Start of Main Content
A red-lit radio studio with an “ON AIR” sign on a brick wall and a glowing microphone logo reading “Stewart in the Studio”.

The State of AI in Mortgage Lending Heading Into 2027

Stewart in the Studio

A Podcast for Mortgage, Home Equity and Servicing Professionals

Episode 31

AI is moving quickly, and mortgage lenders are weighing where to use it, what to invest in and how to manage the risks.

On Episode 31 of Stewart in the Studio, Brian Francis, head of AI for Informative Research, joins Marvin Stone, Rich Kuegler and T.J. Harrington to discuss practical uses of AI in mortgage lending. They explore applications from analyzing financial data to helping development teams work through backlogs, and why lenders need to understand how their vendors use AI.

The conversation also covers governance guidance from Fannie Mae and Freddie Mac, the need for consistent, explainable results and ways to test targeted use cases before making larger investments.

Watch to hear what lenders should consider as they decide where AI fits in their operations.

Key Takeaways

  1. Lenders with limited integrations can end up tracking individual products across separate systems instead of seeing the work tied to each loan. A shared ordering portal such as OneHub offers a way to bring those services together without requiring extensive integrations.
  2. The cost of a mortgage service extends beyond its quoted price. Audits, financial monitoring and ongoing vendor oversight add to the expense, so working with a partner across multiple services can offer savings beyond individual product discounts.
  3. A provider’s financial stability matters long before it closes its doors, particularly when onboarding a replacement can take six to eight months. Evaluating a partner’s staying power and ability to support multiple services can help lenders prepare for disruptions before they become urgent.
  4. Supporting the broker channel requires dedicated teams, operations and account management that take time to build. That established expertise drove Stewart’s acquisition of Nationwide Appraisal Network and illustrates why lenders should evaluate a provider’s experience in their specific channel.

Transcript: The State of AI in Mortgage Lending Heading Into 2027

EP31: The State of AI in Mortgage Lending Heading Into 2027

Marvin Stone (00:00)
This is Stewart in the Studio, from Stewart’s thought leaders. Well, hey everyone. Welcome to another episode of Stewart in the Studio. Today we're joined by our special guest, Brian Francis, who is one of our AI leaders here at Stewart. And we'll get to his intro in a moment, but just know that he's a key figure around here. But first my co-host T.J. Harrington. T.J., give yourself a quick introduction and tell everybody who you are and what you do.

T.J. Harrington (00:24.206)
Hey everybody, T.J. Harrington, SVP of Strategy for Stewart Lender Services, jack of all trades, master of some at this point. I drive strategy across the full spectrum of Stewart Lender Services offerings.

Marvin Stone (00:37.145)
Great. And Rich Kuegler.

Rich Kuegler (00:39.317)
Hi guys, I'm Rich Kuegler. I lead the client success group here at Stewart Lender Services, and that really is sales, business, development, and marketing for any of our organizations that support mortgage originators, home equity originators, capital markets investors and servicers.

Marvin Stone (00:56.655)
Great. Thanks, Rich. And Brian, you know, you are at the center of things here at Stewart with so many different AI initiatives underway across different groups across the company. But one of the things that's really, we're gonna talk about today is how things have changed over the past couple of years and really what lenders can expect going into 2027. You know, we have seen sort of the beginning of it all, which is chat and chat prompts and prompt engineering and those types of things, and then we got into a little bit more bots and agentic AI and things like that. So you've been here, you've been involved in AI and machine learning since long before kind of the recent craze started, I guess in 2022. But with that, before we get started, just give us a brief intro about who you are, what you do here at Stewart and kind of your background, how you got here.

Brian Francis (01:43.64)
So hey, thanks, Marvin. So, I'm currently serving as the head of AI for Informative Research, which is the division of Stewart that is our credit reseller. So we focus on delivering credit reports, tri mergers, as well as verification services. And I joined IR and Stewart through the acquisition of one of those verification services, a product that I developed called Account Check, which is consumer permission to access to financial banking data and payroll data. So that service is employed by over 500 lenders across the United States. But yeah, as you mentioned, AI has really been at the core, even before this whole aspect of large language models and being able to chat with an AI. So AI is core really to the analysis work that we were able to do with Account Check to look at the data in a consumer's bank account.

