Friday Harbor is expanding its AI pre-underwriting platform to bank-statement and certain jumbo loans, addressing a growing operational problem for lenders: determining whether increasingly common complex loans work before they consume underwriting resources.
The Seattle-based mortgage technology company said Tuesday that its platform can now calculate qualifying income for bank-statement borrowers, analyze loan files against investor-specific guidelines, identify documentation and eligibility problems, and anticipate potential underwriting conditions.
The expansion moves Friday Harbor beyond reviewing documents for completeness. The company is attempting to give production teams an earlier indication of whether a complex borrower qualifies under the rules of the investor expected to purchase the loan.
Non-QM loans represented more than 10% of mortgage rate locks in July, according to Optimal Blue data cited by Friday Harbor, more than triple their share when Optimal Blue began tracking the category in 2022. Optimal Blue’s Market Advantage data are drawn directly from its product, pricing, and eligibility engine, which covers more than one-third of U.S. mortgage transactions.
NMP reported last month that conforming loans had fallen below half of mortgage production for a third consecutive month in June, while jumbo and Non-QM products continued gaining share.
“With more than one in 10 loans falling outside traditional agency guidelines, lenders can no longer afford to treat these loans as edge cases,” Friday Harbor founder and CEO Theo Ellis said.
Moving The Decision Closer To The LO
Bank-statement loans allow lenders to evaluate income through a borrower’s deposits rather than relying solely on tax returns or conventional employment documentation. They are commonly used for self-employed borrowers whose taxable income may not reflect the cash flow available to repay a mortgage.
But determining qualifying income can require reviewing months of deposits, separating business and personal cash flow, applying expense factors, identifying one-time deposits, and following calculation methods that vary by investor.
That complexity frequently forces an originator to consult an account executive, scenario desk, or specialized underwriter before knowing whether the borrower qualifies.
“Bank statement loans are a great example,” Ellis said. “The income calculation can be complex enough that loan officers often need to pull in an account executive or someone with specialized expertise just to determine whether the loan works.”
Friday Harbor said its platform performs that income calculation while reviewing the broader loan file. It then compares the documents and loan data with applicable investor guidelines, flags missing or inconsistent information, and identifies issues that could become underwriting conditions.
The platform does not replace formal underwriting or guarantee that a loan will be approved. Its value proposition is earlier visibility: giving the originator and lender an opportunity to correct a file, select a better-fitting program, or abandon an unworkable scenario before it moves further into production.
AI Moves Deeper Into The File
Friday Harbor’s expansion goes beyond extracting bank-statement data or calculating income. The platform is designed to compare the broader loan file with investor-specific guidelines, flag inconsistencies, and anticipate potential conditions before underwriting.
That builds on the trend that lenders are using AI to reduce the manual work that makes complex Non-QM loans more expensive to originate.
The next test is whether these tools can reliably reconcile borrower documents, lender overlays, and investor requirements.
Scaling Non-QM Without Scaling Bottlenecks
As Non-QM volume grows, lenders cannot send every complex income calculation or eligibility question to a small group of specialists without increasing costs and turn times.
Earlier review could improve pull-through by identifying unworkable files before lenders and originators invest more time in them. Friday Harbor, however, did not disclose results showing how the new capabilities affect accuracy, underwriting conditions, pull-through, or closing times.
The expansion follows the company’s AI governance compliance attestation in May, an important consideration as lenders face greater scrutiny over AI outputs, vendor oversight, and human accountability.
For originators, the promise is not merely faster income calculations. It is knowing earlier whether a complex borrower and the intended loan program fit together well enough to produce an underwritable file.
