Underwriting Automation: Loan Application Scorecard Prompt
Prompt
You are implementing AI-assisted loan screening for loans <$500k. [PASTE: LOAN APPLICATION DATA - applicant profile, income, employment, credit score, delinquency history, bank data if available, fraud indicators]. Develop scoring model: 1) Risk Segments (Low/Moderate/High), 2) Quantitative Scoring (credit score 30%, DTI 25%, employment stability 20%, bureau patterns 15%, income verification 10%), 3) Behavioral Flags (fraud red flags, financial stress signals), 4) Decision Rules (approve auto/refer/decline thresholds), 5) Pricing Adjustments (risk-based spreads), 6) Model Monitoring (approval rates by segment, backtesting, fair lending checks), 7) Compliance (prohibited bases, disparate impact, ECOA). Map FICO 620-800 to point scale. Format: decision tree with thresholds and scoring worksheet.
Why it works
Structured approach with clear methodology enables consistent decision-making and scalable execution. Documented framework supports audit, governance, and regulatory examination.
Watch out for
Context-specific application required; generic approach may miss nuances. External constraints and market conditions may limit control. Model predictions require human validation and override capability.
Used by
Finance TeamsData Analysts