Look, I read something from HousingWire that really struck a nerve, and I want to walk you through it because it touches on something we all feel in this business. The industry is spending about $11,898 to produce a single mortgage these days, which is way up from the historical average. According to the reporting, we've basically spent billions on digital tools over the past fifteen years, yet somehow the cost of originating a loan has climbed about fifty percent. We digitized the old processes without really rethinking them, and borrowers are paying for that inefficiency through their rates.
Here's where the money actually goes. You've got loan officer commissions, branch management cuts, lead generation costs, and then the big one: a whole lot of people doing repetitive work. The reporting points out that two-thirds of what it costs to make a loan is just personnel expense. People retyping information from one system into another, people checking what other people checked, and so on. It's costing lenders money too. I've seen files that generate good revenue but still lose money because the infrastructure is so heavy.
The real problem, according to this piece, is that the people who actually know how mortgage work live on one side of the organization, and the people building the technology live on the other. The loan officers have all the expertise locked in their heads, all the edge cases they've learned to handle. But when they leave, that knowledge walks out the door. Meanwhile, the software engineers are building systems based on secondhand descriptions of how things should work. Nobody's sharing the real knowledge.
The idea being floated is something called a loan officer engineer. This is someone who's a licensed originator but also helps improve the system itself. Half the job stays the same: working with borrowers, making judgment calls, being the person people trust when things get complicated. The new half is writing down how you solve hard problems in a clear, structured way so the system can apply that solution consistently to every similar file. It's not about replacing judgment with automation. It's about turning hard-won expertise into something the company learns from permanently instead of losing it when someone walks out the door.
What strikes me about this is the economics of it. Right now, a loan officer's value is capped by their own hours. They can originate more loans, but that means nights and weekends. If that same person is also building the system's knowledge, then their expertise becomes part of something bigger. The company gets smarter with every loan they touch, especially the complicated ones. And that means lower costs, faster closings, and fewer deals that get rejected just because they're too expensive to handle.
What I am seeing locally is that borrowers in the Bay Area and East Bay are getting squeezed harder as closing costs climb. Self-employed folks, mixed income situations, anything that doesn't fit the standard boxes, those are getting shut out or overpriced. The current system treats complexity as a cost problem instead of as an opportunity to build better systems. If loan officers could encode their knowledge and make it repeatable, we'd stop losing deals we could actually do, and we'd stop charging people premium rates for situations we've already solved a hundred times. That's what matters to me.