And make certain identifications from it. If you go back in time 10 plus years, the introduction of this data into Fannie Mae and Freddie Mac, Account Check led that. And they had never seen that kind of data before. It started really as a replacement for physical bank statements. So instead of the consumer providing the data, we go directly to the source.

But we quickly understood by looking at the data, there was so much more that could be derived from it. In order to help us with that derivation, we had to be able to accurately classify and categorize that information so that we could begin deriving things like income and employment and traditional, and I'll use the term traditional AI. Traditional AI, things like machine learning, neural networks, rule-based systems were really the core what allowed us to do that. So taking the gobbledygook of a bank transaction description and being able to extract out of it, hey, this is a payroll deposit for Brian working for the company called Form Free. And that really was what allowed us to open up Fannie and Freddie's eyes to say, hey, there's a lot more here into the data that we're getting.
Now, back in 2016, so we're celebrating the 10-year anniversary of day one certainty for Account Check verification. And unfortunately, it's taken a long number of years, but both Fannie and Freddie are finally at the point where they now realize through their own internal evaluation that yeah, I can look at a person's 12-months worth of bank data and utilizing these types of AI tools, know who they work for, and know how much they make.

Marvin Stone (04:36.387)
Yeah. Well that's a that's a great background, Brian. I mean, as far as, you know, just in how you've come up through the trenches actually developing a product and things. One of the things that's been interesting is kind of the lender perceptions out there. And so T.J. and Rich, maybe you guys can kind of talk about you know, where this has gone from the lender's standpoint. I mean, obviously, you know, Brian's been very deep in the verification space and all through the development of that really pioneering effort. But you know, Rich, from your standpoint, you're always on you know, traveling and talking to lenders all over the country all the time. How did how do they feel about where we came from and how where we are today?

Rich Kuegler (05:17.505)
Well, I think it's probably no surprise that a lot of lenders are approaching this in a very cautious manner and not really sure about the roadmap and what capabilities there are. And you know, Brian was like I think half joking when he said traditional AI, but that was like a whole what 14 months ago. You know, and it's like the trajectory has been so fast that I don't think that a lot of folks really can have captured what it's all about.

And now the question and the and the uncertainty that I'm hearing out there is, well, what do I do with this? Right? Like it makes sense to now have the discussion around can I look at my outsource process that I do overseas because I was going to try to do some kind of arbitrage on labor cost? Can I automate some of those capabilities now?
Can I just have a have a computer talk to a human or have an AI agent as you were calling it now, you know, resolve some simple tasks and then help in triage and those kinds of things. So it's kind of all over the board. But what I'm hearing is people were afraid, they're getting less afraid, but now there's some trepidation on what do I do next and how do I make sure my investment, like you know, you've had a number of things over the past few months of how much some of these major AI companies are investing in data centers and technology, how much do I need to invest to stay up and what do I invest it in? And that's where I would turn to T.J. and say, What do you think, T.J.?

T.J. Harrington (06:49.665)
Yeah. So I've been involved with Account Checks since before the Stewart acquisition with Form Free. So I've known Brian for a while and and was involved with that. And I can tell you that the key to that product was the day one certainty with Fannie and Freddie and that gave lenders the confidence to know that was something the GSCs had approved. And similarly now, you know, the conversation that was the equivalent of ten years ago, the equivalent of AI governance, right? What can what lenders are looking for today is certainty and saying what's acceptable? What are the rules of the road?

And Fannie and Freddie looking at the Account Check product and saying, hey, we feel this works the right way and gives you day one certainty for deliverables on these items and rep and warrant relief. That launched Account Check as the product and certainty. And I think we're waiting for the I think we've seen and Brian, you can share some of the governance items that you've seen and I think there's some great resources we have there. But the GSEs have begun to put their kind of pole in the ground saying, hey, here's the governing stances or for us around the use of AI. And from there I think you're gonna get that rules of the road and then from there you get the investment. And I think that's been the key item. Getting something that is a starting point around the use of it so that lenders have the confidence and then figuring out from there what the products, what the processes, what the best use cases are. I mean in default servicing today, just like Account Check, we see the automation of the claims, the supplemental claims process for FHA, which traditionally has been a big mess of data. And like what Account Check has done, AI has ingested the data smartly, identifies what it is, classifies it, pulls out what's relevant and develops the partial claims package.

