The Amazing Race

Just like the U.S. and Soviet Union were scrambling to get their acts together, the mortgage industry is doing the same. With the release of ChatGPT in November 2022, companies like Big Purple Dot integrated the chatbot into its CRM system.

So it makes sense that companies like A&D Mortgage and Rocket Companies are releasing artificial intelligence-backed programs to get their skin in the race to automation. But, how soon is now? Is new AI coming into the marketplace too soon, or are companies making conscious efforts to market it effectively?

The Calm Before The Storm

Pavan Agarwal, CEO of Sun West Mortgage Company and Celligence, is an AI pioneer, revolving his business around Celligence’s artificial intelligence platform, Angel Ai. Angel Ai has been used by Sun West internally since 2018 and is considered an early piece of industry-specific artificial intelligence. Agarwal says it’s a good time for companies to compete in the AI space. “I do believe it’s smart because you don’t want to be the last person to adopt email. You don’t want [to be] the last person to have a website. It’s the same [now] you don’t want to be the last person doing AI,” Agarwal said. “And this is even more disruptive than the Internet was. So this is a land grab, a mind share grab. This is whoever grabs the most mind share first wins.”

Agarwal likens the marketing of AI to companies like Amazon. “When the first version of Amazon came out, it was just books, very simple … They were able to start off with something simple and did that one simple thing really, really well, and then they expanded systematically. As opposed to so many others. You don’t remember any of the other e-commerce brands that came out at that time,” Agarwal pointed out.

Eyes Wide Open

Agarwal also noted that now is the calm before the storm, with a lot of “mindshare” up for grabs, meaning the amount of consumer awareness or popularity surrounding a particular product, idea, or company. “It’s just the beginning for AI companies. There’s going to be a lot of them that will come in and a lot of them will get worked out. Last year when ChatGPT 4 came out, a bunch of AI companies got worked overnight,” he said. “Right now, the barrier to entry in terms of an AI product is low. I think it’s become easier to build something that you can call AI easier than ever before because the tools are mature. But it also means the expectations keep rising as well. So the real barrier of entry is ultimately dictated by how much market have you captured, and how much mindshare have you captured. That’s the real barrier of entry. So whoever captures the most mindshare wins at the end of the day.”

Kristopher van Beever, a principal at STRATMOR Group, has been in the technology space for 40 years, with 20 of those years simultaneously being in mortgage consulting. He’s concerned about the “rush” to get into the AI marketplace. “From an AI marketing perspective, many of the vendors that we talk to and lenders who are talking about AI really don’t have as deep of an understanding [of] what really the implications are of AI that they should,” he said.

Van Beever added that while looking at new AI programs in the mortgage marketplace, he’s been able to categorize each vendor into one of three categories. “The first category, I almost call them the used car salesman because they are claiming that their technology is AI-enabled. But when you peel back the layers, you realize they’re using rule-based engines, they’re using different ways to make decisions, they’re using expert systems, which is an early form of AI, but it’s not what a lender would expect to get from AI,” he said. “The second group, which I think is the largest group, are people who are very committed to doing things for their clients and their customers better. They market their AI by using AI in various ways. They’re using them as assistants or as chatbots. In mortgage, the group that’s using AI the most, I would say, are the ones that are dealing with documents. They’re taking documents and getting AI and machine learning to recognize, classify and extract data from documents and put it into their loan origination systems or their servicing platforms.”

The third group, van Beever says, are the “innovators,” meaning the vendors leading the charge with AI marketing. “That’s going to be the slowest and the longest pipeline. I would classify underwriting in that category and beyond the use of assistance,” he added.

The Room Where It Happens

Rocket Mortgage is ready to step into van Beever’s “innovators” category. Dan Vasquez, Rocket Mortgage’s AI Strategy Lead, says the idea for Rocket’s own artificial intelligence platform started percolating seven years ago. The program, called Rocket Logic, is an AI-driven technology platform aimed at simplifying and expediting the process of obtaining a mortgage for homebuyers. Vasquez rattled off some of the platform’s functions, namely secure document processing and its latest tool, Rocket Logic-Synopsis, which is generative AI to listen to, transcribe and search client calls, analyze sentiment and record client patterns and preferences.

“Early on, we looked at plugging machine learning models, document recognition capabilities, and image recognition capabilities,” Vasquez said. “These jobs to be done in creating a mortgage that are so critical and happen so many times throughout the process. Early on from the beginning, we were asking, ‘How can we use these new technologies to do these more accurately, to scale up our team members, to help them make better decisions?’ That’s been the driving force of Rocket Logic from the very beginning.”

