On a Tuesday in San Francisco, software is arguing with an underwriter - politely.
Aspecialty carrier has just received a submission for a mid-market property risk in a wildfire-exposed county. The underwriter opens Federato. The system has already triaged the file, scored it against the carrier's appetite, plotted where it would sit in the existing book, and surfaced two endorsements the broker forgot to mention. None of this is dramatic. That, the company would tell you, is exactly the point.
Federato sells software to insurance carriers, MGAs, mutuals and program administrators. It calls the category RiskOps. It did not invent the underwriting workflow - underwriting has been around since Lloyd's of London poured coffee in 1688 - but it did finally write the missing layer between the strategy document on the executive's desk and the spreadsheet on the underwriter's screen.
RiskOps is the framework that connects underwriting decisions to portfolio strategy in real time.- Federato, on its own product category
Insurance has more data than almost any industry. It uses less of it.
That is the contradiction Federato exists to resolve. Carriers sit on decades of policy data, loss data, exposure data and reinsurance data. They also still bind a non-trivial number of risks because someone replied "looks fine" to a Tuesday email. The gap between what the executive committee approved in last quarter's appetite memo and what the underwriter actually does at 4:47 pm on a Friday is the most expensive document in the industry.
Until very recently, the proposed fix was either (a) a thicker memo, or (b) a more expensive consultant. Neither worked. The memo gets archived. The consultant leaves. The Friday email keeps happening.
The most expensive document in insurance is the gap between strategy and the next policy bound.- Editor's note
Two Stanford wildfire researchers, one whiteboard, and a hunch about where the data lives.
Will Ross and William Steenbergen met in graduate school at Stanford. Their subject was wildfire modeling - which sounds like a hazard-science problem until you remember that wildfires also burn houses, and houses are insured. The two co-founders had spent a decade in and around agentic AI before the term made anyone money. In 2020 they sketched the idea for Federato on a single sheet of paper.
The bet was unglamorous. Do not pitch insurers a smarter model. Pitch them a smarter workflow with the model already inside it. Train the AI on insurance-specific data. Embed it directly into the workbench. Let it get smarter with every policy bound. The thesis - that the surface area for AI in insurance was the underwriter's desktop, not a separate analytics product - is the kind of thing that sounds obvious in retrospect and was, in 2020, not obvious at all.
Will Ross
Co-Founder & CEO. Fifteen years around AI. Has been known to describe what he is doing as "Uberizing" insurance - a flattering analogy or a warning, depending on which side of the steering wheel you sit on.
William Steenbergen
Co-Founder & CTO. Co-author of the original wildfire research and the technical architecture underneath RiskOps. Owns the federated-data plumbing nobody sees but everyone depends on.
Founded
March 2020. The pitch was on a sheet of paper. The first round closed two years later. The company is now 170 people.
Headquarters
2261 Market Street, San Francisco. Remote-friendly. The product roadmap is decided in calls that span six time zones.
What the underwriter actually sees.
The product is a workbench. New submissions arrive and get triaged automatically - sorted by appetite fit, winnability, accumulation impact and broker history. Risks the carrier does not want are quietly declined or kicked back. Risks that look good get the human treatment, with the data already pulled, the appetite check already done, the portfolio implications already plotted on a chart the underwriter can read in three seconds.
Underneath, agentic AI is doing the boring work. It pulls third-party data. It checks against the appetite. It tracks policy changes through renewal. It flags accumulation when a single hurricane could light up four counties at once. None of this is magic. It is the same job a junior analyst would do, performed at machine speed and without the part where the analyst has to ask their manager what the appetite is this quarter.
AI that triages submissions, flags risks, checks appetite, and surfaces recommendations across the entire policy lifecycle.- Federato product overview
A Federato timeline, in five rounds and one product category.
The carriers - including a $17B one - say it works.
QBE North America, the U.S. arm of a global carrier with around $17B in gross written premium, said working with Federato moved its operations "two decades forward in a matter of weeks." That is the kind of quote a marketing team would write for itself if it could get away with it. The quote is in fact attributed to the carrier. Insurate, a workers' comp data-science MGA, calls Federato "a trusted partner." Ledgebrook uses it to attract broker submissions and underwriting talent.
Reported customer outcomes are blunt: a 90% improvement in time-to-quote, a 3x improvement in the proportion of good business bound, and a 50-90% reduction in the number of separate systems an underwriter has to touch in a day. Underwriters who have used Excel and Salesforce in the same hour will understand what that last number means.
What customers say changed.
The bigger argument: insurance becomes a real-time discipline, or it becomes obsolete.
Federato's stated mission is to help insurers grow profitably by aligning underwriting decisions with portfolio strategy in real time. That is a careful sentence. It does not say "replace underwriters." It does not say "disrupt insurance." It says, in plain language, that the act of binding a policy and the act of managing a portfolio should not live in different software, and that closing that gap is, in fact, the entire game.
It is also a sentence with consequences. If the gap closes, the role of the underwriter shifts from data wrangler to judgment-maker. The role of the chief underwriting officer shifts from memo-writer to system-designer. The role of the broker shifts because the broker can finally see, at submission time, whether a deal has any chance of being bound. None of these shifts are small.
The role of the underwriter shifts from data wrangler to judgment-maker. None of these shifts are small.- YesPress, on what Federato is actually doing
Back to Tuesday, San Francisco - and the polite argument.
Return to the underwriter at her workbench. The submission is still there. The wildfire risk is still real. The portfolio is still finite. What has changed is that none of the relevant facts are stuck inside someone else's spreadsheet anymore. The appetite is live. The accumulation is live. The recommendation is live. The argument the software is having with her is not adversarial - it is just complete. She binds the policy or she does not, and either way the portfolio knows.
Five years in, this is what Federato sells. Not a smarter model. A finished argument. The question for the rest of the industry is whether they want to be part of the argument, or be the subject of it.