Representative Engagements

CrossCheck was engaged by a credit union to assess whether the use of its AI-supported underwriting solution altered fair lending outcomes across the full credit decision process. The analysis reviewed application outcomes from initial AI score assignment through automated decisioning, referral logic, and final manual review. Using proxy-based prohibited basis analysis and regression testing, we evaluated whether denial disparities emerged, increased, or declined at different stages of the underwriting funnel, allowing the client to see where risk was introduced, mitigated, or carried forward.

In addition to measuring overall disparities, we isolated marginally qualified applications and analyzed outcome transitions to determine whether AI testing results, credit union decision thresholds, or manual review processes affected similarly situated applicants differently. This analysis included comparing predicted and actual outcomes, identifying underwriting funnel decision points that warranted deeper review, and isolating how the credit union’s traditional underwriting criteria interacted with AI-enabled processes. The engagement gave the client a practical framework for ongoing AI solution monitoring and governance by explaining how AI model outputs, operational rules, and human intervention worked together to shape fair lending risk.