BLOG VIEW: The Federal Open Market Committee (FOMC) on Sept. 18 announced its decision to cut the Fed Funds rate by half a percentage point. It was the first rate cut since the early days of the pandemic. In fact, the last time the FOMC cut rates by half a point was during the 2008 global financial crisis.
Federal Reserve Chair Jerome Powell finally provided the housing industry much needed relief by slashing rates for the first time in more than four years. The Fed’s initial cut is likely to bring more buyers and sellers to market, potentially opening the inventory floodgates and momentum for price competition.
The biggest beneficiary will be borrowers looking to lower their rate – borrowers who secured 7% to 8% rates over the past 24 months. They will likely take advantage of the recent rate dip for some relief.
CoreLogic recently posted that an estimated 4 million loans are candidates for refinancing if mortgage rates float below 6.75%.
“An estimated $1.45 trillion outstanding mortgage balances originated during the high-interest rate period in 2023-2024,” the firm says.
Experienced mortgage bankers recognize that this could result in a tsunami (or at least a tidal wave) of activity. What will lenders do? Traditionally, they would would ramp up hiring – more loan officers, processors, and contract underwriting. Maybe some will realize technology weaknesses need to be buttressed.
But here is another perspective:
Advances in artificial intelligence, including the application of generative AI and expanded tools for predictive machine learning models, will be instrumental in how the industry approaches this surge. There has been considerable progress, even in the last 30 months, in terms of how technology is impacting fulfillment of loans and streamlining operations.
In our firm’s Touchless Lending Platform, AI is being leveraged to digest collateral more effectively, validating the estimated appraised value. But more importantly, using Large Language Models to analyze appraiser sentiment in comments to automate the process of creating conditions. Analyzing property images in the appraisal to underwrite against GSE guidelines and create conditions such as water damage, signs of recent repairs, lack of images from all rooms. Thus automating the creation of conditions on the collateral.
Similarly, the platform leverages image analysis in indexing and extracting data from documents. By analyzing the extracted data, platforms can automate the clearing of conditions. For example, leveraging AI, the platform can review bank statements for non-QM DSCR loans to identify eligible and ineligible income transactions, analyze rent-rolls and spreading of financial documents for commercial properties.
There are also tremendous opportunities to leverage AI in the origination/production side. First in portfolio retention, loans were traditionally identified as “in the money;” relying on predictive models to determine what loans could be candidates for refinance. But now two primary things have changed. One is the availability of supplemental deterministic data that can be included in predictive models. Data around borrower financial health, assets, savings, life events such as new job, marriage/divorce, birth of a child. As the availability of data expands, models get better and more precise. Targeting the layer of eligible loans also becomes more precise.
Second, the ability to create predictive machine learning models has improved dramatically. In the past creating these models required a specialized data science, mathematics, and statistics background. These were PhD level resources who had special experience and training to conduct supervised training of models. However, there are tools such as Azure Machine Learning, Amazon SageMaker and Google AutoML that support the un-supervised training of complex predictive models. The industry now has access to a greater number of skilled resources using robust tools to be able to optimize across all the new data that could be analyzed to build better more predictive models.
When examining the origination/production side of the house, it quickly becomes evident there are other areas where AI can have dramatic impact. For example, using Generative AI as a scenario generator to help new LO, processor, underwriters get familiar with product offering and pricing. Not just using a ChatGPT like Q&A tool but an actual GenAI solution to create dynamic scenarios for new employees to practice actual responses to borrower or broker loan questions.
Additionally, this enables these same employees to position products and address questions, issues, or concerns in a training environment that mirrors real situations and scenarios. Gen AI is a powerful learning tool.
In addition, the generative nature could be a tremendous marketing boost. Rather than a limited number of marketing template emails, AI can be leveraged to conduct more comprehensive campaigns morphing the messaging, channels and methods of approaching a customer – email, text, phone. Dynamically analyzing results and generating alternative strategies and approaches to solicit customers.
Mail-merging static templates is no longer a worthwhile strategy, in today’s market there are much better, more intelligent options.
While one might be able to see the black/white of this refinance wave (loans in/out of the money), do not forget the grey zone, the borrower whose rates are still not low enough to refinance because they are locked in at 2% or 3%. Using AI to help them unlock their home equity is equally as important. Our firm’s “Home Equity in a box” provides for a turn-key solution to end-to-end pre-approval, origination, fulfillment to close of HELOC and closed end HE Loans.
In conclusion, it is pivotal that the mortgage lending industry does not approach this impending cycle with the same lens. It’s time to take an innovative approach to a cyclical problem by leveraging advanced AI.
Sundeep Mathur is the vice president of the fintech business for Tavant.