That's amazing. And that's the exact use case for AI where it's not making decisions, but it's taking a stare and compare process, a human process, and enabling AI in a way that you know is so much more efficient, faster, timelier, more accurate than a person ever could be, and saves a tremendous amount of costs. And I think we're looking to find those use cases across the originations and servicing and default servicing cycles.

Rich Kuegler (09:00.897)
Yeah, T.J., that's a great way to say it. Like how do you enable better decisions rather than replace the decision-making process?

Brian Francis (09:08.011)
And I think one of the biggest challenges lenders are facing right now, especially in this ecosystem, is you're seeing every technology vendor out there figuring out how they bolt AI into what they deliver. I mean, I won't name names, but there was a service provider in the industry that came up with a new tagline that says they are the AI agentic experts for mortgage. I'm like, are you kidding me? Have you ever seen like their products that they've actually delivered? So lenders really have this challenge of…

Rich Kuegler (09:42.187)
Bold strategy, Cotton.

Brian Francis (09:45)
…where are you getting all of this? How are they, how are the lenders able to understand and dissect this? How do I know that vendor A's AI really is real? And how do I know that vendor B's is just a load of malarkey? And that's one of the big things that, with the new guidelines from Fannie and Freddie coming out, is the lenders are going to be responsible not just for the AI that they may use internally, but also for their vendors' AI. So it's going to become very important that lenders can accurately assess what their vendors are delivering. And so as a vendor, it's incumbent on us to deliver to the lenders the necessary verifications and validations. Think of it as a SOC2, but as a SOC2 that's been enhanced specifically for AI. And in my new role as head of AI at Informative Research, that's been the first thing. That's been the biggest focus is how do we get our governance in place so that when our lender partners come to us and say, hey, where are you using AI and how are you controlling it? We have that ready to go. Just like we can tell them what our business continuity and our disaster recovery is, it is already written. There's no marketing fluffed associated with that. And that's really one of the keys that lenders need to look at is are your vendors certified in the information they're delivering to you about how they use AI?

Rich Kuegler (11:12.277)
And Brian, that's that's an important note too, because across the Stewart organization with our AI council, I mean, that's that's been a big focus as well, is knowing that we touch a good number of transactions across the enterprise, right? Whether it's the credit verifications that come out of IR, whether it's the valuations and appraisals that come out of our Stewart Valuation Intelligence business, or it's the plethora of title products that come that we that we deal with the purchases, refi’s, home equities, et cetera.

All those lenders are looking for that kind of governance control and that I know has been a big focus of the AI council that Stewart has put together with a lot of senior resources to try to figure out exactly how we do that.

T.J. Harrington (11:53.368)
We met with a large national lender in Dallas at Five Star and the first question out of the vendor manager's mouth was, what are you doing for AI governance? And how are you using AI in your business? And to Brian to your point, I think for at least the title and appraisal portions of the world. We're not we're not making AI a product. We're really saying AI is part of our business process and here's how we govern it and here's where it we're we're using AI to be more efficient. But we're not saying hey we're AI, AI as a product set. It's really what we're providing today as traditional solutions, but AI enabled in a way that enhances the product offerings. But subject to the kind of governance guardrails that you'd expect out of a publicly traded insurance company.

Brian Francis (12:35.917)
Yeah, and I think that's really, T.J., where you get into kind of separating the hype of AI with the business value that can be delivered with AI. And if you if you step back and look at it and you look at like the way that chatbots have evolved with a large language model, the key word to look at there is that the response that you get from a large language model is what's called non-deterministic.
Which means you can ask it the exact same question with the exact same data, and you might get a different answer. So, it's very important that, yeah, there's a lot of great things that you can use agentic AI and AI harnesses to be able to speed up development, to rapidly process and gather and summarize large amounts of information. But at the end of the day, lenders have to know they are delivering a deterministic product.

