AI may turn 30-year mortgages into ‘floating-rate’ home loans as refinancing gets faster

Lenders say artificial intelligence will help homeowners whose mortgages are ripe for refinancing secure a new, cheaper loan far faster. The result could also squeeze investors in the $9 trillion market for mortgage bonds.

When interest rates fall, only about a third of homeowners who could save substantial sums by refinancing actually do it, according to research from Morgan Stanley. That’s because candidates don’t know they’re eligible or don’t want to go through the notoriously drawn-out and tedious process.

The number of takers is likely to rise, though, as mortgage lenders embrace AI to churn out approvals in a fraction of the usual time. Rocket Mortgage says a borrower can get from application to rate lock in just 30 minutes, and it’s aiming to cut that to 10 minutes. Rival United Wholesale Mortgage says initial approval can take as little as 15 minutes. Better.com, another digital lender with a small slice of the market, claims the firm can do it in only two.

Faster turnaround could double the percentage of eligible homeowners who refinance to perhaps 60%, according to a report from Morgan Stanley strategists including Jay Bacow, co-head of securitized products research. If that happens, they wrote, AI could make the 30-year mortgage seem like something “closer to a floating-rate instrument that only floats down.”

While this would be a boon for lenders and borrowers, it might erode returns for mortgage bond investors as their stream of income from higher-rate home loans gets cut off faster than they bargained on, and they’d have to reinvest at the prevailing lower interest rates.

“Faster speeds can impact returns,” said Charles Sorrentino, head of investments at Rithm Capital Corp. “Whether it’s through a rate rally or an operating efficiency or just a change in consumer sentiment, investors will need to deal with it.”

One possible outcome is that home loans become more costly as investors demand extra interest to compensate for the added risk — perhaps one or two tenths of a percentage point, the analysts wrote. In financial markets, mortgage-backed securities could wind up trading like pure callable bonds.

It’s also possible that the savings in time and money from AI-assisted refinancings will be so significant that homeowners come out ahead anyway, and that the easier process will motivate customers who were reluctant to act under the old system.

“AI should crush behavioral friction,” said Kirill Krylov, a fixed-income strategist at Baird.

True, the technology will do nothing for the borrower who simply doesn’t meet key requirements, like having enough income, he said. But “the borrower who was eligible but inattentive, lazy about shopping, intimidated by paperwork or simply never contacted becomes much more refinanceable.”

The impact of AI is hard to gauge with mortgage rates holding stubbornly above 6.5% and thus muting demand for refinancing. Ziggy Jonsson, Better.com’s chief technology officer, said future borrowers could consider it when rates fall by just 10 basis points; if the costs and hassles drop, even small loans close to their end date could benefit.

Less Friction

The formalities of a closing — including legally required time for customers to review documents — still could add plenty of length to the process, and the Morgan Stanley analysts noted that regulations require the participation of licensed humans, rather than relying solely on an AI “black box.”

But labor is about two-thirds of the cost of creating a mortgage, they said. “If originators shift from human labor to AI, perhaps closing costs fall close to zero and the refinancing wheel becomes near-frictionless,” the analysts wrote.

Rocket has plowed more than $500 million into AI, automation and related technology over the last six years to make it easier to reach refinance candidates and get the deal done faster, Chief Executive Officer Varun Krishna told investors in May. Agentic AI can help Rocket sort through current clients from whom it’s collecting payments, figure out how to best reach them including the time of day and whether it’s better to text or to email, and assess whether they’d be interested in refinancing.

Personalized Deals

Some lenders are setting up customized software that can assimilate borrowers’ financial documents, eliminating a process that could otherwise take weeks by pulling data from tax forms and bank statements into a personal profile that supports the loan. Borrowers can also adopt AI programs that prompt them to refinance, Morgan Stanley said.

Not everyone is convinced that AI is poised to transform the business. Jeffrey Kvalevog, chief strategy officer at lender New American Funding, said the industry has seen similar pronouncements about optical character recognition for scanning documents, automated phone calling and email.

“People tend to sensationalize how much AI is going to save you,” Kvalevog said.

Even if AI saves time, refinancing still can come with sizable upfront fees for appraisals, lawyers, mortgage taxes and other expenses that could discourage applications. So far, there’s not a lot of evidence that borrowers are changing their behavior.

“The existing prepayment data is more evolutionary as opposed to revolutionary,” said Mark Tecotzky, co-head of credit strategies at Ellington Management Group.

The real test for mortgage bondholders will come when there’s a significant drop in home loan interest rates and more borrowers try to refinance.

“It’s in every conversation,” said Walt Schmidt, a mortgage strategist at FHN Financial. “We haven’t had a good refinancing wave since AI became a thing.”