An insurance product is metadata + rules. A property/fire line is modeled as data (line → ISO form → provisions — the individual coverages/clauses → options), and three kinds of rules evaluate as you configure a policy: compliance (regulatory, mandatory), constraint rules (data-validity bounds on fill-in values — a violation blocks submission) and risk / underwriting (carrier appetite — the risks the insurer is willing to take). A climate risk model scores wildfire / flood / wind by address and feeds the underwriting decision — exactly where an ML cat-model (catastrophe model — software that estimates disaster losses) plugs in. The provision/option defaults you see when a form loads are not hand-typed — they are seeded from prefilled property data (assessor records, aerial-imagery AI, protection class — the area's fire-department & hydrant grade) — you confirm or override them as part of policy design. How that data is bought, stored and mapped: prefill integration spec ↗ · where the service sits on the platform: architecture — Property Data Prefill ↗. Mock data only, ISO-standard form names (HO-3, DP-3 — ISO is the industry body that publishes standard policy forms). A Zen Algorithms product — sammuthu.com.