Previous Mortgage lenders are facing a growing reality: the complexity of modern lending is outpacing the capabilities of traditional software. In this HousingWire conversation, Zeb Lowe speaks with Michael Kelleher from Sapiens Decision about why AI-driven decisioning is quickly becoming essential infrastructure across mortgage operations. What began as a niche technology is now gaining traction across non-QM lenders, banks, credit unions, and correspondent channels.
Throughout the discussion, Michael breaks down the forces pushing lenders toward more sophisticated decisioning tools, including regulatory uncertainty, expanding product complexity, and the operational limits of legacy technology. The conversation also touches on Sapiens’ recognition as a leader in AI decisioning by Forrester and what that validation means for lenders evaluating the next generation of mortgage technology.
“There’s been a lot of changes since last time [we spoke],” Kelleher said. “I think the most important change is we’ve gone from what was a concept to a category.”
This shift was also confirmed when Forrester acknowledged Sapiens as a leader in decisioning AI. “Forrester has now named us a leader in decisioning AI,” Kelleher said. “And in addition to that, we now have real use cases with banks and credit unions, non-QM lenders, correspondent lenders, wholesale lenders, so we have a lot of momentum in the category. I would even say we have an acceleration going into 2026.”
According to Kelleher, the acceleration is driven by tangible results lenders are beginning to see. “One, we have real use cases now,” he said. “We have lenders that understand today’s technology has a difficult time managing decisions, and more importantly, keeping up with the changing of decisions.”
As those use cases spread across the industry, the impact of technology on decision-making becomes even more visible.
“When you have a use case in the market, and other lenders hear about decisions that used to take three days taking three minutes for certain lenders, they come to conferences like this. They start to talk,” Kelleher said. “When you’re able to solve pain and you’re able to bring that competitive edge to it, those two things can be a powerful combination.”
One segment where Kelleher sees particularly strong momentum is the non-QM market. “Non-QM is on fire,” he said. “I couldn’t be more excited about non-QM for the industry as a whole.”
But the growth of non-QM lending has also exposed operational challenges.
“There’s a lot of manual components to it,” Kelleher explained. “You have manual underwriters and different underwriters making different decisions, which produces different exceptions, and the turnaround time starts to get longer.”
Many lenders, he said, are asking whether they can achieve the same automation seen in agency lending. “These non-QM lenders would come to us and say, ‘How can we have an AUS (Automated Underwriting System) like Fannie or Freddie? Is there a way we as a non-QM lender, could possibly have an AUS?’” he said.
Sapiens’ approach is to translate lender guidelines and matrices into an automated decisioning framework.
“Our authoring tool is able to take their guidelines, their matrices, and bring it into a world where we’re able to look at eligibility, exception management, produce document checklists and conditions,” Kelleher said. “Overall, the outcome becomes eventually being able to automate it and govern those decisions.”
For banks and credit unions, the challenges may look different, but lead to a similar need.
“They typically have the same thing,” Kelleher said. “They have their guidelines on Excel sheets, or if they’re a more advanced bank, they have what’s called loan cards, but they’re usually very out of date.”
By converting those guidelines into structured decisions, lenders gain governance and transparency. “They have an auditability trail. They have versioning,” he said. “They’re able to say this was the decision made by this person on this version, and this was the outcome at a click of a button.”
Kelleher also emphasized that traditional mortgage technology was not designed for today’s level of complexity.
“The traditional mortgage lending software is just very linear,” he said. “We take an application . . . approve it, close it and sell it.”
But the number of decisions — and how often they change — has grown dramatically.
“I’ve actually found building some of those decision rules is the easy part,” he said. “How do you maintain them over time? . . . How do you handle the changes?”
Compliance is another area where decisioning automation is gaining traction.
“We got our feet wet in compliance,” Kelleher said. “Freddie Mac brought us in to work with them to be the engine behind LP, so we understand the importance of having an audit trail and versioning.”
By applying automated decision-making to loan data, lenders can streamline regulatory reviews. “If the auditor comes in, they’re actually able to press a button and show that entire audit trail,” he said. “It just lowers your regulatory risk or exposure.”
For Kelleher, the biggest takeaway for lenders considering AI decisioning is to start small.
“This industry believes that technology solves everything in mortgage,” he said. “[But] mortgage companies do not need to be technology companies. They win on the decisions they make.”
His advice: begin with one problem that matters. “Find that one decision in your company,” Kelleher said. “That one piece that you think is slowing down the process . . . that one piece that keeps you up at night.”
“Start with us. Let us just solve that,” he added. “And I always say it’s like The Matrix—you’ll start seeing decisions everywhere, and then you’ll wonder how you lived without it.”
