An Industry Overdue for Disruption
Insurance is one of the largest industries in the world, and yet it remains one of the least touched by modern technology. Many of the dominant car insurers have been operating for more than a century, running on technology stacks that are decades old. For customers, that stagnation is unfortunate: it means higher prices, slower service, and experiences that have not kept pace with the rest of their digital lives. The opportunity, then, is not to incrementally improve a legacy system but to rebuild insurance entirely from the ground up around artificial intelligence and machine learning.
That ambition is hard to execute precisely because insurance is so difficult to enter. Building a carrier means clearing significant regulatory hurdles, deploying capital intensively, and accumulating enough claims data to construct genuinely predictive models. Crucially, you have to be a licensed insurance company to gather that data in the first place. These barriers are exactly why the industry has gone so long without meaningful disruption — and why a company built on AI from its first day is positioned to be the disruptor rather than the disrupted.
Why Technology Drives Both Growth and Profitability
The clearest evidence that an AI-native approach works is financial. Posting the most profitable quarter in a company's history — while still growing significantly against a difficult market backdrop, and after already sustaining profitability for two years — demonstrates real differentiation rather than a temporary tailwind.
The engine behind that performance is pricing. At its core, insurance pricing is a prediction problem: forecasting who will get into an accident and who will not. Modern quantitative techniques and AI are, fundamentally, advances in predictive science, and they can be applied directly to this question on a modern technology infrastructure. The key insight is that risk is highly concentrated. Roughly ten percent of drivers cause the majority of accidents. If a company can reliably identify and avoid that small group, it can offer a genuinely better deal to everyone else. That model delivers real, measurable benefits to customers — and as long as the focus stays on the customer, the financial results tend to follow.
A Data Advantage That Compounds
What makes this approach durable is the behavioral data underlying it. Signals drawn from smartphones, from new vehicle technologies, and from autonomous vehicles allow the modeling of specific behaviors — speeding, hard braking, and other patterns — and how each one changes the likelihood of an accident. The system can even account for how an over-the-air software update to a vehicle alters its risk characteristics. Because nearly every step of being an insurance carrier has been automated end to end, the company can move far faster than incumbents to incorporate these signals.
This creates a compounding dynamic. Each new market produces fresh claims data, which feeds the models, which learn, which enables faster and safer growth, which generates still more data. The result is an exponentially growing data moat — an advantage that widens over time and is exceptionally difficult to replicate for any organization still in the middle of modernizing decades-old technology stacks.
Embedded Distribution as a Second Moat
Pricing is only one source of differentiation; distribution is another. Embedding insurance directly into the vehicle purchase process — selling more than 200,000 policies through an exclusive partnership with Carvana — has proven to be a powerful, delightful customer experience. When coverage is offered at the moment a car is bought, the experience is seamless, and customers in that channel often are not shopping on price at all. This builds a second moat around the business: in these partnerships, the embedded insurer is frequently the only insurance company present, giving it differentiated access to customers. That partnerships channel has grown thirty percent year over year, and there is active expansion of similar tie-ups beyond the initial relationship.
A Long Runway for Growth
The growth story is far from complete. In the United States, insurance is regulated state by state, which is why operations currently span 36 states — about 80 percent of the population. The plan is to march toward national coverage by the end of next year, with New Jersey expected to launch this year. The arithmetic here is striking: moving from 80 percent of the population to 100 percent represents roughly a 25 percent growth lever on its own, entirely separate from gains within existing markets.
Geographic expansion does come with a learning curve. Entering a new state means starting to collect claims data, feeding it into the models, and letting them learn before growth accelerates. For that reason, new markets are entered cautiously, with expansion building over the following few quarters. This deliberate pace is itself a sign of confidence rather than caution about the underlying model.
Confidence Reflected in Capital Allocation
Markets do not always react rationally to strong results, and a sound strategy is not to try to predict them — a company that could forecast markets would not be in the insurance business in the first place. The more productive posture is to treat market reactions not as a verdict on intrinsic worth but as an opportunity to act on. A $75 million share buyback program is exactly that kind of action: a direct expression of confidence in the business and a willingness to own more of it when others hesitate.
Taken together — predictive pricing built on compounding behavioral data, embedded distribution that bypasses price competition, a clear path to national scale, and disciplined capital allocation — these elements describe a business that views its strongest quarter not as a peak but as a beginning. The deeper point is that an industry left largely untouched by technology for over a hundred years is finally being rebuilt around it, and the advantages of doing so accumulate rather than erode.