Seven commercial underwriting systems and three specialty line extensions — all anchored to The Underwriter's Decision Stack. Built in public. Each produces a working system, GitHub repository, whitepaper, and engineering offer.
A five-agent AI pipeline that reads a commercial property submission — broker cover note, schedule of values, loss history, risk engineering survey summary — and produces a structured Underwriting Brief with a routing recommendation. Eliminates 45-90 minutes of manual extraction per submission without losing the unstructured intelligence that manual processes routinely miss. Built on Python and Claude with a human checkpoint before any routing decision.
Wraps existing pricing workflows connecting the model to portfolio comparables, reinsurance economics, and a structured deviation capture mechanism that makes pricing rationale auditable for the first time.
Maps claims experience to underwriting decisions at inception. Identifies portfolio patterns before they appear in loss ratios. Produces renewal briefs closing the underwriting-claims loop.
Converts 40-page survey reports into queryable underwriting intelligence. Compares current against prior surveys. Validates risk grades against loss history.
Real-time treaty consumption monitoring, net pricing at point of decision, accumulation alerts, and fac placement intelligence. Visible before bind, not at month-end bordereau.
Live accumulation monitoring, drift detection, pre-bind impact simulation, and appetite calibration. From retrospective dashboard to forward portfolio steering.
Structured judgment capture at point of decision, consistency analysis, bias detection, model drift monitoring, governance reporting. Transforms the audit trail from a filing system into a learning system.
The Underwriter's Decision Stack applies beyond commercial property. These three projects extend the framework to parametric insurance, renewable energy underwriting, and data centre underwriting — each following the same build-in-public methodology.
AI-assisted parametric trigger design — trigger identification, 20-year backtesting, basis risk quantification, payout structure optimisation. Applicable to weather, climate, catastrophe, agriculture, renewable energy, and index-based structures.
AI-augmented underwriting intelligence for solar, wind, and BESS — irradiance and wind resource signals, technology risk grading, EML/OAR, BESS thermal-runaway scoring, coverage condition recommendations.
AI-augmented underwriting intelligence for data centre risks — Tier scoring, SLA-based BI quantum, equipment breakdown profiling, geographic accumulation, coverage condition recommendations.