Vasquez added that Rocket Logic is aimed at Rocket’s team members and partners in the TPO space. “You can’t tell the story of Rocket Companies, what we’re doing with AI, how we are trying to help everyone achieve homeownership, without talking about Rocket Logic,” Vasquez said. These days, there is no such thing as a technology company. Every company is a technology company, and Rocket Logic is the heart of our technology.”

Why now? Vasquez said that after being backed by Amazon, Rocket felt confident to contribute to the tech space due to Amazon’s notorious brand loyalty and market dominance. “What we’re trying to do right now is make sure that wherever these innovations may come from, we have a seat at that table,” Vasquez added, noting that Rocket’s partnership with Amazon is going on six years. Per a company press release about Rocket Logic-Synopsis, the platform is “built on Amazon Web Services (AWS) and using Amazon Bedrock, the Rocket Logic-Synopsis technology is active for calls handled by Rocket’s client experience teams, including loans from mortgage brokers who work with Rocket Pro TPO. The tool is also used by the company’s servicing team, who oversee Rocket Mortgage’s 2.5 million serviced loans.”

But, this wasn’t a freefall launch. Despite seven years of development, Rocket Logic is still in its infancy. “When it comes to AI, I think it’s always important to crawl, walk, run,” Vasquez commented. “With AI, it’s important to not rush into things in a way in which you miss risk vectors or potential gaps in your strategy. And do we have a good, safe, ethical way to mitigate that risk? Or are we unsure? And in an industry like ours, we can’t afford not to be sure.”

AI Adjacent

Just a week after Rocket Logic’s launch, A&D Mortgage unveiled AIM, described as “the next evolution of [our] partner portal with enhanced AI integration.” A&D’s Chief Information Officer Nikolai “Nick” Avgustinov said that while continuing to develop A&D’s existing portal, including AI functions was a no-brainer. He noted that the previous portal’s interface wasn’t intuitive, nor did it create an easy user experience. “One of the challenges was we figured out that we had a lengthy approval process, complex documentation, and we needed some better risk assessment and better collaboration with our broker. So the answer was to significantly improve the interface of the portal,” Avgustinov said. “From the technological point of view, it’s a modern technology stack. Also, of course, we implemented pieces of AI to manage condition automation better and to speed up the approval process.”

From Avgustinov’s perspective, there was no better time to update the portal than as soon as possible. “We’re making automation not just for the broker, but also for our internal team, so they will do their work faster with the good tools and with the good technologies,” he said. And here we’re thinking about the text recognition, the workflow automation, their workflow automation … it’s not just a nice piece of software, it’s also a powerful tool to provide cost-effectiveness and streamline process.”

Van Beever would most likely categorize A&D in the first category of AI vendors. “Artificial intelligence is a beautiful word, but in reality, it’s just the machine copying human thoughts and human interaction,” Avgustinov said. “Right now, we are utilizing more machine learning and natural language processing. But if we’re talking about future development, of course, we have a lot of ideas.”

Fear of Missing Out

There’s an element of missing out on the next opportunity to crack the AI code, especially for an industry that is still cautiously approaching the use of artificial intelligence. According to the Mortgage Lender Sentiment Survey published by Fannie Mae in October 2023, even though 65% of lenders are familiar with AI, only 7% have deployed it within their current mortgage process. Another 22% say they’re trialing the use of AI solutions on a limited basis. “I think there is this fear of missing out. You’ve got that, ‘Hey, my competitor has AI, therefore I should have AI.’ But I do think some lenders are looking for better ways to do a better job for their customers,” van Beever said. “[They’re gauging whether] AI can give them the ability to be more effective, more efficient, do a better job of providing a consistent solution to their clients, keep up with regulations in a faster way.”

For van Beever to consider a product worth checking out, he wants to see a proven track record of evolution and accuracy. “If you have limited capabilities and you can provide an enhanced ability for a lender to use your product, then you introduce it at a lower scale … When I look at vendors who can actually show me that their product is good, that’s something that I can now start to adopt. Because if it can be demonstrated and shown that it’s good and better,  you still have to have a few humans in the loop, or you still have to have people that oversee that to make sure that the claims that they have of 85 or 95 or 99% accuracy really are accurate,” he said. “Because whether they use that technology or not, they’re still accountable for the results. If you rely on an AI, you still have to be able to defend that decision or that use of that technology in an audit or if you have to go to court over something.”