Viewpoint: The ‘Unbundled’ Insurer Emerges in Age of AI as Value Chain Fragments

A roofing contractor in Florida buys a general liability policy. Between him and the capital that stands behind the policy sits a supply chain: a retail agent, a wholesale broker, an underwriter, a licensed carrier, a reinsurance panel and, increasingly, a pension fund or a credit manager who has never heard of the contractor. The chain runs from the customer at one end to capital at the other, and its only job is to connect the two as efficiently as possible. For about a hundred years, most of the ecosystem lived under one roof. A carrier underwrote the risk, issued the paper, paid the claims, held the data and put its own surplus behind the promise. Historically, the cheapest way to coordinate expensive, difficult work was to put all of it inside one company. “The insurance value chain is separating into capital providers focused on investing and service firms that handle underwriting, distribution, program administration and claims.” In 1999, insurance consultants notably predicted that would end: underwriting, distribution, administration and capital would separate, creating virtual insurance companies. They were not wrong; they were just early. The insurance value chain has been fragmenting for years, and AI is already accelerating the breakdown. Carriers Embrace MGAs A carrier is a bundle of underwriting, licensed paper, claims, data and capital. The components are slowly unbundling with the underwriters first. Conning’s 13th annual MGA study describes a “migration of underwriting expertise” from carriers to MGAs. MGAs wrote roughly $128 billion of U.S. premium in 2025, up 12% in a year where the market only grew 5%. MGAs’ share of U.S. premium grew from 3.6% in 2012 to about 11.6% in 2025. Carriers are embracing MGAs, and momentum shows no signs of slowing. In Conning’s survey, 61% of insurers said they are expanding their use of MGAs and none said they plan to cut back. That growth is also fueling demand for fronting carriers, which provide the licensed insurance capacity that connects underwriters with capital. Today, 78% of MGAs use a front to access capital. Read more: MGAs: An All-Weather Distribution Channel for Soft and Hard Markets The model is longstanding, but its scale is changing rapidly. Fronts grew almost 400% in five years from $6 billion in premium in 2020 to over $22 billion in 2025, according to Conning. (Editor’s note: Some of report’s findings, including the section of the growth of fronts, sit behind Conning’s paywall). Large carriers like Chubb and Berkley will continue to manage everything under one roof, but the marginal dollar is being deployed through the new, intermediated model. While still a rounding error relative to the global insurance market, fronting is becoming commonplace. What’s Left Is a Balance Sheet When underwriting and paper is pulled out of a carrier, what remains is a balance sheet and investment portfolio. The carrier begins to look more like an investment manager than an insurance company. In June, Howard Hughes Holdings, Bill Ackman’s new venture, closed its $2.1 billion acquisition of Vantage and handed the insurer’s investment portfolio to Pershing Square. In September, The Fidelis Partnership, described itself as the world’s largest independent MGA, reported $5.39 billion in premium for 2025 while retaining almost none of the underlying risk on its own balance sheet. One is a capital manager that rents underwriting. The other is an underwriter that rents capital. Insurance linked securities (ILS) have been an important tool to bring new capital into the market as capacity. Sidecar capital, packaged ILS, reached $23 billion this summer, up roughly 50% in just a year and a half. The capital funding sidecars is not insurance capital, but rather institutions looking for alpha. Stone Point’s Jim Carey said at Monte Carlo: insurers “are starting to see this as a really efficient capital management play… They’re as much evaluating the fixed income asset manager as they are evaluating the insurance partner.” Conceptually, none of this is new. It is Warren Buffett’s play: collect premium now, pay claims later, and invest the float in between. GEICO is the operating half of that machine, a direct writer with a 12.4% expense ratio last year compared with the industry’s 25.8%. In 1967, Buffett kept underwriting and capital under one roof because that was the only way to get the float. However, in 2026, all you need is a fronting agreement and capacity. Carriers are blending into asset managers while underwriters and brokers assume more of the operational burden of the insurance supply chain. Considering where the next wave of operational leverage will come from, carriers are the last place to look. 20-Year Hiatus The fragmentation of the traditional carrier was a real theory with strong backing but nothing happened for 20 years. Conning’s assessment captures the industry’s