loading...
Corgi reached a $4B valuation in under two years by replacing the underwriter entirely — not automating them. The opportunity isn't competing with Corgi. It's building the infrastructure a category this new still lacks.
Insurance underwriting is a data processing job dressed up as a professional service for a century. A business applies for coverage. An underwriter reviews the application manually, requests documents, runs the numbers through a model built in the 1990s, and issues a decision in weeks. Corgi removed the underwriter.
Nico Laqua and Emily Yuan founded Corgi in 2024, both from YC S24. The pitch was straightforward: take a D&O application, feed it to an AI model trained on commercial insurance filings and claims history, and return a bindable policy in under five minutes. Same-day coverage for risks that previously required a two-week underwriting cycle.
What followed was unusual even by 2024 standards. Three funding rounds closed in eight weeks. A $108M Series A at a $630M valuation. Then $160M Series B at $1.3B. Then $106M in a B1 extension at $2.6B. A fourth round was reported in July 2026 at a $4B valuation. The company disclosed $40M ARR in May 2026 and announced a target of $450M by year-end.
Corgi started with management liability lines — D&O, Tech E&O, EPLI, Cyber — categories where the risk is information-dense and the underwriting manual is essentially a structured data extraction problem. It has since added trucking and, notably, AI liability insurance: coverage for losses caused by AI model errors. The AI liability product matters for reasons beyond market size. No actuarial history exists for this category. Every policy Corgi writes is simultaneously a product and a data collection instrument. The company building the claims dataset in year one defines pricing for the decade that follows.
The marginal cost of underwriting one additional policy on AI infrastructure approaches zero. That is not an incremental improvement on the incumbent model — it is a different business entirely.
Traditional carriers hire underwriters, train them for years, and pay them to evaluate individual risks. Corgi's cost structure looks nothing like this. As volume scales, the per-policy economics improve rather than degrade. The $40M to $450M ARR target represents ten-times annual growth, not expansion into adjacent markets.
The AI liability line is the more durable competitive position. First-mover ownership of the claims dataset that will define this category is a data moat, not a market share lead. Competitors who enter later inherit whatever pricing Corgi's experience base has already established. The product Corgi is building is part insurance carrier, part risk research institution.
The lead opportunity is not another AI insurance carrier. It is AI error attribution infrastructure.
For Corgi to price AI liability accurately — and for any insurer that follows — someone has to build the tooling that tracks AI failures, attributes causation, and produces actuarial-grade claims data. This does not exist in any standardized form. The gap is methodological: there is no agreed framework for determining whether a loss was caused by a model hallucination, a bad prompt, an integration failure, or user error. Every AI insurer needs this infrastructure. None of them will build it internally — the incentives point to a neutral third party.
The opportunity is picks-and-shovels positioning on a new risk category: a platform that monitors AI system behavior, logs failure events with structured metadata, and generates the kind of documentation that survives a claims process. Revenue model is SaaS subscription plus per-claim data licensing. The company that owns the attribution standard owns the category.
A secondary opportunity is vertical expansion of AI underwriting into industries Corgi has not yet entered. Trucking is one signal that the model generalizes. Construction, restaurants, and specialty trades share the same structural characteristics: high application volume, definable risk parameters, slow incumbent underwriting cycles, and price-sensitive customers who would accept self-serve coverage if the process were fast enough. The technology Corgi built for management liability translates with vertical-specific training data. The constraint is distribution, not the underwriting model itself.