The cost to originate a loan remains high, but Dark Matter's CEO says automation is starting to change that
The cost to produce a mortgage has been climbing for years, and despite new technology helping make the process more efficient, costs still remain high.
The Mortgage Bankers Association, in its Q2 performance report released on Tuesday, said the average cost of producing a loan was $10,936, slightly down from the $11,898 in the first quarter but still high.
Changing that math is the central argument behind a growing wave of AI-driven mortgage technology, and one mortgage technology executive said the cost question and the technology question are finally converging.
Vikas Rao (pictured top), CEO of Dark Matter Technologies, said the origination cost problem is the right place to start.
"We're at a point in time where the cost of originating a loan is the highest it's ever been," Rao told Mortgage Professional America. "So how do you use technology to reduce friction? Where do you introduce automation and AI to bring down that cost of origination?"
Finding cohesive solutions
Rao said one of the most common things he sees across the industry is lenders accumulating software that individually promises ROI but collectively hurts the budget. The same way consumers accumulate streaming services they barely use, lenders accumulate tools that looked like solutions when purchased but are never fully adopted.
"How many software programs do we pay for that we never use?" he said. "You deploy it because some branch or top producer asked for it, but you're not looking at the drag it's having on your overall cost of originating a loan."
He said the adoption failure is structural, not just a matter of discipline. When AI tools are deployed as standalone additions rather than embedded within existing workflows, the learning curve becomes an obstacle.
"A bunch of point solutions that are independently great but don't work cohesively together only increase cost, not add value, because a human then has to stitch everything together," he said. "When something is dramatically new and adds a step outside of your normal process, there's a significant learning curve."
He said that’s why software solutions that carry the user throughout the loan process are better suited to not only improve efficiencies but drive better ROI as well. They allow companies to really understand where their money is being spent.
"Even with all Dark Matter offers, all in, we're often less than 2% of what a lender spends on originating a loan," he said. "It’s staggering where the money is actually being spent: on people and on all the other point solutions that people are buying."
From doing to reviewing
Dark Matter's Empower LOS covers retail, wholesale, correspondent, home equity, and assumptions channels, and Rao said the AI tools being built into it are designed to be part of the existing workflow from day one rather than a separate layer on top.
Rao said what AI actually changes in mortgage is what humans are being asked to do, with the work shifting from executing repetitive tasks to reviewing what AI has already completed.
One example he cited was post-close reviews. The company’s Aiva Intelligence solution checks signed packages from title companies for missing pages, data mismatches, and missing signatures, with reviewers focused only on exceptions that need attention.
"Think about the lift that people get from things like this," he said. "Historically, you would staff people to just comb through pages. Now the AI flags just the things that need your attention."
He said the same model applies on the borrower side. Dark Matter recently launched an AI assistant, currently in beta, that guides applicants through the loan process outside of office hours. A companion tool summarizes what the borrower did and what documentation is still needed before the LO even opens the file.
Rao said the goal is to make sure every user within Empower eventually has their own AI agent assistant, so that routine tasks happen through a prompt rather than manual navigation of screens.
"As a user, you should not have to do things that the AI agent can do. It should either autonomously do them, or you should be able to prompt it without navigating screens. We feel that is going to be a paradigm shift from a lending perspective."
Like so many in the industry, Rao is confident that eventually there will be a window of lower rates leading to a refinance surge. He said lenders who use the current slow market to build infrastructure for that moment will not be scrambling to staff up when volume returns.
"Do we really want to go back to a world where lenders hire like crazy every time refinance volume spikes?" he said. "Or, as a lender, do you want to have the right technology and processes deployed so that you are prepared to take advantage of those circumstances?"
Stay updated with the freshest mortgage news. Get exclusive interviews, breaking news, and industry events in your inbox, and always be the first to know by subscribing to our FREE daily newsletter.