Which means if I'm taking the inputs to a mortgage loan, the consumer, their capacity, and their collateral, that if those inputs are the same, that my outputs are going to be the same. And if you have AI in the middle somewhere, then you have to be able to prove that given the same inputs, you're always going to get the same outputs. And that's really a key differentiator. And what it does is it really focuses you on where can I make AI effective for me in both the short term and in the long term?

Marvin Stone (14:10.475)
Yeah, for sure. Thanks, Brian. So, and great conversation here. And so one of the things that I mean lenders obviously are looking to deploy AI, everybody wants the advantage, right? And so and to cut cost especially. And so we hear this, you know, Brian, you're kind of saying that the sweet spot is to develop, use AI to develop deterministic processes where you can so that you know what the answer is going to be every time you put A in, you're gonna get B out. But you know, otherwise I guess it's like asking your kids something, you know, it's like you never know if the answer is gonna be the same twice, right?

So but I guess the thing is, you know, there's is there sort of an inventory across the mortgage processing or the mortgage manufacturing process, like an inventory of places where AI can fit. Especially maybe not on output where it really needs to be deterministic. But at least in filtering, sorting, bucketing loans, where kind of like from the beginning of the to the end of the pipe. Where are the sort of sweet spots to be looking at?

Brian Francis (15:08)
Well, I'll start and then Rich and T.J. can weigh in. But kind of my background has been in software engineering. So serving as the chief technology officer to develop the Account Check product. I'm really looking at how can core systems be developed. So every lender that's out there, if they go and talk to their CIO or their head of development, they're going to hear two things. They're going to hear backlog and they're going to hear tech debt.

Which means I've got a long list of features and capabilities that the business has been asking for years, and I've never been able to get to it. And maybe when I got to it, I patched it in there or I kind of rushed it out and didn't do it in the optimal way. But what you've seen with the advent of tools like Claude Code and Codex to be able to one, rapidly ingest your entire code base, your entire set of business requirements directly from tools like Jira. Analyze that, take a look at what the backlog is, and rapidly clear that backlog all the way down to delivering code that's ready for human review. And that's where you get into that gate. Go ahead, Marvin.

Marvin Stone (16:35.481)
Well, so what does that look like? So you're saying point AI at your Jira ticketing system for backlog projects and use AI to determine which projects are low hanging fruit versus which ones are maybe a little more task rich.

Brian Francis (16:50.195)
You can address this in a crawl, walk, run scenario. Crawl being exactly that. Identify, look at the Jira tickets, look at the code base, and say, all right, of all these tickets, analyzing the code base, these are the 20% that can be knocked off immediately. Then the next step, generate the actual development tickets to do that and hand that off to your development staff.

The next step is instead of handing that to the development staff, hand that to a coding agent and have the coding agent run that. And then once it exits the code, then it goes through the rest of your software development process, going through QA and then going through release. I'm not advocating, nor do we advocate this of all that running automatically directly through release, but you can speed up a large percentage of the process, clear that list, and then what it allows your senior architects to do is analyze, validate and then quickly move to release.

Marvin Stone (17:56.464)
So, you kind of mentioned and guys you can confirm because you're in lender shops all the time, but you kind of hit on a couple of the areas that I think that we hear of a lot that in lenders need to be thinking about these three areas. One is what who's asking me across the organization to develop a custom application? And that's what you just described here is I need a custom app to do this thing. Whereas typically we buy off-the-shelf apps and make them do what we want them to do with a lot of configuration and customization. The second one is really the third-party AI products, which are baked into the process now, where I like what Brian Vieaux with MISMO said is you may not own the platform, but you own the risk. And so you have to do all this due diligence around it because you didn't create it.

But then the third one that we should probably talk about is equipping the organization. I mean, we, you know, Microsoft has Copilot out there, Claude has their desktop application now I think there's Cowork for a couple of them. There's all these applications that help users get better and a lot of them are actually pretty good now I mean some of them like you know I would say Copilot in Excel is now very good and so it can speed tasks and so I just want to know from you know maybe T.J. starting with you and then we'll kind of get back to Brian what does it look like as far as investment in AI if you're in the leadership spot at a lender. Where do you stake your bets?

