hyperexponential reposted this
One year ago today, I joined hyperexponential. The people, culture, remit all mattered. But this was also a thesis-driven bet: if any kind of work was well suited for AI agents, commercial P&C underwriting was one. AI helps with a lot of everyday things. But a full agentic transformation only makes sense when the work is high-stakes, complex, and, at the same time, governable. Underwriting is all three. 📈 The stakes are obvious. Even a slightly better underwriting decision can move loss ratios, retention, and profitability in a big way. In a market with roughly $1 trillion in global commercial P&C premium, small improvements compound quickly. 🧩 The work is genuinely complex. Commercial underwriting is not an “if this, then that” logistics problem. Every account brings its own mix of business risks, pricing tradeoffs, and portfolio considerations. The numbers matter, but they only tell part of the story. Meanwhile, speed to decision is critical. Slow answers lose business. 🛡️ And at the same time, the work can be governed. Insurance carriers have spent decades building rules around underwriting to control risk. The opportunity now is to apply those same controls to agents. Sure, a lot easier said than done 🙂 Nothing about portfolio strategy, compliance rules, audit trails, pricing logic, authority thresholds, or the underwriting context scattered across systems, workflows, and institutional judgment comes out of the box when you plug in an AI model. That was the second part of the thesis. If agents are going to do meaningful underwriting work one day, they have to be built into the real machinery of underwriting: the data, rules, models, controls, and decision history that carriers already use to run the business. That’s where hyperexponential comes in. The company has spent years helping carriers encode underwriting logic, controls, and judgment into production workflows, with $60+ billion in premium already running through the platform. One year in, I’ve more confidence in this thesis than ever.