Turn the ignition in your car and a series of remarkable machines spin to life. A starter motor engages the engine’s flywheel, initiating a combustion cycle via a crankshaft. To most of us, it’s as good as magic. Few of us even know what a fuel injector is, let alone understand how a car engine works.
A steering wheel, two pedals and a gearbox — that’s all we need to make our cars go. The operation of a car has been “abstracted away,” to use the jargon of Silicon Valley. This is a phenomenon seen across industries: As fields get too complicated, a need arises for a new level of “abstraction.” Think about how you plug an appliance into a power socket without ever thinking about voltages, Ohm’s law, power substations and the like.
A similar process is underway as artificial intelligence-powered tools gain adoption in the U.S. mortgage industry. Speaking with a grizzled mortgage veteran can feel like a conversation with a walking encyclopedia — the minutiae of lender guidelines, the subtleties of Regulation Z, the finely honed sense of each step of the mortgage “manufacturing” process.
The level of knowledge required to succeed in the mortgage industry is a barrier to entry. It’s the primary reason that, by one estimate, the average age of a mortgage loan officer has climbed to nearly 45 years. Artificial intelligence (AI) will bring a similar reduction in job complexity to the “drivers” of the mortgage industry: originators.
Daunting complexity
Drivers of a modern motor vehicle don’t think about what is happening under the hood. This frees their attention to focus on the road, where they’re going and maybe a conversation with the person in the passenger’s seat. Similarly, an originator should be focused on their conversation with the passenger (the borrower): discussing product options, solving problems and building confidence that the loan will close on time.
The mortgage industry has gotten overly complex, but AI-powered technology is what can bring order and simplicity back. The best uses of AI will eliminate noise and complexity. The goal is to put the originator back into the driver’s seat.
The 2010 Dodd-Frank Act is 849 pages long. The Fannie Mae selling guide is 1,176 pages long. There are lender overlays, state regulations and investor bulletins. A veritable forest of complexity has grown around the industry.
As a result of this complexity, operational productivity in the mortgage industry is dismal. According to a recent Mortgage Bankers Association report, the average underwriter in 2023 reviewed 16.3 closed loans per month, while the average loan processor reviewed 7.5 closed loans per month. These figures are down by about 50% versus 2020. Artificial intelligence has the unique ability to clear-cut the forest of complexity and focus our attention on the questions that actually require human judgment and common sense.
Highest use
This all leads to a very natural next question: “OK, but what exactly is AI-powered technology?” When many hear about “AI,” they think of those remarkable GPT creations people sent each other in early 2023 — the flowery birthday emails, Shakespearean sonnets in the voice of Cosmo Kramer and the rap songs about the Pythagorean theorem. (Anyone else?) And these “generative AI” creations” were indeed impressive.
Generative AI, however, is not what will truly revolutionize originators’ lives. A good comparison of “generative AI” is with the early use of gunpowder in fireworks. When gunpowder was invented in China around A.D. 800, one of its earliest uses was fireworks to impress the emperor’s subjects and to celebrate major events. Fireworks are indeed impressive — a beloved part of Independence Day celebrations today.
Chinese inventors quickly discovered, however, that gunpowder was even more valuable for their societies when used for warfare, mining and construction. In other words, the original use case was not the highest and best use for gunpowder.
Something similar will happen with AI. The business world doesn’t need “generation.” In fact, a quick scroll through social media will reveal that the industry is already drowning in human-generated content. The business world, and especially the mortgage industry, needs synthesis, automation and a reduction in overall noise — not “generation.”
Reducing tedium
AI’s true role in loan origination will be in tedium reduction and increased memory. Let’s explore a few examples of how AI can reduce the complexity from mortgage.
While every borrower provides the same documents with their loan application, there are inevitably additional requests from underwriting — the dreaded “conditions.” Thousands of pages of guidelines dictate what documents are needed for every borrower situation: letters of explanation for large deposits, tax returns, self-employed income, documentation for alimony payments, etc., etc.
Imagine if you had access to an AI that had the memory of an expert underwriter and could anticipate all of these document needs instantly, at the beginning of the loan process. This AI would also automatically fetch verifying information from third-party sources.
You could give your borrower a dynamic, specific needs list within minutes of them completing a loan application. Your borrower wouldn’t have to endure endless back-and-forth and weeks of waiting to know their loan was clear to close.
Many are already familiar with tools that can scrape the necessary information from borrower documents (optical character recognition or OCR in the jargon). These tools are getting significantly more accurate, cheaper to run and less reliant on human error correction. This means they are becoming more useful in practice.
Reading handwritten documents or images taken from iPhones is now becoming standard. When up to 50% of processor and underwriter time is spent in “stare and compare” tasks, AI offers the possibility of immediately bearing more of this burden.
Dynamic solutions
The best originators know that what sets them apart from the competition is being able to take even the most challenging borrower scenarios and nevertheless get the loan done. This requires extensive product knowledge and the creativity to restructure a deal.
“Mortgage practitioners too often fret about learning the minutiae of underlying AI technology. Familiarizing oneself with all this jargon … is like learning how a car engine works. Fascinating, but irrelevant.”
AI can augment originators by extending their memory from a handful of products to hundreds. It can dynamically suggest solutions to loan hurdles — for example, when a slight change in downpayment, documented reserves or qualifying income could open up new, more affordable loan products or assistance programs for borrowers.
Mortgage practitioners too often fret about learning the minutiae of underlying AI technology. Familiarizing oneself with all this jargon — transformers, neural networks, large language models — is like learning how a car engine works. Fascinating, but irrelevant to the task of becoming a better driver. Mortgage professionals ought to ask of new technology: “Will this tool make my job simpler?” Simplification is what this industry needs and what AI must deliver.
