Sunny Ray interviews Graham Topol of MGT Insurance about rebuilding insurance for the 35 million small businesses in the US. Topol explains the company's mission to make insurance faster, simpler and better, drawing on his path through Morgan Stanley, FTV Capital and a tech-enabled brokerage called New Front. He argues that selling software to incumbents only treats symptoms, and that controlling the license and product as a carrier is what allows real change. The conversation covers what AI native means day to day, why insurance is a strong use case for applied AI, and how probabilistic models fit a probabilistic problem. Topol describes the hiring culture at MGT, including the weekly Flounder and Golden Fish awards. He also looks ahead to more accurate pricing, lower expenses, and emerging AI liability coverage. He offers advice for business owners on choosing coverage, agents and carriers, and notes that MGT is the first commercial carrier with MCP based quoting in Claude. He closes by pointing listeners to their local brokers and to MGT on LinkedIn and Twitter.
Graham Topol explains why insurance is one of the best real-world uses of AI and how MGT is rebuilding it for small businesses.
Why rebuild small business insurance from scratch instead of selling software to the incumbents?
The product hasn't changed much, but the experience is broken. At a brokerage I saw a line at the printer every day because carriers still required print, mail or fax, then took two to four weeks to respond. A baker doesn't open a shop for paperwork. Better data and a faster process help small businesses, brokers and carriers.
Why does being a carrier, not just a broker, matter so much?
If you don't control the policy, you can't change the industry, and you can only issue one if you're licensed as a carrier. I watched a $3,000 bakery policy pass through four people and take weeks. To fix that workflow, you have to control the license and the product.
What does AI native mean inside an insurance company day to day?
Applied AI works best in manual, document heavy industries, and insurance is both a big data business and a document business. Our product is a legal document with a promise to pay. AI doesn't reinvent the product itself, but it reinvents how we work to deliver that product.
When did you realize AI could price small business risk better than the old way?
I'd be careful about pricing, since everything we do stays compliant with regulators and our state filings. But reading the transformer paper at Stanford in 2018, it clicked that you need a probabilistic thinker for a probabilistic problem like insurance. The tools only got good enough in early 2025.
What kind of person thrives at MGT?
We hire more senior people who have taste and know what good looks like. They must handle ambiguity, show resilience, and act with conviction instead of waiting on a web of decision makers. We hold them accountable. Every week at all hands, someone claims the Flounder award for a mistake they learned from.
How will AI change what small businesses pay for insurance over five years?
Pricing should become more accurate, so lower risk businesses pay less and higher risk ones pay more, working with regulators. We also run lower expenses than competitors, and efficiency savings should eventually pass to policyholders. Separately, AI liability will become a real coverage need, and MGT is working on it.
What should every small business owner ask before buying a policy?
Make sure you have the right coverage for your location and business, and work with a qualified agent. Ask who the insurance company is, whether it is admitted, and what its rating is. The cheapest quote matters less than whether you're covered when a $50,000 claim hits.
What will buying insurance feel like in the years ahead?
We're the first commercial carrier to go live with MCP based quoting, so you can quote MGT in Claude through chat. I think that could become a viable channel. Brokers still have a role, and they and their clients will likely use personal AI agents to find the right policies.
Building something daring? Sunny talks to founders like this every day. Fifteen minutes to see if your story belongs on the stage.
Claim your pre-interviewSend a founder here and we will know it was you.