Tesora Is Building What Harvey Built for Law, but for Insurance Underwriting

The Macro: Insurance Is a Trillion-Dollar Industry Running on Email and Excel

I have a theory about which industries get disrupted last. It is not the ones with the most regulation or the most money. It is the ones where the expertise is hardest to externalize. Where the knowledge lives in people’s heads and in informal processes that nobody has ever written down.

Insurance underwriting is a textbook example.

Here is what a typical underwriting workflow looks like at a managing general agent or carrier. A submission comes in. An underwriter opens the email, downloads the attachments, and starts pulling data from five or six different systems. They cross-reference loss runs, check policy forms, look at historical claims data, consult pricing models in Excel, and make a decision based on a combination of quantitative analysis and gut instinct. The gut instinct part is important, because a lot of the decision-making criteria live nowhere except in the head of the person who has been doing this for twenty years.

When that person retires or leaves, the knowledge walks out the door with them.

The insurance industry generates over $6 trillion in global premiums annually. The underwriting process is the engine that determines whether that money gets deployed wisely or poorly. And yet the tools underwriters use have barely changed in two decades. Guidewire and Duck Creek handle policy administration. Majesco and Sapiens do some of the heavy lifting on the actuarial side. But the actual decision-making workflow? That is still spreadsheets and email and phone calls and tribal knowledge.

Harvey proved that AI could encode expert legal workflows in a way that actually works at scale. They did it by going deep on legal-specific training data and building interfaces that lawyers would actually use. The question is whether the same approach can work for insurance. The answer, I think, is yes. But the company that does it needs to understand both the technology and the domain. Insurance is not a market where you can fake domain expertise.

The Micro: A McKinsey and Google Team That Actually Understands Insurance

Vivek Rao (CEO) comes from McKinsey and Sycamore Partners, a private equity firm managing roughly $14 billion in assets. He is a UChicago and Booth graduate. That background matters because it means he has spent time inside the kinds of organizations that Tesora is selling to. He knows what an MGA’s operations look like from the inside, not from a demo deck.

Federico Reyes Gomez (CTO) has a BS and MS from Stanford in Computer Science and Philosophy. He was an engineer on Google’s Document AI team, which is about as relevant as prior experience gets for this problem. Document AI is literally about extracting structured data from unstructured documents, which is a core component of what underwriters do all day. Before Tesora, he joined Kumo, a Sequoia-backed startup.

Tesora came through Y Combinator in Summer 2025. The team is currently two people.

The product connects disparate systems, encodes expert workflows, and eliminates the spreadsheet and email bottlenecks that constrain growth at MGAs and carriers. That description sounds generic until you think about what it actually means in practice. It means Tesora has to ingest submission data from multiple formats and sources, understand the context of that data, apply underwriting guidelines that vary by line of business and by company, and surface recommendations that underwriters trust enough to act on.

The “trust” part is the hard part. Underwriters are not going to hand over pricing decisions to a black box. They need to see the reasoning. They need to be able to override it. And they need the system to learn from those overrides.

I have not seen detailed product demos or customer case studies yet. The website is live but light on specifics, which is normal for an enterprise B2B company at this stage. They are clearly in the early innings of customer acquisition and product development.

What I find compelling is the specificity of the problem. This is not “AI for everything in insurance.” It is AI for underwriting and actuarial workflows, which is a focused enough wedge to build deep domain expertise while still representing an enormous market.

The Verdict

Tesora is early. Two-person team, limited public information about customers or traction. But the thesis is strong and the founders have the right combination of domain knowledge and technical ability.

The competitive landscape includes legacy players like Guidewire and Duck Creek who are bolting AI features onto existing platforms, plus newer entrants like Federato and Sixfold who are also going after AI-powered underwriting. The question for Tesora is whether being AI-native from day one gives them a structural advantage over companies that started with traditional software and are adding AI later. I think it does, but only if they ship fast enough to establish relationships with carriers before the incumbents catch up.

Thirty days, I want to see a design partner or two. Sixty days, I want to hear about a pilot generating real underwriting decisions. Ninety days, the question is whether the encoded workflows are actually better than the spreadsheets they replace. If underwriters are faster and more accurate with Tesora than without it, this company has a real shot at becoming critical infrastructure for the insurance industry.