T.J. Harrington (19:21.579)
Yes. To Brian's point I think there's been such a focus on the vendor space on productization of AI. I really think it's more about the core systems and being able to speed development and run your business the way you run your business, but taking care of all the things to modernize it. I mean, I think about all the systems and things we rely on today, the little patchwork items we have. Marvin, you and I spend so much time spotting issues with our partners in integrations and data coming over and data sanctity and all the little customizations and tweaks that have been made over the years as patches on just little break-fixes that just never quite get fully retired.

Those are the types of things off the shelf that AI can identify and fix and improve a business immediately. And I haven't even talked about, you know, any outsource process. I haven't talked about anything about layering AI into the actual manufacturing process, but purely from how you manage and run your business in the day to day from technology, everything else. It's really in my mind a change management boost. You know, you think about change management, regulatory things, product things. There's an implementation…

Rich Kuegler (20:27)
Yeah, speed to market. Right.

T.J. Harrington (20:29)
speed to market. It's things of that nature for running your business, not necessarily having a product or process. I mean just an example on the title side of the house at Stewart, right? I started at fifteen years old as an abstractor at a registry deeds office. And one of the things that I had to do and they made me do because I had the best eyes at fifteen years old was read all the cursive writing on all the ancient documents, right?

Rich Kuegler (20:57.665)
I thought you were gonna say make you go out and get lunch for everybody.

T.J. Harrington (20:58)
Yeah well that too. But they'd put me in a clean room with documents back from the patent of the king and having somebody with fresh eyes read that very intensive cursive writing. And now with with the AI models we have, it reads that better than I ever did as an abstractor. And so that's something that we're layering in as a tool set that quantum leaps our ability to be accurate and precise in our day to day.

So we're see things like that where, you know, beyond the platforming and the changing of how it and to Brian's point too, you go out there and you have SaaS models of layering in solutions to lenders today out of the box that model's almost gone as a viable model because you give me a couple of weeks in a codex and a coding agent, I could build that from scratch and maybe in a way that's that's bigger and better than what has been deployed as a product set. So I think it's really putting the future in lenders' hands to be able to run and build their business the way they want to, using these tools to design their process flows and how they work and organize their business in a way that we've never seen before.

Marvin Stone (22:10)
Yeah.

Rich Kuegler (22:11)
Yeah, T.J., I was gonna add to, excuse me, Marvin, was just gonna add to that like if you think about where certain lenders want to focus, you know, what I'm hearing is deployment of different AI technology and resources is really being driven by what are my critical business issues? How do I wanna formulate specific areas of the business that I want to go after? So if I've got some customization that might not be available in my off-the-shelf LOS, how do I replace the ancestors of AI, which I would say are all these bots that people, that some, some you know, tech-forward lenders put in put into play a long time ago. And now they can replace those on a more frequent basis and be more, just you know, shorter development time and shorter implementation time to make sure they can attract those niches and be successful in those areas.

T.J. Harrington (23:04.673)
Rich, I just keep thinking of that movie iRobot with all those old robots and containers in the valley. That's what we're headed towards, right? We're replacing RPA with just yeah, sorry.

Rich Kuegler (23:14)
Yeah, exactly. What do you do with all those old ones out in the in the robot graveyard?

Brian Francis (23:16.919)
Yeah, I think the other advantage you get with that, Rich, is because you have compressed the time to develop, it now allows lenders to take more of a risk and say, well, no, I don't have to make a multimillion-dollar multiyear investment with this technology partner that I think might improve my business. I can do something very quickly, target at maybe one branch or maybe one region and say, hey, is this really going to make a difference? And all of investment…

Rich Kuegler (23:43.455)
Right. You can pile it and fail fast. Yeah.

Brian Francis (23:45)
Exactly.

T.J. Harrington (23:50)
Yeah. And that's

Marvin Stone (23:50)
Yeah. So go ahead.

Brian Francis (23:50)
And all you've invested is tokens to do that.