shift, “the failure of the virtual insurer was not a failure of concept, but of conditions.” Data was slow, contracts were crude and alternative capital was a rounding error. In Conning’s words, “what strategists imagined as a coordination problem was, in reality, a technology problem. The cost of connecting separate functions remained prohibitively high.” Economists call this transaction costs: when coordinating through the market costs more than doing it in-house, you end up with a firm; when those costs fall, the firm unbundles and specializes. Since 1999, a unified model has made sense. Now, the world has also gotten meaningfully more complicated since 1999. Catastrophe losses have topped $100 billion for six consecutive years in 2025. Nuclear verdicts and litigation finance have pushed casualty severity in one direction for a decade. Headwinds complicate the supply chain making it harder to break the components apart. Ostensibly, a firm should make even more sense today than two decades ago. Over the last five years though, something has shifted. Technology has addressed the gaps that made the fragmented supply chain change seem impossible two decades ago. Data is more structured, usable and transferable than ever before, which makes it easier to structure contracts in a way that align incentives between underwriting and capacity. Technology advancements opened the door to the first major wave of MGA and fronting carrier growth. And AI is about to accelerate it. Last Expensive Step The insurance value chain is separating into capital providers focused on investing and service firms that handle underwriting, distribution, program administration and claims. Service firms bear the operating costs, with human capital as the largest line item. A human insured calls a human retail agent, who emails a human wholesale broker, who re-keys the submission for a human underwriter, who asks a question that travels back down the chain. A typical brokerage spends over 60% of their revenue on human capital. Evan Greenberg’s critique, that risk now passes “through four or five layers of intermediaries, who all take commissions,” is right as a description of cost. But it misses one piece: the cost of human capital is decreasing and that’s making the intermediated supply chain cheap. Software can now transact with humans as humans; an example is the read of a messy submission, ask the clarifying question, draft the response, reconcile the bordereau, all without a person in the middle of every step. Humans are not removed from the process but can now handle a higher volume of accounts while delivering a better standard of service. In Novella’s case, we have built a brokerage with almost no service team. AI agents regularly handle endorsements, cancellations and binding. As more AI native companies emerge across the supply chain, the costs of selling a policy will plummet. We spent years as an industry figuring out how to structure our data to benefit from APIs. It turns out that APIs were a temporary solution. Technology can now understand unstructured data as effectively as structured, and most insurance IT teams have yet to realize this. Instead of building an API, an AI agent can read documents, ask follow-up questions, understand nuances of risks, and submit a risk through a portal or email. As these capabilities become embedded in everyday operations, the structure of the supply chain will be completely transformed. Some companies will specialize on a singular function, while others will integrate data across the chain in the way that major players like Aon or Amwins do today. When specialized human talent is paired with specialty trained AI, better outcomes are available for lower costs: more accurate underwriting, faster broking. These gains are unfolding at the operating layer, not the carrier layer. The promise of AI is operating leverage, the modern carrier is searching for investment alpha. The opportunity is to reinvent the insurance supply chain with AI at the core and focus on the operations, not the investing. Where This Ends Up The sole job of the insurance supply chain is to distribute risk to capacity. For a hundred years, the industry kept these functions within a single company because it was the most cost-effective way to operate. This is no longer the case. Specialist firms are detaching operations from balance sheets, technology is driving down the cost of connecting them, and the roofer in Florida ends up with a lower premium and an easier time securing coverage. It isn’t yet known whether all of this holds through a market cycle. Conning’s own test is casualty, where a delegated underwriting mistake can hide for years, and that test is being run right now. What is apparent is that the 1999 prediction was right about everything except timing. Capital is abundant. Underwriting is scarce. Coordination and operations are about to be cheap. Topics InsurTech Data Driven Artificial Intelligence Carriers