Risklytics (YC S26) Killed Its First Idea Mid-Batch and Became an Insurance Brokerage. Would YC Fund Your Services Business?

The latest Launch HN on the Hacker News front page is not a SaaS product, an AI agent, or a developer tool. Risklytics (YC S26) is an insurance brokerage. Two founders on leave from Harvard, Sam and Alex, place insurance coverage for companies building robots, drones, autonomous systems, and satellites. They earn a standard broker commission of 10 to 20 percent from the insurer and charge founders nothing.

If you are drafting a YC application right now and your idea is not software, this launch answers a question applicants ask constantly: does Y Combinator actually fund services businesses?

What Risklytics actually does

The pitch is simple. A frontier tech company's first customer or pilot usually requires insurance coverage before the machine is allowed on site. But insurers evaluate these companies with application forms written for ordinary businesses, and the safe answer for the person reviewing an unfamiliar risk is no. One Risklytics client was denied coverage for using CAD software in their workflow. Another, a robotics wholesaler, was denied because some of the property they rent out uses AI, even though they never use it themselves.

The timely part: in January, ISO, the industry body whose standard forms most insurers license, published add-on clauses that void coverage for losses arising out of AI in any part of a process. Each insurer decides whether to adopt those clauses account by account, and there is no public record of who has. The wording is unsettled too. It defines generative AI as systems trained on data that produce text, images, audio, video, or code. Whether a robot's control model, which produces motor commands, falls under that definition is an open question. Risklytics says underwriters give different answers across the board.

So their product is partly software (a six-step application that translates what you are building into the structured form insurers expect) and partly accumulated knowledge: a map of which insurers add the AI clauses, which will remove them, and which cover frontier companies on purpose.

The part most applicants will miss: the pivot happened inside the batch

Risklytics did not apply to YC as a brokerage. They came into S26 with a wildfire risk model they wanted to sell to insurance companies. In their own words, insurers liked it and wanted to build their own, "which is the polite way to say no."

While that idea was dying, they kept hearing the same uninsurable story from hardware companies, got licensed as brokers in a week, and started placing coverage. In the week before their launch they got three companies covered that had been denied elsewhere, including a tele-operated robotics service that needed coverage within a week for its first pilot.

We have written before about how YC accepts teams, not ideas (Intuned got in with a completely different idea). Risklytics adds the lesson that post misses: the new idea came from a batchmate's unsolved problem. A founder who launches satellites could not get insured, and asking around turned one anecdote into a pattern. Founders consistently underrate proximity to other founders as an idea source, which is a large part of what YC is actually selling.

Does YC fund services businesses?

Yes, and it always has, as long as the service has a wedge that compounds. The clearest precedent is Zenefits (YC W13), which gave away HR software for free and made its money as a licensed insurance broker collecting commissions. The software was the distribution; the brokerage was the business model. Risklytics is running a close cousin of that play thirteen years later, aimed at frontier tech instead of small business HR.

Paul Graham has a name for why these opportunities stay open: schlep blindness. Insurance licensing, state-by-state credentialing, reading policy exclusions line by line. These are exactly the tedious, unsexy problems most technical founders refuse to look at, which is why the ones who do face little competition. Stripe is his canonical example in the payments schlep; brokerage for uninsurable robots is the same shape.

What is much harder to fund is a services business with no leverage: an agency where revenue scales linearly with headcount and nothing accumulates. The useful test is whether something compounds as you grow. For Risklytics, every application they process deepens their map of which insurers will actually cover AI risk, an asset no incumbent broker is building and no new entrant gets for free.

How should you pitch a services business on your YC application?

Name the compounding asset explicitly. Not "we provide a service," but "every transaction teaches us X, which nobody else is recording." If your margin depends on a commission structure, state the take rate plainly the way Risklytics does; vagueness about how you get paid reads as not understanding your own industry.

Show the regulatory moat as a feature, not a bug. "We got licensed in a week" is a schedule fact that doubles as a character reference. If your business requires licenses, certifications, or state filings that scare other founders off, that is your moat forming, and partners know it.

And be honest about what is service and what is software. YC partners read thousands of applications and can tell when an agency has been dressed up in platform language. The Zenefits framing is the honest version: the software earns distribution, the service earns revenue, and you should be able to say which is which in one sentence.

This is also the kind of claim worth stress-testing before you submit: the YC alumni who review applications at YC Roaster flag the services-dressed-as-SaaS pattern constantly, because it is one of the fastest ways to lose a reader who has seen the trick before.

The takeaway

Risklytics got into YC with one idea, watched it die, and rebuilt as a commission-based brokerage in the second half of the batch, and that is the version they just launched to the world. If your idea is a services business with a real wedge, the precedent is on your side. What matters is not whether you sell software or service. It is whether something compounds, whether you move at licensed-in-a-week speed, and whether you can say what your asset is in one sentence.