'Garbage In, Garbage Out': Miami Broker Warns AI Mortgage Approvals Are Failing Non-Traditional Borrowers

Written by:

Steve Marcinuk

Date:

24 Jun 2025

Share

Share via Facebook

Share via Linkedin

Share via X

Copy to clipboard

Leading Miami mortgage broker David A. Krebs says artificial intelligence and automated underwriting systems are creating a dangerous false sense of security for borrowers with non-traditional income streams.

“These automated systems can be manipulated,” warns Krebs, Principal Broker at DAK Mortgage. “If there’s any kind of bonuses or RSUs where you get stock options, they’re not calculated correctly. You input the wrong information and it’s garbage in, garbage out.”

The Hidden Failures of Automated Approvals

According to Krebs, major banks’ increasing reliance on AI-driven pre-approvals is leading to a surge in last-minute loan denials, particularly affecting self-employed professionals and business owners.

“People go to the big box guys, and they typically will take a verbal application, how much you make, how much are your expenses. They plug it in and don’t verify it,” Krebs explains. “Then they spit out a pre-approval, you go to submit an offer, and then you go through underwriting, and you get a big red X.”

This automated approach particularly fails borrowers with complex income structures, like the roughly 25% of loan applicants who are self-employed or have significant investment income. These borrowers often write off substantial business expenses, making their adjusted gross income an incomplete picture of their true financial capacity.

Why Human Underwriting Still Matters

For non-traditional borrowers, Krebs argues that experienced human underwriters remain essential. “These types of deals are not automatically underwritten through a computer algorithm platform,” he says. “You have to really roll up your sleeves. Know how to calculate your ratios, front end, your back end, debt ratios.”

He points to a recent case involving a plastic surgeon seeking a $5 million home purchase. After being capped at $3 million by a major bank’s automated system, Krebs secured 80% financing by manually underwriting the application.

“I smoothed out all the bumps. I basically mitigated all the risk that I saw associated with the deal, made sure that the doors were closed that the underwriters not going to ask for additional documents,” Krebs explains. “I delivered everything they wanted on a silver platter with a red bow on top.”

The Future of Mortgage Underwriting

While Krebs acknowledges that AI and automation have their place, he argues they should complement rather than replace human expertise. This is especially true as lending standards tighten and credit markets become more complex.

“I’ve built my book of business on these scenarios, people who go and mismatch rate with risk,” says Krebs. He estimates that about a quarter of all loans originated across the country require this kind of specialized underwriting approach that AI systems currently can’t replicate.

For borrowers with non-traditional income streams, Krebs recommends working with lenders who understand their unique situations rather than relying solely on automated approvals. “At the end of the day,” he says, “there’s no algorithm that can fully replace an experienced underwriter who knows how to structure complex deals.”