Panta Runs an Insurance Brokerage With AI Agents Instead of Account Managers

The Macro: Commercial Insurance Is a $1 Trillion Industry Powered by Email

Commercial insurance is enormous. Over $1 trillion in global premium every year. And the way it works has barely changed in decades. A business needs coverage. A broker gathers information from the business, formats applications for carriers, negotiates terms, and coordinates paperwork across multiple parties. The entire process runs on email, phone calls, and PDF attachments.

The Excess and Surplus (E&S) market makes this even worse. E&S handles risks that standard carriers will not touch. Construction, cannabis, hospitality with liquor liability. These placements require specialized knowledge, relationships with wholesale markets, and patience. A typical E&S placement can take weeks of back-and-forth between the broker, wholesaler, and carrier.

The bottleneck is not information. It is coordination. Everyone involved has the data. The problem is moving data between systems, formatting it for each carrier’s requirements, following up on quotes, comparing terms, and managing the paperwork. This is exactly the kind of structured, repetitive, rule-bound work that AI agents should handle well.

The Micro: A Google ML Engineer and an Apple Full-Stack Builder

Vincent Chen and Frank Wang founded Panta. Vincent was a Machine Learning Engineer at the parent company of YouTube, leading research on foundation models for Vertex AI, NotebookLM Enterprise, and Ask Photos. Frank was a Senior Full Stack Engineer at Apple and is a top-10 contributor to the Rust and Gleam open-source ecosystems. They have a four-person team from San Francisco, part of YC Winter 2026 with Gustaf Alstromer.

Panta operates as an actual licensed insurance brokerage, not just a software tool. They are licensed across all 50 states with a 99% placement rate across 100+ A-rated carriers. The product covers general liability, commercial auto, workers’ comp, property, excess liability, cyber, D&O, and product liability.

The AI agents handle the entire back-office workflow: gathering application data, formatting submissions for carriers, managing quotes, comparing terms, and coordinating the placement process. Human approval gates and audit logs maintain regulatory compliance.

They have a California Surplus Lines Broker license (#4512382), which means they can handle the complex E&S placements that standard brokerages find difficult. The infrastructure includes integration with Canopy for importing existing insurance policies.

The Verdict

Panta is doing something unusual: using AI to run an actual brokerage, not just to sell software to existing brokers. The advantage of this model is that Panta captures the full margin instead of sharing it with broker customers. The risk is that operating a brokerage requires regulatory compliance, E&O insurance, and the kind of operational overhead that software companies do not usually deal with.

The competitive field includes established insurtechs like Bold Penguin, Pie Insurance, and Next Insurance. But most of these focus on specific lines or specific customer segments. Panta’s full-line approach with AI-powered back-office automation is differentiated.

In 30 days, I want to see the average placement timeline. How fast does Panta get a quote compared to a traditional broker? In 60 days, the question is loss ratios. If Panta’s AI is placing risks effectively, loss ratios should be comparable to or better than industry averages. In 90 days, I want to know about premium volume growth. An insurance brokerage is ultimately measured by GWP (gross written premium), and the growth rate tells you whether the AI-first model is working.