Two founders, both named Luke, are pointing AI agents at the least glamorous corner of insurance - wholesale placement - and getting brokers coverage in minutes instead of days.
The insurance industry runs on a promise: if something goes wrong, somebody pays. What it rarely promises is speed. A retail broker with a hard-to-insure client - a trucking fleet, a warehouse in a flood zone, a business nobody wants to touch - often waits days for a single price. Emails go out. PDFs come back. Portals time out. Follow-ups pile up. Somewhere in that fog, the customer gets impatient and walks.
Luke Button watched it happen. At his previous insurance brokerage, he says the team waited an average of eight days for a quote, and lost customers because of it. That specific frustration - not a market-size slide, not a trend report - is the seed of Hedge, the AI-native specialty insurance wholesaler he co-founded in 2026 with Luke Rosa. The company is part of Y Combinator's P26 batch and based in San Francisco.
To understand Hedge, you have to understand the layer it lives in. When a risk is ordinary - a small office, a delivery van - a retail broker places it with a standard carrier and moves on. When a risk is unusual, it enters the world of excess and surplus lines, or E&S. Here, retail brokers don't deal with carriers directly. They go through a wholesaler, a specialist middleman who knows which obscure market will write which strange risk.
Wholesalers are the connective tissue of specialty insurance, and they are famously manual. Hedge sits in exactly that seat - between the retail broker and the specialty carrier - but rebuilds the workflow around software. Its AI agents read the broker's submission, select which markets to approach, complete the carrier applications, negotiate quotes, prepare binders, issue documents, and even handle regulatory filings. A small team of humans directs and supervises that work through an internal command system, rather than typing it all out by hand.
The pitch to a broker is concrete: coverage placed roughly 50% faster and 15% cheaper, with a first response that lands in about 30 minutes instead of the better part of a work week. On the risks that usually get ignored, Hedge claims a hit ratio about three times higher than the manual status quo.
Speed is not a vanity metric in this business. In specialty insurance, the broker who quotes first often wins the account, because the end customer rarely wants to wait around. So the gap between eight days and thirty minutes is not a nicer experience - it is the difference between keeping the client and losing them. That is the wedge Hedge is driving into.
The demand response has been blunt. Within roughly eight weeks of launch, Hedge was working with more than 50 brokers, and by its own account the interest came back faster than a two-person team could keep up with. That is why the careers page reads less like a startup and more like a hiring spree: founding underwriters, founding wholesale brokers for trucking, property and commercial auto, engineers, and salespeople.
The origin story has a coincidence baked in: the company was co-founded by two people who share a first name. Luke Button, the CEO, is a repeat Y Combinator founder. He previously built Fernstone, an AI-powered retail insurance brokerage, which makes Hedge less a pivot than a second, sharper swing at the same industry. Before that he led growth at Traba, a Founders Fund-backed labor marketplace, was head of growth at Antimetal, and co-founded Contrast, which was acquired.
Luke Rosa comes at the problem from the systems side. As head of scaled systems at Traba, he replaced human processes with AI agents that managed operations tied to more than $20 million in revenue. That is the exact muscle Hedge needs: not chatbots, but agents that can be trusted with real money and real paperwork. He studied economics at the University of Chicago.
Hedge's market access spans the full spread of E&S lines. That means commercial property, including catastrophe-exposed buildings; general and excess liability; cyber and technology errors-and-omissions; employment practices; environmental and pollution; commercial auto; management liability; inland marine; and medical malpractice - among others. The unifying theme is not an industry but a difficulty level: thin files, distressed accounts, and risks that standard carriers decline.
On the experience side, Hedge leans on things brokers actually complain about missing. Each account gets a dedicated, named broker rather than a rotating cast of contacts. File notes are live, so everyone can see where a submission stands in real time. Support runs seven days a week. It is a deliberately old-fashioned service promise wrapped around a very new engine.
Hedge makes money the way wholesalers do - commissions on the coverage it places. The bet is that automating the placement grind lets it move more volume, win more hard risks, and do it with a far smaller headcount than the incumbents carry. The stated long game is to fold underwriting in alongside distribution, so that a single AI-native operation both finds the risk and prices it.
The competition is not a plucky field of startups. It is the entrenched giants of specialty distribution - the Amwins, RT Specialty and CRC Groups of the world - who move enormous premium volume and still, in Hedge's telling, do much of the quoting by hand. That is the whole argument in one line: the incumbents are big, profitable, and manual, and Hedge is betting the third word is the one that matters.
Where does it fit in the market? Squarely in the middle, by design. Not a retail brokerage chasing end customers, not a carrier holding the risk - a distribution layer that turns a broker's messy submission into a bound policy. If the AI-native approach holds, that middle seat is a good one: high-friction, high-margin, and largely untouched by software until now.