Vincent Chen spent the better part of his career at Google teaching machines how to recommend the right podcast or organize your photo library. Now he's applying that same expertise to an industry that still relies on fax machines.
Commercial insurance brokerage, it turns out, might be one of the last places in American business where "automation" means someone scanning a PDF instead of retyping it. Chen's new venture, Panta Insurance Services, is wagering that AI agents can do in hours what human brokers spend weeks doing: hunting down underwriters, filling out forms, chasing quotes. The proposition sounds simple enough. The execution? That's where things get interesting.
In his second week operating as an actual insurance broker—not just building the technology, but writing real policies—Chen's AI-powered system beat a 20-year industry veteran's quote by 38%. Whether that says more about the technology or the state of the industry is an open question. Probably both.
The Half-Trillion-Dollar Spreadsheet Problem
Here's how commercial insurance usually works: A restaurant owner needs liability coverage. Their broker gathers information—often over multiple phone calls and email chains. Then comes the data entry. The broker logs into carrier portals, sometimes a dozen of them, manually re-entering the same information over and over. They fill out ACORD forms, the industry-standard application documents that haven't changed much since the pre-internet era. They email underwriters. They wait. Sometimes for days.
Panta calls this the "human API." It's a generous description for what is, essentially, an expensive game of telephone.
The U.S. commercial lines market cleared $500 billion in premiums last year, according to S&P Global Market Intelligence. Despite that scale—and perhaps because of it—much of the broker workflow still runs on phone calls, faxes, and spreadsheets. McKinsey estimates generative AI could unlock 10-20% productivity gains across insurance operations, with the strongest impact in submissions and underwriting. The consultants might be underselling it.
For businesses that need coverage to operate legally, the waiting isn't just annoying. It's expensive. Construction projects stall. Trucking fleets sit idle. Medical spas can't open their doors.
Zero Data Entry (in Theory)
Panta's pitch is straightforward: AI agents handle the broker back-office while licensed humans maintain oversight. Data collection, application generation, submission routing, underwriter follow-ups—all automated. The company promises "zero data entry" for brokers, though that feels like the kind of claim that probably requires an asterisk or three in practice.
The platform integrates with Canopy Connect, a tool that lets businesses share existing insurance and loss data securely. Panta ingests that information, prepares submissions, and shops them across what the company describes as a network of more than 100 A-rated carriers. The focus is surplus lines and hard-to-place risks: construction trades, trucking operations, cannabis businesses, hospitality venues, medical spas. The stuff traditional insurers tend to avoid.
Coverage lines include the usual suspects—general liability, commercial auto, property, workers' comp—plus cyber, directors and officers, product liability. Panta says it's licensed in all 50 states. Public records confirm at least a California surplus lines broker license (#4512382). Whether the rest of those licenses exist is harder to verify, which is perhaps telling about how fragmented insurance regulation remains.
Chen, in a LinkedIn post, described how an AI agent "autonomously secured" insurance for an explosive oil and gas laboratory. One of the riskier placements you can make in commercial insurance. The company says it wants to write space insurance policies in 2026, which seems ambitious for a team that just got its broker license.
From Foundation Models to ACORD Forms
Chen spent years at Google building foundation models for Vertex AI and agentic systems deployed in consumer products like NotebookLM and Google Photos. His co-founder, Frank Wang, came from Apple—a senior full-stack engineer with open-source work in Rust and Gleam. Not the obvious background for insurance brokers, but then again, maybe that's the point.
The two founders pivoted through multiple AI projects in 2025. CallZero.ai. CallOS.ai. Others. Y Combinator rejected their first application. After deep interviews with brokers—the kind of customer discovery startup advisors always recommend and founders often skip—they reapplied with what Chen later called a "broken" demo.
YC accepted them anyway for the Winter 2026 batch. Sometimes a compelling thesis matters more than working code.
The team numbers three to four people, operating out of San Francisco. CB Insights lists a $500,000 convertible note with Y Combinator as an investor, though Panta hasn't confirmed funding details publicly. For context: that's seed-stage capital, barely enough to keep the lights on for a year in the Bay Area.
Everyone Else Had the Same Idea
Panta isn't alone in believing AI can transform commercial insurance brokerage. Harper, an AI-native brokerage, launched in March 2025 and reported more than $6 million in annualized premiums across 35 states within months—impressive traction, though still a rounding error in a half-trillion-dollar market.
Newfront, a YC alum from way back in Winter 2018, built its own AI capabilities—GPT-powered document parsing, automated contract review—before agreeing to be acquired by Willis Towers Watson in December 2025. That acquisition, more than any pitch deck, validated the thesis that automation would reshape brokerage economics.
On the software side, Applied Systems rolled out "Applied Insurance AI" across agency workflows in October 2025. Patra offers AI-powered policy checking and quote comparison. Casey, another Y Combinator company from Fall 2025, automates broker submissions. Even Pathpoint, a digital excess and surplus lines wholesaler with years in the market, added AI-powered quoting tools. Everyone saw the same opportunity at roughly the same time.
The regulatory environment is catching up, too. The National Association of Insurance Commissioners adopted a Model Bulletin on AI use in December 2023, establishing expectations for governance, documentation, and vendor oversight. Multiple states have since implemented similar frameworks, though enforcement remains spotty.
Meanwhile, some insurers began excluding broad AI-related liabilities from commercial policies in late 2025, adding another layer of complexity for brokers advising tech clients. The irony—insurance companies worried about insuring AI companies while simultaneously deploying AI themselves—has not been lost on anyone paying attention.
Service Business with Software Dreams
Panta's website promises 10x faster turnaround times on complex quotes. It positions the brokerage as a "service business with software economics," the kind of phrase that sounds appealing to venture investors who prefer software margins to service headcount.
Whether AI agents can truly replace the expertise, relationship capital, and regulatory judgment human brokers bring to complex placements is the question everyone in this space is trying to answer. A 20-year veteran didn't spend two decades learning how to fill out forms. They built relationships with underwriters who pick up the phone. They developed instincts about which risks to place where. They learned which carriers are hungry for premium and which are pulling back.
Can an AI agent replicate that? Maybe eventually. But probably not in 2026.
Still, in an industry where speed and pricing increasingly determine who wins accounts—especially for commoditized risks—Panta's bet is that automation will matter more than tradition. Chen's early metrics suggest there might be something there. A 38% better quote in week two isn't nothing.
The real test will come when the technology fails. When an AI agent misreads a submission and a client ends up underinsured. When a regulatory audit reveals gaps in oversight. When an underwriter relationship that took years to build gets burned by an automated follow-up email that didn't quite strike the right tone.
Those moments will determine whether Panta is building the future of commercial insurance brokerage or just a faster way to do what spreadsheets and phone calls already accomplish. For now, though, Chen and his tiny team have Y Combinator's backing, a broker license, and a handful of bound policies.
In insurance, that counts as a promising start. Even if it's not quite space insurance yet.
