Kastle CEO Rishi Choudhary recognized artificial intelligence voice agents could transform the mortgage industry when he and his colleagues saw what companies were failing to do earlier in his career.
"One of the big problems we noticed is mortgage lenders were just not picking up the phone and reaching out to leads that they were paying a lot of money for," he said.
"Some consumer direct lenders are really good at speed to lead, but retail loan officers that move into consumer direct are not used to operating the phones on a regular basis."
The opportunity to address that particular sales gap led him to launch Kastle in 2024, and the fintech's
"When we first started, voice AI technology was just not ready. It was very slow. The voices were not good, and you could tell that you were speaking to an AI agent."
At the pace artificial intelligence is improving now, though, Choudhary can see a future where mortgage customers have full-blown conversations with licensed AI agents, who possess the ability to lay out loan options in seconds based on detailed knowledge of a borrower's personal finances.
That future might be closer than many imagine, with the components to create it already available and becoming better.
While
Defined as the ability for a tool to make choices or achieve goals independently without human guidance, agentic AI is used today within the mortgage industry to create customized communications and remove the back-and-forth behind underwriting data verification, substantially reducing overall processing time.
The promise technology leaders see with it goes beyond efficiency and lower costs. It also will eventually offer a level of personalization that previously wasn't possible, some industry experts say.
"The feeling that you are the most important customer in the process, and there's always focus on you — that is the ultimate experience I think anybody wants when they're dealing with a company. I think that is possible with AI," said Jay Stockwell, a co-founder of mortgage lender Beeline and CEO of affiliated software firm Magicblocks.
Building customer trust
The finest technology won't scale if customers aren't willing to engage with them, though. The mortgage AI community is betting that consumers will overcome reluctance, leading them to forge ahead in building what they hope are game-changing products.
"If you're buying a lot of leads, you may not have the human beings — even if you wanted to call them with humans — to call fast enough. With AI voice, you can turn on 1,000 agents for the next hour and then turn them off," said Dave Grannan, co-founder and CEO of Marr Labs, another AI-based voice software provider. Marr's product is being used in various financial services industries, including mortgage and real estate.
"We think over the long term, people will probably have a certain level of comfort and prefer, at times, to talk to AI voice agents. An AI voice agent never gets upset, never gets impatient, never is tired," Grannan added.
Yet even though it might take on human-like qualities, AI is still a concept the public has not fully embraced. A late 2024 survey conducted by technology provider Cloudvirga found 60% of recent home buyers saying
A well-constructed compliant artificial intelligence framework of the future ought to be able to break down some of the mistrust borrowers might have, Stockwell said.
While immediate responses from AI remains part of its appeal, a better customer experience will emerge when it accumulates and retrieves knowledge to accurately answer a borrower's questions, not in generalities, but specifically, he added
In 2023, Beeline was among the first to introduce a customer-facing AI-backed chatbot. As it looked toward building out its capabilities, the company recognized optimal customer experience based on detailed knowledge of individual circumstances needed to sit at the center of framework design.
"We've built it so that it forms memories. It can maintain memories of previous conversations so that it can then fold that into future conversations," Stockwell said about the Beeline/Magicblocks software.
In Marr Labs' experience, consumer hesitation also has a tendency to melt away as a customer becomes accustomed to the technology.
"We always declare that we're AI when we're on calls, but when it's natural and human sounding, you get a much higher task completion," he said. "People don't feel like they're in that uncanny valley. Often, it's clear that people either don't know at all, or they forget they're talking to an AI, Grannan said.
Addressing compliance and the future impact of agentic AI
With lenders diving deeper into advanced AI offerings, the mortgage industry continues to encounter concerns that typically accompany the introduction of new automation and technology —
Any AI system is only as accurate as the information and instruction it is given, meaning both the provider and business user have both the responsibility and final oversight of the output that comes out of deployed tools to ensure regulatory compliance.
