Risklytics opened a brokerage for the risks AI exclusions leave behind

Why it matters

AI systems are moving into customer workflows and physical sites faster than insurers can accumulate loss data. Risklytics is betting policy review and carrier access can become critical infrastructure for those deployments.

On August 13th, Samuel Gold and Alexander Risio launched Risklytics, a San Francisco commercial insurance brokerage built for companies deploying AI agents, autonomous systems and machines at customer sites.

The two founders are on leave from Harvard, according to Y Combinator's company profile. Gold, Risklytics' CEO, studied computer science and economics. Risio, its CTO, worked on computational genomics at Harvard Medical School and previously co-founded Primeshio, an AI-assisted 3D modeling product for Blender, and Arckis, a public-sector data project.

Their pitch starts with an unglamorous constraint on the physical AI boom: insurance policies can determine whether a robot gets deployed, an enterprise contract closes or a loss leaves the buyer holding a bill it thought it had transferred.

Risklytics says conventional commercial forms were written before AI agents started performing customer work and robots began operating away from their manufacturers' facilities. The brokerage takes a client's operations to carriers willing to consider the risk, then reviews the forms and exclusions before coverage binds.

A human broker behind the AI intake

Risklytics uses AI at the front of the application process. A customer answers three questions about its business and where its technology touches the physical world. Software turns those answers into a draft application, which the customer reviews before submission.

The binding decision remains with people. Risklytics says a licensed producer reviews each submission, places it with carriers and examines the returned policy language. Nothing is automatically bound online.

That distinction matters in insurance, where an application error or overlooked endorsement can become consequential only after a claim. Risklytics is selling interpretation and placement, with automation handling the paperwork around those services.

The brokerage lists general liability, professional liability, cyber, directors and officers, commercial property, equipment, workers' compensation and umbrella coverage. Its packages map roughly to a frontier-tech company's operating stages: an early prototype or pilot, systems deployed at customer locations and a scaled fleet requiring higher limits.

Risklytics cautions that those packages are application templates. Actual coverage remains subject to carrier appetite, underwriting, licensing, policy terms and availability.

The founders found the gap inside insurance

Gold and Risio told YC that they were already building insurance models when they noticed peers developing frontier technology struggling to find appropriate coverage. They argue that insurers respond to unfamiliar risks by adding exclusions while they wait for claims data, leaving technology companies to deploy before the insurance market has caught up.

The founders compare that stage of the market to the early development of cyber insurance: exclusions arrive first, followed by specialists and eventually products built around the new category of risk. Risklytics is trying to become one of those specialists before physical AI deployments produce enough loss history for carriers to price them routinely.

The founders have chosen brokerage rather than underwriting the risk themselves. Risklytics earns commissions from carriers when it places a policy and says it does not charge applicants an additional fee. That model limits the capital required to start selling, while making carrier relationships and policy expertise central to the product.

Risklytics is part of Y Combinator's Summer 2026 batch, where YC lists it as a two-person company. YC's published standard deal invests $500,000 in each accepted company: $125,000 for 7% and $375,000 through an uncapped most-favored-nation SAFE. Risklytics has not announced a separate financing round or valuation.

Frontier-tech insurance is already becoming a startup category

Risklytics is entering a market with other young companies working at different points between technology builders and insurers.

RiskCube, a YC Fall 2024 brokerage, sells commercial coverage to startups in AI, robotics, space and defense. Its pitch emphasizes speed, including certificates of insurance that it says can often be issued within a day. Risklytics puts more emphasis on translating unfamiliar operations for carriers and finding exclusions inside the policy forms that come back.

Valgo, from YC's Winter 2026 batch, is working further upstream. It builds probabilistic models intended to help insurers estimate losses for autonomous systems that lack decades of historical claims data. Risklytics is focused on placing policies available in the market today.

Those companies reflect the same underlying problem. Robotics and autonomy startups are moving from controlled demonstrations into warehouses, roads, industrial sites and customer facilities. Every new deployment adds contractual insurance requirements and exposures that standard software-company packages may handle poorly.

Risklytics' near-term test is practical: whether its carrier access and form review can produce policies that satisfy customers, investors and procurement teams without quietly removing the AI-related risk that prompted the purchase.

Risklytics' public materials also show a second thesis. Gold's LinkedIn launch announcement described Ember, a beta model intended to score natural-disaster risk property by property instead of relying on broad postal-code judgments. Risklytics' current website and YC profile center on commercial brokerage for frontier technology. Risklytics is therefore presenting property-risk modeling and frontier-tech brokerage under the same name, with the brokerage now occupying the storefront.

The sharper opportunity may be the one Gold and Risio are pursuing today. AI intake can shorten an application, but their business ultimately depends on old-fashioned broker work: understanding what a client built, finding a carrier willing to cover it and reading the contract before the robot leaves the building.