AI Underwriting: Frequently Asked Questions

AI underwriting uses machine-learning models to evaluate loan applications, learning default patterns from historical data instead of relying on fixed rules. It analyzes more variables, finds nonlinear relationships, and typically predicts default risk more accurately than traditional scorecards.

A traditional scorecard typically achieves an AUC around 0.72; Underwrite.ai models routinely reach 0.82 to 0.85. But the metric that matters is lift over your incumbent FICO-based model — and one installment lender cut its first-payment default rate from 32.8% to 8.5% after switching to machine learning.

Underwrite.ai evaluates 60+ approaches including XGBoost, LightGBM, decision trees, random forests and support vector machines, selecting the best performer for each portfolio. Models are supervised binary classifiers.

No. Large language models are never used to make credit decisions. An LLM may be used only to translate a model’s feature attributions into a plain-language explanation. The LLM explains; it never decides.

Yes. Underwrite.ai models are fully explainable. Techniques such as SHAP report which variables drove each decision and by how much, producing adverse-action reasons. Underwrite.ai does not build black-box models.

Yes. Models are designed to avoid protected characteristics and their proxies, are tested for disparate impact, and comply with FCRA, ECOA and GDPR. Underwrite.ai provides disparate-impact analysis tools.

No. Underwrite.ai does not use social media data or any data that could be considered discriminatory.

An anonymized loan tape and application tape — historical applications linked to their loan outcomes. Underwrite.ai builds a custom model calibrated to that portfolio, with a 30-day free trial to evaluate performance.

Milliseconds. A UK auto lender reduced decision time from 2–3 days to 2–3 milliseconds while maintaining prediction accuracy above 98%.