Maintaining a structure where every action or determination made by an AI agent is transparent and explainable is key to build trust needed among consumers, regulators and especially, the lenders using it in review and borrower approval.
"They don't want AI to make the decision. They want AI to list the reasoning — where they got the information from, so that they can review and approve," said
"We built a framework to actually enable that and then, not only explain the reasoning, but also tell you the document it finds information from," she said.
AI advancements in reading and understanding borrower documents will lead to much sought-after efficiency and savings lenders have longed for down the road, according to technology executives.
"If we can actually trust the data that we got at the beginning, then we can streamline the process of all of this downstream all the way from origination to securitization," said Jonathan Kearns, vice president of technology at the Mortgage Industry Standards Maintenance Organization at an Information Management Network home equity conference in March.
"Generative AI has the ability to really help us in verifying and trusting the information that we get, the data we spend as an industry an enormous amount of time, energy and money on," he said.
Since everything AI generates comes out of what is contained in an artificial brain, lenders can also put the appropriate controls and backstops in place to avoid revealing any type of information, such as pricing, which it doesn't — or shouldn't — want shared with a borrower.
"We created that type of guardrail only operating within a range, which makes our customers feel safer," Yu said. "Based on configuration, if a company does not want the borrower to have a conversation about the mortgage pricing with an AI agent, it would respond by saying, 'I cannot answer the pricing related questions, please contact your loan officer.'"
A full compliance engine of industry rules and disclosures governing consumer protections and fair lending also needs to be within systems for the AI to learn on, according to Choudhary.
Before it took a single call, Kastle's product, which was aimed exclusively at servicers when the company launched but has since branched into originations, was pretrained on required guidelines.
"That is a big differentiator that helps customers deploy the solution," Choudhary said.
Automation throughout history has frequently changed the look of the workforce, with disappearing jobs among the outcomes. While the latest developments in technology will likely
Humans are still needed to provide individualized service, understand lending requirements and ultimately, make final loan decisions, they said. Instead, agentic AI should be looked at as a tool that elevates worker potential.
"it should always be seen as a tool working for you. It's working for humans," Stockwell said,
Comparing mortgage employees to a popular superhero character, he added, "Tony Stark is not a superhero. He doesn't have any superpowers, but when he puts the Iron Man suit on, he does.That's how I think AI should be thought of. It's giving everybody superpowers."
The technology's near and distant future
Although concerns won't go away, artificial intelligence is already firmly established in home lending and looks set to become a bigger part of its future, driven in part by investments already
In the near term, lending fintechs that launched this decade see current interest fueling a wealth of opportunity ahead. Having come onto the scene just as agentic and generative AI began taking off, they can work on development, unencumbered by older technology or rules.
Mortgage lenders cut from a similar mould are also more likely to be open to new ways of using technology and also stand to benefit thanks to not having old habits they need to change or redevelop.
"We're at the point where it still is very much an early-adopter market," Grannan said. Voice AI uptake today is coming from tech-savvy companies that want to gain an edge on competitors by employing it at the beginning of a perceived future wave.
"They're willing to take probably a little more risk," he added.
Although AI can be cost prohibitive, many companies still seem willing to consider exploring options because their addition doesn't require a complete overhaul of an existing system, boosting appeal, Choudhary said.
"AI sits on top of it. I think helping executives understand you can get an across-the-value chain product without having to rip and replace is a big unlock in AI implementation," he remarked.
While not likely for years, the ability for an AI agent to pass the Nationwide Mortgage Licensing System exam at some point is not outside the realm of possibility given how much faster and smarter artificial intelligence is becoming, fintech developers also said.
At that point, the mortgage workforce would take on an entirely new look.
"I want to see a world where AI agents actually are able to get an NMLS license or even a servicing license," Choudhary said.
"They don't get all the rights of a human licensed originator or a servicer, but they do get some benefits that allow them to take the customer deeper into the funnel. We see this industry where each lender or servicer has a set of human employees, and there's a licensed AI workforce as well, collaborating hand in hand."
