Mortgage AI in Lending & Secondary Markets, Promise & Risk

Mortgage AI is promising for automating labor-heavy steps across the loan lifecycle, but it needs rigorous testing and ongoing oversight to avoid introducing new risk, SitusAMC’s Julia Curran told HousingWire. She said leaders must treat mortgage AI as use-case specific rather than a single technology, warning that a proof-of-concept test with 27 vendors found 26 couldn’t reliably handle tasks needed for secondary-market reviews. Curran emphasized the need for subject-matter experts, testing across a wide variety of scenarios using a ground-truth database covering hundreds of reviewed loans, and regression testing after model updates. Bias testing is also critical. She said AI can help with inputs to LOS, POS, and servicing systems, while human judgment still matters.