster, rebrand the industry for Gen Z, run campus programs. Those are legitimate efforts. But they fundamentally misread the nature of the risk.
What is walking out the door is not headcount. It is judgment.
As Swept AI's analysis lays out, when a senior claims adjuster with 25 years of experience retires, the loss is not a vacancy on an org chart. It is the accumulated knowledge of which body shops inflate estimates, which injury patterns in specific jurisdictions lead to litigation, which contractor networks in specific regions are unreliable. That knowledge exists in no database and no policy manual. It is not captured in historical claims data. And if you automate the workflow before you capture the judgment, the AI learns from incomplete inputs — and nobody notices until a loss pattern emerges the model never learned to recognize.
This is where the talent exodus collides catastrophically with AI deployment timelines — producing what is arguably the most underanalyzed structural risk in insurance right now: AI governance failure by omission.
Here's the failure mode in plain terms. A carrier trains an AI underwriting model on historical data. That historical data reflects the decisions of expert underwriters who are now retiring. The model learns the statistical patterns — but not the contextual overrides. Not the jurisdictional experience. Not the "this submission looks standard but something is off" instinct that took two decades to develop. The AI deploys. The experts retire. The governance framework signs off on a model built on incomplete foundations.
KPMG's 2025 Insurance CEO Outlook, surveying 110 insurance CEOs globally, captures the tension precisely: 73% are prioritizing AI investment to streamline underwriting and claims — while simultaneously, 77% identify workforce transformation and upskilling as a top constraint on growth, and 56% cite ethical challenges and data readiness as the primary obstacles to scaling AI. These three numbers belong together. The confidence in AI investment and the concern about talent and data quality are not separate issues. They are the same issue.
The IIS 2026 Global Priorities Report adds another dimension: for the first time in five years, changes in regulation surpassed cybersecurity as the top political and legal concern among insurance executives — specifically AI governance regulation. Regulators are watching the AI deployment race and asking whether carriers can explain and defend the decisions their models make. If the people who understood why certain decisions were right are gone, defending those models in an examination becomes very hard.
Risk & Insurance's analysis of the talent gap frames the downstream stakes clearly: the institutional knowledge lost when these workers leave could have major impacts on compliance, customer service, and strategy. And at the same time, the types of roles the industry needs are shifting — toward data science, AI literacy, and technical translation between business and technology teams. The gap is not just getting wider; it is changing shape.
These two forces — the rise of AI-native carrier infrastructure and the mass retirement of institutional knowledge — are not separate trends. They are the same story told from opposite directions.
AI-native carriers can build governance into the architecture from the start, with no inherited knowledge gaps to manage. Traditional carriers hold the data depth and regulatory relationships that newer entrants lack — but they are racing to deploy AI at precisely the moment the people who understand the domain are leaving. McKinsey notes that established insurers have deep reserves of structured historical data that give them a genuine advantage — if they can activate it before that knowledge disappears into retirement.
The carriers who close the gap will be the ones who treat the 400,000 problem not as an HR crisis but as a knowledge engineering imperative. Concretely, that means:
Shadow AI deployment before replacement. Run AI systems alongside experienced professionals, comparing model decisions to expert decisions on the same cases. A six-to-twelve month shadow period generates a dataset of expert-vs-model disagreements that is more valuable than any historical claims database — because the disagreements are exactly where institutional knowledge diverges from what the model learned from statistics alone.
Behavioral knowledge capture. Convex's research with behavioral scientists found that what experienced underwriters thought were five steps in their process were actually 15 nuanced decisions. That kind of documentation — done before retirement, not after — is what makes AI models defensible when regulators ask how they were built.
AI governance as a competitive asset, not a compliance checkbox. As KPMG's Jacques Cornic emphasizes, carriers need an inventory of every model using AI, and a clear description of how data is being used, to earn trust in the output. The carriers who build that framework now will be the ones regulators trust to move faster later.
As underwriting talent thins at the carrier level, and AI models take on more decision-making, the quality of submission data and application completeness becomes more important — not less. Platforms that help businesses submit cleaner, richer risk data upstream reduce friction for AI-driven underwriting and become preferred distribution partners. The best response to an industry knowledge crisis is not to wait for carriers to solve it — it is to build workflows that reduce the burden on underwriters, experienced or otherwise.
Wipfli's 2026 Insurance Industry Outlook puts the strategic choice plainly: firms that use AI to capture and preserve fast-vanishing institutional knowledge will outperform those that treat it purely as an efficiency tool. And industry forecasters project insurance AI spend to grow by more than 25% in 2026 alone — meaning the decisions being made right now about how to deploy AI, and on what knowledge foundations, will compound for years.
The 400,000 problem is already past tense in parts of the market. The AI-native carrier is already licensed and operational. The industry is not approaching a tipping point — it is past one. The question for every player in the value chain — carrier, MGA, distribution platform — is whether your strategy was built for the industry that existed, or the one that is being built right now.