The mortgage industry’s AI problem is not what you think it is

Every conference I go to right now has the same conversation happening in every room. AI is going to transform the mortgage industry. AI is going to eliminate jobs. AI is going to make everything faster, cheaper, smarter. And maybe all of that is true, eventually. But I keep sitting in those sessions thinking we are spending a lot of energy on the wrong question.

The question everyone is asking is: How do we adopt AI as fast as possible? The question worth asking is: What has to be true about your business before AI actually helps you?

When the market tightened in 2022, the conventional wisdom was to cut. Trim the operations team. Outsource processing. Run lean and wait for things to turn. It was a rational response to a brutal environment, and a lot of companies did it.

But here is what that decision actually produced: companies now patching together workflows across multiple vendors, training new staff from scratch and trying to deliver a consistent client experience through a system never designed to hold under pressure. The overhead went away for a while. The problems came back compounded.

The companies best positioned to use AI well are not the ones that moved fastest on technology. They are the ones that never let their operational foundation fall apart in the first place. AI amplifies what you already have. If what you have is a patchwork, that is what you are amplifying.

We asked ourselves during that period: If we cut here, what breaks? Almost every answer led back to the client experience or a referral relationship; the two things that are nearly impossible to rebuild once they’re gone. A referral partner who leaves quietly rarely tells you why.

The number that lenders need to watch is pull-through rate. When operations are fragmented, pull-through suffers and pull-through is the metric your referral partners actually care about, even if they never say it in those words. Protecting the people and the process that keeps that number strong should be a priority. An astute lender who implements the use of AI responsibly will never have to call a Realtor to apologize for a deal that fell apart because of internal chaos.

The real problem with how we’re talking about AI

Most of the AI conversation in mortgage treats technology as the solution to a business problem, when in most cases the business problem came first.

A five-minute phone call with a borrower can change the entire structure of a loan. They mention something in conversation that no intake form captures, and suddenly you’re looking at a completely different product or timeline. AI cannot replicate that yet. The companies treating it like it can are going to find out the hard way.

Deploying AI should be intentional, not to adopt it for the sake of adopting it. Internally, our COO and CFO are using AI to transform reporting and business tracking; so what used to take days now gets done in a fraction of the time. That’s a real operational advantage that doesn’t make it onto conference slide decks, but it compounds.

On AI underwriting: The accuracy rates look impressive on paper, but impressive is not the same as complete. If the output still requires a human to review every decision, and it does as of now, you haven’t removed the human from the process. You’ve added a step. We’ve run trials. No one has cracked this yet. When someone does, we’ll be ready to move. Until then, we’re waiting for the solution that’s actually finished, not just promising.

What smaller companies should actually be doing

The largest lenders have a structural advantage in AI that is not going away. A small shop cannot out-AI a mega-lender, and trying to probably isn’t the best use of resources.

What smaller companies can do is compete on the things AI is genuinely bad at: local market knowledge, long-term relationships, the ability to pick up the phone at 7 p.m. when a deal is in trouble, the judgment to tell a borrower something they don’t want to hear because it’s the right advice. These are not consolation prizes. They are genuine competitive advantages in a market where borrowers are increasingly skeptical of automated everything.

Automate the back-office work your team finds tedious and error-prone. Keep the client-facing work human. And be very careful about signing long-term contracts with vendors in a space moving so fast that today’s cutting-edge solution may be obsolete before implementation is finished.

The part nobody wants to say

The mortgage industry has always been a relationship business that happens to use technology and not a technology business that happens to do mortgages. The loan officers who will still be thriving in five years won’t be the ones who adopted AI earliest. They’ll be the ones who understood what it was actually good for, kept doing the things AI couldn’t replicate and built businesses strong enough to absorb a technology shift without losing what made them worth coming to in the first place.

The infrastructure question and the technology question are the same question. Getting that foundation right is not the boring part of the conversation. It is the whole conversation.

Emmanuel St. Germain is the CEO and Founder of Choice Mortgage Group.
This column does not necessarily reflect the opinion of HousingWire’s editorial department and its owners. To contact the editor responsible for this piece: zeb@hwmedia.com.