🔒 Score, tier, and policy action are all computed in this browser tab from the trained model's exported coefficients - zero network calls. The one exception is "Generate AI case summary" below, which calls a Netlify Function; the Anthropic API key lives only in that function's server environment and never reaches this page.
Underwriter Workbench
Pick a case from the queue. The risk score and policy recommendation are deterministic; the AI summary only narrates them and is checked against the same tool data before it's shown to you.
Case queue
Select a case from the queue.
Policy thresholds (the $35,000 large-loan escalation, the 35%/40% DTI cutoffs) are illustrative for this demo and are not real LoanTap policy.