T.J. Harrington (23:55.3)
That’s precisely the model of IMBs, but I actually see banks beginning to do that too, right? Banks who have invested in good AI infrastructure can now feel free to within the confines of the AI infrastructure and governance take those kinds of chances.

And that's something where you've not seen a lot of. When you think of banks you think of governance, you don't think of innovation. But they also have probably the biggest war chest for investment compared to IMBs. So maybe you do see large scale deployment and you know, banks have always had the lowest cost of capital, but what they've had is the most difficult, costly manufacturing process. That's been kind of their Achilles heel versus the agility of IMBs in the broker channels. Maybe you see that flipped on its head with AI.

Brian Francis (24:46.221)
Well, I can tell you this that you're gonna find the best COBOL program in the world and the best AS/400 system admin in the world is gonna be Claude Opus 5.5.

Marvin Stone (24:56.059)
Okay, so we got about five minutes left, and I want to hit governance and risk really hard here because that's really the those are the questions we get all the time. And Brian, you know, you kind of touched on the governance piece a little bit, but I think we're gonna have a paper that we can, really a document, that we can help lenders with kind of understand where the governance is coming in because it's not just lenders governing their own operations just from wanting to eliminate risk and change management and stuff.

No, it's coming down from on-high really. So, Brian, if you'll talk about that and then we'll talk about, you know, how we're gonna put some resources together for the lenders watching today.

Brian Francis (25:33.144)
Yeah, so I think the biggest challenge has been for lenders is well, how do I live in a regulatory environment and be able to speak to those regulators that I have to work with and be able to deliver that information. Now, the two entities primarily for mortgage are going to be Fannie and Freddie. Now, bank lenders, they've got their whole other side of their regulatory framework.

But from the Fannie and Freddie side, it's taken a little bit of time. And now we finally have Freddie finally released theirs in September of this year. Fannie's was a little bit earlier this year. And, as they want to be Fannie and Freddie, they're similar but they're different. So one of the things that we want to focus on being able to deliver is how do those combine together? Where do they overlap? Where are their differences? And it really comes down to what do I have to manage? What do I have to report on? What are the things that I as a lender have to look to? And then what do Stewart and IR have to do as vendors to provide inputs to you for when you do report that information up to Fannie and Freddie? But it's really focused around things about where are decisions being made?

So, are analyses being done by AI, are decisions being made by AI? It's not saying you can't make the decisions, but you have you as a lender have to quantify exactly where that decision takes place. Think about it from the credit perspective. If there's an adverse action, then you've got to be able to respond and know exactly, here's why this action was taking place. And there's risk involved in that.

If you're going to hand off particular portions of the decision to an AI, then you have to take the risk. It's very similar to what's been going on in the kind of the national world, the national discussion of an AI right now, as to you've got advocates that want to have regulatory capture that say we need government regulation so that we as the frontier AI companies don't have to put the guardrails around our system. We need the government to say that.

There's others that are saying, well, no, you are a company, you have liability. If you put out a product that doesn't meet those requirements or fails from a liability, then there is recourse that has to be done. So that's still an ongoing discussion. We'll see where that ends up. But from a perspective of how do lenders live in this world? It's just like it's just like doing a SOC2. It's just gonna have more categories and more information that you have to deliver. And it's our responsibility as vendors to deliver our portion of that information to you.

Marvin Stone (28:38.511)
Yeah, and real quick, you know, MISMO came out with their FRAME framework for helping lenders that are members toward this path. And it's very comprehensive. I mean, the MBA was involved and really did a lot of great work there looking at the regulations that are out there, regulations that are coming, Fannie Freddie guidelines on this. But it's kind of it kind of reminds me and T.J., Rich, you guys go back in this business. You remember the three-page service provider bulletin from the CFPB, and all of a sudden everybody that launched all of this activity.

Because basically it said the same thing with that Fannie and Freddie are now saying about AI, which is you're responsible. And so lenders get that. T.J.?

