Can AI Crack The Mortgage Approval Gap?

Mortgage applications processed through AngelAi at Sun West Mortgage had substantially lower decline rates in 2025 than applications in the national market, according to a new company white paper. The gap appeared across Asian, Black, Hispanic, and white applicant groups.

Sun West CEO Pavan Agarwal offered a measure of the company’s performance from one of its warehouse lenders. “One of our warehouse lenders is Truist [Bank] … and they told us that in their book, we're the only lender that's growing in volume and profitable,” he said.

Agarwal credited several factors for Sun West’s performance, including fair-lending practices and what he called “the simplicity of the AI.”

The paper’s central claim is that AngelAi’s automation makes complex applications easier to process, allowing brokers and lenders to approve more borrowers and scale more easily. “Because of the AI, we're able to scale our sales team very rapidly,” he said, adding that Sun West can train new hires to become “expert AEs” in about two weeks.

Its findings also raise an important question for lenders: How many potentially approvable loans are lost because a file is difficult or costly to complete?

Sun West Reports Far Fewer Declined Applicants

Sun West declined a far smaller share of mortgage applications in 2025 than lenders nationally across all four applicant groups examined, according to the paper’s comparison of AngelAi operating data with federal Home Mortgage Disclosure Act (HMDA) data. Relative to each group’s national rate, Sun West’s reported decline rate was approximately 88% lower for Asian applicants, 58% lower for Black applicants, 65% lower for Hispanic applicants, and 72% lower for white applicants.

Sun West reported a 4.01% decline rate for Asian applicants, compared with 32.98% nationally. Its rate was 18.55% for Black applicants, versus 44.32% nationally; 13.73% for Hispanic applicants, versus 39.10% nationally; and 9.24% for white applicants, versus 33.31% nationally.

The paper suggests AngelAi’s handling of application intake, documents, income analysis, and underwriting conditions may help explain the difference. It argues that the tools can help lenders work through files involving multiple income sources or extensive documentation with fewer delays and less inconsistent handling.

A longer view of HMDA data from the Urban Institute shows denial rates rose among all four ethnicity groups between 2021 and 2024. The gap in denial rates also increased for Black and Hispanic applicants during that period. The Black–white gap widened from 10.5 to 13.2 percentage points, and the Latino–white gap edged up from 5.7 to 5.9 points. The Asian–white gap narrowed in 2024, but the overall denial rate was higher by 0.30% for Asian applicants.

Potential For $230 Billion In Additional Volume

The paper estimates that if decline rates nationwide matched those reported by Sun West, more applications could reach approval, representing $229.9 billion in potential additional loan volume. (That figure is a hypothetical scenario, not a finding that those loans would qualify or that AngelAi would produce the same results at other lenders.)

The authors applied the decline-rates from Sun West to the estimated number of 2025 HMDA applications to show how much additional volume could be recovered for each group.

At Sun West’s reported rate, only about 4 out of 100 Asian applicants would be declined, compared with about 33 nationally — a gap of roughly 29 applications. On that basis, the paper estimates $32.48 billion in potential additional loan volume for Asian applicants.

For Black applicants, about 19 out of 100 would be declined at Sun West’s rate, compared with about 44 nationally — a gap of roughly 26 applications. The paper estimates $27.42 billion in potential additional loan volume for Black applicants.

For Hispanic applicants, about 14 out of 100 would be declined at Sun West’s rate, compared with about 39 nationally — a gap of roughly 25 applications. That equates to $39.09 billion in potential additional loan volume for Hispanic applicants.

For white applicants, only about 9 out of 100 would be declined at Sun West’s rate, compared with about 33 nationally — a gap of roughly 24 applications. That would potentially create $130.90 billion in additional loan volume.

The paper argues that additional lending could also support broader economic activity, including homebuilding, renovations, and housing-related services. To illustrate the possible scale, the authors assume that 10% to 30% of the $229.9 billion in potential loans would translate into new residential investment or related production. Hypothetically, that would amount to $23 billion to $69 billion, or 0.075% to 0.224% of 2025 U.S. GDP.

Bottom Line

Sun West’s paper makes a business case for fair lending: If AI lowers the cost of working through complex files, lenders may be able to serve more borrowers without loosening credit standards. For brokers, the question is whether that holds beyond Sun West’s own pipeline.