Editor’s note: This piece originally appeared in the June edition of MReport.
As homebuyers become more digital-savvy, lenders are looking for new ways to reach out to their customers in an environment that is becoming increasingly high-tech. Of all the fintech technologies flooding the market today, artificial intelligence (AI) holds enormous potential, not only to help streamline lender operations and processes, but also to make the marketing of mortgage products more efficient. However, a recent Fannie Mae study revealed that while 63% of lenders were familiar with this technology, only 27% had used AI tools for their mortgage business, while nearly 58% said they expected to adopt some AI solutions within two years.
During an MReport webinar sponsored by AI Foundry, experts delved into how AI is changing the mortgage industry and helping lenders achieve a high-tech, high-touch environment for their customers. They explored how the use of AI could enhance the customer experience, streamline costs, and improve turnaround times.
Using AI in Mortgage
Despite the strides fintech has made in recent years, mortgage lending remains a business where it can take weeks to process a loan and where applicants that were “approved” in minutes can still get turned down by underwriters.
“Here’s where AI is coming to the rescue,” said Steve Butler, Founder and President of AI Foundry. “While it takes about three weeks, on average, to close a mortgage today, advances in AI are set to drive a new generation of software robots that automatically process mortgages, replacing slow and costly manual processes.”
The good news, according to Butler, is that an increasing number of financial institutions have begun to understand the need and urgency of adopting technology enablers such as AI to prepare their operations for the future.
“If I have to look into the near- and long-term future of the mortgage industry, in 2019, we will begin to see approval times drop
for the current duration of three weeks, beginning a journey where, one-day approvals will become the norm within five years,” he said.
However, for lenders to undertake this journey, it is important for them to understand the finer nuances of AI and how it is changing the way consumers perceive their digital experience.
“AI is changing the way we do everything,” said Peter Piela, Head of Solution Development for AI Foundry. “What we’re seeing now is that AI is covering most of the major human sensory systems, from speech recognition to visual recognition and behavioral analysis.”
While the results of this technology have been truly amazing, the question is, how do we map those strides against what we’re doing in our industry?
According to Piela, AI is currently impacting mortgage in two broad areas: operational efficiency and enhanced customer or homebuyer experience. From an operational viewpoint, he said that one obvious application is the automation of mundane tasks. However, he adds that AI is capable of doing much more than those “stare-andcompare” activities.
“If we look at our human operators like loan officers and underwriters, they make a variety of decisions,” Piela said. “Some are fairly straightforward; some are more complex and require nuanced thinking. Why can’t we offload some of the more straightforward decision-making to AI and let it make them in a fully automated manner?”
To drive home the growing reach of AI, Rocky Stubbs, SVP and Head of Consumer Direct and Digital Mortgage Lending for Flagstar Bank, gave the example of a demonstration of Google’s AI system, Duplex. It uses natural language processing, and for this particular demonstration, it scheduled a haircut appointment without the person at the other end of the phone ever realizing they were engaging with a virtual assistant.
There’s already a high general awareness of voice technologies, with more than 70% of consumers having used this tech, according to Stubbs. Not surprisingly, this trend is being driven by younger consumers or digital natives, and most of the usage is coming from smartphones and mobile devices like Siri or the growing share of in-home devices.
“Though they’re being employed more frequently, most of these technologies are being used for quick, highly transactional functions, and that’s why AI has not yet gained large-scale adoption or broad thinking through a mortgage or banking transaction,” Stubbs said.
Stubbs continued, “As access to more complex processing becomes widely available, the use cases for mortgage lending will continue to grow, given the nature of the complexities we face in our business.”
A Case to Market
One such use case is product recommendation, which is different from product comparison, where the assistant just explains the features and benefits or does a compare and contrast between sales of two types of mortgage products. According to Stubbs, AI can be used for “actual recommendations based on a deeper understanding of the consumer’s financial situation and the inputs both within the data set that the bank might have and third-party data sets coming into one frame of reference for the assistant to actually make the product recommendations.”
From the user experience point of view, Piela said that AI could be used in the form of simple chatbots to answer casual questions, as well as in more sophisticated digital systems to help users navigate through complex decision-making tasks.
“If we have a more dynamic and parallel process driven by these AI real bots, we can provide better, more personalized feedback to the end user, and that, in turn, improves the customer experience.”
Another area where the use of AI can be leveraged by lenders is for intelligent applications.
“I don’t just mean a voice form where you’ve got a robotic voice going through the checklist and processing information,” Stubbs said. “This would be an intelligent process that is asking the minimum number of necessary questions and perhaps even second- or third-layer questions as the application process continues.”
