Competitive Lenders are Using AI/ML: Here’s Why.

By Frank Poiesz

March 28, 2023 | AIVA, Artificial Intelligence

Artificial intelligence and machine learning (AI/ML) technologies are transforming our world, and competitive lenders are adopting AI/ML to streamline time-consuming tasks in residential real estate.

We are still likely years away from solutions that eliminate costly human work, but competitive lenders are already using AI/ML for mundane processes like finding data elements in documents and classifying documents based on the images or the language they contain.

Traditional mortgage software is deterministic, meaning it will repeatedly produce the same output from a given initial state over time. AI/ML is non-deterministic, which means outputs evolve over time, based on previous iterations. This characteristic of simulated “learning” from prior experience is similar to the way humans gain knowledge, which is why people use the terms “intelligence” and “learning” to describe what is really just statistics and probability.

This non-deterministic system behavior holds several competitive lenders’ advantages.

Competitive advantages of AI/ML

AI/ML algorithms help competitive lenders save time through the scripted analysis of often-large sets of data, freeing up teams to focus on borrowers while the AI handles tedious chores, such as document classification, in the background.

AI mortgage software can use optical character recognition (OCR) to scan text for key words or patterns and classify selections and whole documents based on how closely they match designated criteria. Once an AI system runs these data sets, results can be evaluated for accuracy.

As these models are run over and over again, they become honed and refined, offering competitive lenders yet another competitive advantage. Since AI systems mimic thought, the more a system is run, the more feedback lenders receive.  As that feedback is incorporated into the model, the system becomes better equipped to accurately answer the question. Just like humans, who learn from experience, machines “learn” as more quality inputs are added and algorithms are tuned to produce more accurate and reliable outcomes.

This means lenders can leverage highly specific AI/ML systems to tackle mortgage lending problems at scale. AI systems can already review insurance and validate borrower income, as well as classify and enter key data directly into loan origination systems. Some even take the first pass at important loan documents so competitive lenders and their teams can focus on managing processing exceptions and problem solving with borrowers. Imagine the advantage of freeing up hours currently tied up in “stare and compare” processes by handing off to a reliable, accurate AI system that can return results in minutes.

AI/ML also stands out over more deterministic automation capabilities such as RPA (robotic process automation). RPA “bots” are often deployed to work alongside human supervisors, performing repetitive, scripted tasks defined by a given set of rules. While bots can be helpful in accelerating some processes – particularly those involving highly structured forms – they lack the flexibility of AI when it comes to finding and extracting data from text documents and have no ability to self-calibrate.

AI systems, upon encountering uncommon or odd samples, can leverage contextual understanding to assess the unfamiliar content and determine what to do with it, reading and inferring a document type even if a type itself is not specified.

That type of learning ability will scale into the future as well. Competitive lending AI solutions are continuously updated with new skills to address even more complex origination tasks like analyzing pipeline behavior to improve fallout estimates, prioritizing activities to complete a loan application based on a given closing date, validating signatures, and so much more.

Best of all, some AI systems won’t necessarily require lenders to make a significant upfront investment of capital. These systems are independent and function across a full suite of mortgage platforms, and they adapt to your software while working alongside any other third-party vendor. Competitive lenders’ AI/ML systems can also operate on a per-loan or per-transaction basis and can be on-boarded rapidly.

Navigating regulatory concerns

For all that AI/ML can already do, the technology is still in its infancy, especially in its applications for the mortgage world. At its core, AI/ML’s non-deterministic approach to solving real-world mortgage finance problems is akin to highly sophisticated educated guessing. In such a tightly regulated industry, educated guessing does always not sit well with regulators.

Since AI/ML is increasingly being used to solve practical problems in the mortgage industry and represents a new technological venture for many financial institutions, any competitive lenders AI/ML strategy must consider different risk management and regulatory compliance challenges. Existing financial accounting rules, government regulations, business process standards and auditing methods all have one key commonality: they are based on testing known inputs versus expected outcomes. They rely on deterministic processes. Such testing does not hold up when looking at non-deterministic AI/ML solutions, and so far, no single standard yet exists for alternative testing approaches.

AI/ML systems can be perceived as mysterious “black boxes,” with inner workings that can be difficult to grasp without study by experts. That can be problematic in the highly regulated mortgage industry, where fair lending rules and other laws require transparency into loan origination decisions. Regulators, auditors and business managers alike are all trying to understand if and how AI/ML solutions can be relied upon to operate properly.

That’s why at the heart of any viable competitive AI/ML lending strategy, there must be “AI explainability,” the concept that lenders can use existing technology and documentation standards to check the AI’s answers to a given problem and update its decisioning models to better answer that question in the future.

Conclusion

AI/ML technology is the next frontier for competitive lenders. While the technology offers potential that is growing broader by the day, it also comes with the need to navigate regulatory challenges carefully and responsibly.

For a look under the hood at this incredible technology, and to learn more about how a competitive lender can adopt AI/ML while still supporting regulatory compliance, read our white paper here.