T.J. Harrington (29:16.139)
And that's yeah, and that's so speaking of the CFPB, right? We've we've known the CFPB has been greatly diminished under the Trump administration and what we've seen is States Act stepping up. So one of the interesting underlays or overlays we're seeing is there's some states that are very savvy to the ongoing discussions in AI. And we see Arizona much really kind of on the forefront, believe it or not. Their struggle has been with the deployment of AI in the REALTOR® context.

What are agentic agents doing, touching and concerning, providing advice and guidance in this in the home search process, and they put out bulletins around what is the minimum requirements of a licensee on the real estate commission. I think you expect to see the similar things for LOs and their lenders. You expect to see the same thing for title agents and escrow agents. So there's going to be things that are considered licensable activities that must be done by a licensed responsible agent that if you do delegate to AI, there is a throat to

choke so to speak from a responsibility standpoint. And so that's a layer that we'll see over time. As Brian said, there you might deploy a system but you own as you said through from Brian, you own the risk of that. That's going to be the same thing for originators, REALTORs®, and title agents.

Marvin Stone (30:30.455)
Well said. Rich, I know you come from an engineering background as well. One of the things that was interesting, and I think you were in maybe in this meeting as well, they were talking about how yes, it's like doing a SOC2. Yes, it's like you know your normal governance process that a lender might put on you or something. But it's different because you have to go do these ongoing tests. You know, okay so if I have an AI system you know that's in place today and I document it there, one of the things that they came back to was how often are you testing the output? Because we know the output can change and there can be model drift and those things. So just one last note here on sort of the mechanics of it all. That's one of the things that I'm sure you're hearing from lenders out there.

Rich Kuegler (31:13.795)
Yeah, it's like how do you drive more of an in-process control rather than an end state QC check after the fact? You know, so you can do as many audits as you want. And I think Brian mentioned it earlier. You know, if if a lender makes the decision that one of their big business cases is a certain process, they also need to be able to demonstrate what controls have been built around that process, what guidelines are they are they working toward as all these other guidelines, whether it's Freddie, Fannie, FHFA overall, the s individual states themselves could be a number of different entities that are gonna weigh in. But in the meantime, what kind of decisions can you support and justify not only from an ROI perspective, but truly from a governance and responsible lending perspective?

Marvin Stone (31:53)
Yeah, for sure. Well, this has been great. And so, Brian, as we wrap up, thank you so much for taking time out to kind of talk tech with us here. And you know, if we're not all replaced, we'll come back—by AI—we'll come back on another episode. Maybe it'll be bots, maybe not. But this has been great. One of the things that we'll we'll be doing, I think we alluded to it, is we'll be making basically show notes available for this episode, but also we will have some sort of guidance on you know how to find the materials that really lenders need to know about in terms of governance. So we'll have some great resources there that we'll make available as well. So that's it for us. Just thank you for everyone for joining us. And that's it for this episode of Stewart in the Studio.

Marvin Stone (32:36)

That's it for Stewart in the Studio. Find more episodes and insights at stewart.com/lender. We'll see you next time.

Disclaimer (32:43)
This podcast is for informational purposes only and reflects the views of the speakers. It should not be considered legal or business advice, and listeners should consult their own advisors before making decisions.

About Stewart in the Studio

Hosts:
Marvin Stone, Senior Vice President, Director of Strategic Initiatives
Rich Kuegler, Senior Vice President, Director of Client Success
T.J. Harrington, Senior Vice President, National Product and Sales Enablement

Stewart in the Studio is a monthly podcast from Stewart Lender Services designed to keep today’s mortgage professionals informed, inspired and ahead of the curve. Marvin, Rich and T.J. share 80+ years of combined experience and dive deep with industry experts to uncover the trends, topics and tech shaping the mortgage lending landscape. Like, subscribe, and join the conversation. There’s always a seat in the Studio.

The information provided in this podcast is for general informational purposes only and reflects the opinions of the individual speakers at the time of recording. It is not intended as legal, financial, or business advice. While our team members are experienced professionals, listeners should consult with their own advisors before making any business or investment decisions. References to products or services are provided for informational purposes and do not constitute a guarantee of results. Stewart Lender Services makes no representations or warranties regarding the completeness or accuracy of the information discussed and assumes no liability for any actions taken based on this content.