Insurance is really just a big data problem

Michael Topol, Co-founder and Co-CEO at MGT Insurance, explains why insurance is quietly becoming one of the most interesting data and AI problems in tech.

We get practical about turning messy legacy data into usable signals, how agentic tools change decision making, and why culture and team design matter as much as the models.

MGT Insurance is building a fully verticalized AI and agentic native insurance company for small businesses, pairing experienced insurance operators with top tier technologists. Michael breaks down what changed in the last few years that makes real disruption possible now, and what modern product delivery looks like when prototyping is cheap and iteration is fast.

Key takeaways

• Insurance is a data business at its core, but most incumbents cannot use their data fast enough because it lives across silos, mainframes, and old systems.

• Modern AI lets teams combine internal data with public signals to speed up underwriting and improve consistency, without losing human judgement.

• Vibe coding and rapid prototyping collapse the gap between idea and implementation, bringing product, engineering, and the business closer together.

• Senior talent gets more leverage in an AI driven workflow, and small teams can ship faster by focusing on problem solving, not just building.

• Pod based teams, fixed outcome planning, and strong culture help regulated companies move quickly while staying inside the rules.

Timestamped highlights

00:44 What MGT Insurance is, and what “AI and agentic native” means in practice

02:09 Why small business insurance matters more than most people realize

06:06 The real blocker for incumbents, data exists but it is not usable

08:55 Vibe coding in a regulated industry, where it helps first

12:54 Requirements are shifting, prototypes bring teams closer to the real problem

17:26 The pod structure, plus the Basecamp inspired approach to scoping and shipping

20:52 Better, faster, cheaper, why AI finally makes all three possible

22:11 Where to connect, and who they are hiring

A line you will remember

“Insurance is really just a big data problem.”

Pro tips you can steal

• Build cross functional pods early, include a domain expert, a technical product lead, and a senior engineer from day one.

• Scope for outcomes, not perfect specs, then let the team decide the depth as they build.

• Use AI to automate collection and synthesis, then keep humans focused on the decisions and trade offs.

Call to action

If you enjoyed this one, follow the show and share it with a builder who is trying to ship faster with a smaller team.