Prediction Lab is an AI-native predictive-modeling platform for insurance pricing. Generative AI already changed how software gets built. We're bringing that same shift to how insurers build the models they price on.

A pricing actuary's real job isn't producing a model. It's understanding the book well enough to set rates, defend a filing, and explain to the business why risk is moving and what a rate change does across the portfolio. Today's tools can't answer those questions. Legacy tools hand you a model fast but won't let you ask why, iterate freely, or control how the model gets built. Seeing the coefficients isn't the same as understanding the book. The alternative is writing your own code, which can answer anything but turns a pricing team into a software team maintaining a fragile stack it was never meant to own. So the insight, the pricing decision, and the filing all wait.

We built Prediction Lab so humans and AI agents model together. We expose the modeling primitives, the same controls an actuary uses, as guardrailed tools an AI agent can operate. The agent stays grounded: it can't drift the way it would on open code, and it has real tools to work with, unlike a chatbot bolted onto a black box. It does the work that used to take modelers longest and surfaces the insight. The actuary keeps full control through those same primitives, and anyone in the org can interrogate a model in plain language instead of waiting on the modeling team. The actuary stays the expert; the AI does the grunt work.

Why now: Google says 75% of its new code is AI-generated, 90% of developers use AI daily, and a whole new tool layer got built around the work. The frontier moved from chatbots that answer to agents that act, and agents get reliable when they're scoped to real tools. Predictive modeling is making the same move, from building models to designing the analysis and acting on the insight, and it needs the same new kind of tool.

Traction: our design partner, a P&C carrier, builds pricing models on Prediction Lab today, alongside their previous primary tool. A first model that takes a day in their existing solution takes about 15 minutes on Prediction Lab. The product is live and SOC 2 certified.

Market: we sell per modeler seat at about $25k. That's roughly 15,000 P&C pricing seats globally, a focused ~$380M market. Each account expands through the same buyer into the rating engine and reserving workflow, and from there into the $20B+ predictive-modeling market.

Team: founder Ryan Steil is an actuary, data scientist, and engineer in one. He built insurance models at Liberty Mutual, ran actuarial engineering at Hippo, and led actuarial innovation at PURE. He used every platform on the market and built Prediction Lab because none of them fit. He's bootstrapped it solo to a live product since January 2025.

We're raising a seed round to convert our pilot to revenue, land the next carriers, and ship the platform's next layer.