Editor-in-Chief Sarah Wheeler sits down with President and Chief Operating Officer Melissa Langdale mortgage cooperativetalk about recent events federal hydrofluoric acid TechSprint is focused on next-generation artificial intelligence use cases in the residential sector. The use cases developed by Langdale and her team are considered the most promising of the gen AI consumer experience use cases.
FHFA’s second annual technology sprint consisted of 12 teams demonstrating use cases in four areas: consumer experience, assessing creditworthiness, operations and risk management, and compliance. Langdale’s team came up with an app that could provide an Uber-like experience for multifamily tenants. Langdale has been in the mortgage industry for over 20 years.
Sarah Wheeler: Technical sprints always require a large number of tasks to be performed in a short period of time. How does your team decide what they want to develop?
Melissa Landale: The problem statement we chose was designed around solving the consumer experience, and our team happened to be multifamily. My entire background is in mortgages and single-family, with some exposure to commercial underwriting and multifamily, but very limited. Our team is very diverse in experience levels and backgrounds, and I think that helped us come up with a great solution around consumer experience, tenant screening, and some credit and underwriting.
SW: What are some of the things your team considers?
Machine learning: Well, mortgages have a very streamlined process that all consumers go through, whereas on the multifamily side, it’s very fragmented and there’s not really consistency from a consumer experience standpoint. On some search sites, they can see rental ranges and availability, but the credit criteria, for example, are not consistent. So we try to create an opportunity for consumers to control their experience and provide tools to standardize the tenant screening process, which landlords can then leverage to improve operational efficiency. What it’s really about is creating some standardization for the industry, letting consumers control it, and then delivering significant benefits to landlords on the back end.
SW: How do you think your experience with single-family homes influenced how you approached this project?
ML: When it comes to mortgages, we place a strong emphasis on the consumer experience and are accustomed to standardized processes. But our team has very diverse backgrounds—from engineering to open artificial intelligence—and everyone’s expertise fits the different pieces we’re trying to integrate. So it was really cool to see how it all came together.
SW: What made you want to participate in this tech sprint specifically focused on artificial intelligence?
Machine learning: Oh, I’m a huge nerd and I’m so passionate about this! I host the TMC SparkLab show where we talk a lot about innovation in the industry. I do this because I have been in the industry for over 20 years and my parents were in the industry before that. During this time, I had the opportunity to see how, as an industry, we move people from their desks to pieces of paper in front of us and then create technology solutions that facilitate paper-based processes. Over the past decade, technological developments in the mortgage sector have not kept pace with those in other sectors.
Artificial intelligence in general, and especially the new generation of artificial intelligence, is starting to enter the industry and help lenders rethink their workflow design and what is possible from a consumer experience and operational efficiency perspective. But all of these tools have rewards, as well as additional risks. Coming into TMC, I realized there was a real opportunity to help the industry understand what both of these aspects mean so that we can navigate this evolution in the right way while protecting our companies and consumers. This is my “why”. So this is an opportunity for us to be involved in something that will move the industry forward in some way, shape or form.
SW: What did you get out of the technical sprint?
Machine learning: There are a billion things! I learned a lot about multifamily. I learned a lot about Open AI and the capabilities of chatGPT. I had the opportunity to learn from each of my teammates as we put together the infrastructure for this use case.
SW: Does your use case require human involvement?
Machine learning: No, this is not built into our use case. In multifamily, this may not seem as necessary as it is on the mortgage side—there are a lot of complexities in mortgages and a lot of risk that lenders consider when making decisions. So if a client’s loan document doesn’t get through a very specific pipeline, they need someone who can step in and help that client. But it’s also complicated because someone is spending a lot of money over a long period of time and you need to make sure the consumer really understands what they’re doing.
On the mortgage side, we’ve been building very specific tracks that drive the process, and because our industry is so complex, it’s difficult for those tracks to account for every possibility. A self-employed borrower with 16 rental properties is very different from a self-employed borrower with 16 rental properties federal housing administration First-time homebuyers, so we built specific trajectories to capture every scenario a consumer appears in. But this new generation of artificial intelligence opens the door to seeing things differently, as systems can learn, grow and digest information from a larger pool, enabling them to facilitate transactions in different ways.
But again, there are risks and rewards. There is no magical next-generation artificial intelligence solution that will bail out all of humanity. On the multifamily side, they also face a variety of other risks that a mortgage doesn’t have to take into account, such as insurance.
SW: After your presentation, the judges took notes and raised the risk of fraud, specifically assisted by gen AI. How do you take this into account when developing use cases?
Machine learning: One lens we use is the fraud that exists today. So, are there solutions we can use to help reduce the risk of fraud, even just by a small amount? Today, there is some bias in the tenant screening process because tenants have to physically walk into the apartment building, but there is less element of fraud if they have to present documents in person. But our use case is really no different than mortgages, where many applications are already online. So we’re thinking about that as well – if we reduce the bias that consumers encounter in the tenant screening process, that might help us reduce the incremental risk a little bit.
SW: What was it like to be a part of this tech sprint? You all worked on this project outside of your regular roles!
Machine learning: Yes, everyone is doing their day job in addition to our team’s work. But we did have the opportunity to spend a lot of time together over the course of a few days, and everyone was very committed to making the solution something that we felt really good about. When you’re in that environment, it’s an opportunity to really get to know people, which is awesome. Right now our team has a newsletter group and we are still discussing things. So that part is fun and I love the opportunity to network and meet new people in the industry.
SW: FHFA just started hosting tech sprints last year. What do you think this type of activity means for industrial innovation?
Machine learning: This is our opportunity to explore possibilities. Anything we can do as an industry to lower our origination costs, lower the risk of mortgage origination and sell it on the secondary market is something I’m very passionate about exploring.
SW: I know your team is busy developing its own use cases, but what are your thoughts on the other use cases that were introduced?
Machine learning: They are all fun! I just happen to know that some of the things shown are already being built, which means they’re on the right track – they’re looking for solutions that actually work in the market. . I have to be biased – I like our solution best, but there are 11 other great solutions.
