AI can read the mortgage file, but who decides which facts are true?

Mortgage technology is moving quickly toward a future in which artificial intelligence can read guidelines, analyze borrower data, calculate income, identify conditions and participate in increasingly consequential decisions. That progress is real. But underneath it sits a problem the industry has not solved nearly as well: When multiple legitimate sources describe the same borrower differently, who determines which facts can actually be relied upon? Consider a borrower with variable overtime income. Payroll data shows one monthly figure. The loan origination system shows another. Year-to-date earnings suggest something different. Tax transcripts reflect an earlier period, while bank activity supports part of the picture but not all of it. None of these sources is necessarily wrong. Each may accurately describe a particular fact, period or purpose while still presenting a different view of the same borrower. A source may be authoritative for what it records without being sufficient, by itself, for the downstream fact a lender needs to establish. AI can read all of it. It can compare the records, summarize the differences and infer what probably happened. But mortgage lending requir...