The applications for AI are also numerous within the underwriting process, especially when it comes to searching for guidelines and regulations.
“As the nature of our credit policies and regulation become increasingly complex and fast-paced, leveraging intelligent assistants to find relevant guidelines, even to interpret data to find out what the applicable guidelines might be and then going back to the underwriter, is a way of turning the tables on how AI is being used today,” Stubbs said.
Rick Bechtel, EVP and Head of U.S. Residential Lending for TD Bank, also supported the idea of consumer outreach and bringing down the cost of marketing through AI.
“The traditional means of marketing have become incredibly inefficient, expensive, and at some point, they just don’t satisfy anyone,” Bechtel said. “What we’re looking at is if we can start to be predictive about what consumers need next before they need it.”
By using AI, Bechtel said lenders could string together a mass of relevant data about their customers, either through internal existing data or through a view of external data such as the media borrowers consume. This would enable lenders to make offers when and as appropriate, and targeted to borrowers’ specific needs.
The result, Bechtel said, would be severely reduced marketing expenses.
“The payoff for this is that marketing costs go down radically, especially when you’re hyper-focused on the customers’ needs,” he said.
While the possibilities to save on marketing and operations are numerous, it all comes down to correctly managing the complicated balance between modeling and implementing these systems. This, said Bechtel, is a journey that many lenders struggle to take.
Starting the Journey
One of the first lessons that lenders must understand while implementing AI is that the process of implementing this technology is a long-term objective. It’s a plan that “you’ve got to commit to because none of this is going to happen quickly or easily,” Bechtel said. He revealed that TD Bank had recently purchased an AI firm in Toronto that was helping educate the group about this technology.
“A lot of the work that we’re doing in the U.S. and around the world is happening locally,” Bechtel said.
Before lenders get too excited about the when and how of AI, it is important they ask why they should be implementing it in the first place.
“As AI increasingly becomes a mainstream part of their business, lenders can get distracted,” said Robert Bush, SVP and Senior
Strategy Manager for Citizens Bank.
Bush said that Citizens’ approach is to first understand the current and potential future business challenges, and from there, to determine if AI could be a solution to those headwinds.
“The reason we’re taking this approach is that, at the very least, it will ensure a lot of buy-in from the masses who have to help us execute on our strategy whether they sit within the business or help support it,” Bush said. “It helps us to frame up the expectations, the metrics we are trying to influence by implementing AI. We feel like this will help us get the momentum we need to solve future business challenges.”
Bush said that lenders should approach the implementation of AI the same way they would for any other business transformation.
“Many companies have gone through various iterations and transformations. In fact, all of us are in different stages of digital transformation across institutions,” he said. “Focus on a business problem and launch off from there—take a more bottom-up approach.”
Another area that Bush said lenders must consider is the actual execution of a project, ensuring that decisions are made early on about whether they will tackle it internally or find a partner.
“At Citizens, with AI being fairly new in financial services and without a lot of internal expertise in the area, we quickly realized we were going to need a partner to help us support this initiative and implement it so that we would also get some more knowledge as we took things forward,” Bush said.
Piela agreed, stressing that it is imperative for lenders to find the right resources that have the necessary background in AI.
“Those resources are not plentiful, as all that technology is still very new. So finding the right people to work with is important,” Piela said. “Lenders must take a call on whether they want to build that core experience inhouse or find partners.”
According to Bechtel, it was also important for lenders to form a tech roadmap to help them navigate the questions that come with implementing such technology. When thinking about technology partners, he said that lenders should look for partnerships that provide something that automates, speeds, and provides consistency to a process. Once that partnership is formed, they can ask what else they can do together.
The Human Angle
Every year that I’ve been in this business, there was something that made us think, ‘We’re five years away from being out of a job.’ Here we are in 2019, and we’re still not out of the job,” Stubbs said. “Today’s technology has changed the jobs we do, and it’ll continue to do so as it simplifies the mortgage process.”
“The people in the industry will start playing the role of the expert as technology takes on the more mechanical areas of the mortgage process,” Bechtel said. “We need to be thoughtful about things that are hard to automate, especially give the complexity of this business, which requires a lot of human awareness. It requires smart humans to deal with because, when we think about people, we think about expertise.” While some areas of the industry remain more ripe for disruption than others, Stubbs pointed out that, despite there being more technology in the industry over the past five years, customer satisfaction had decreased. “Part of that is because the technologies are still experimental, and they’re choppy, so they’re not always exactly right,” he said. Stubbs suggested that the borrower’s preference to engage with humans isn’t going anywhere. Therefore, it makes sense to “let computers do what they do best and let people do what they do best.” “The robots aren’t coming for us,” Bechtel said.