How AI Is Being Used in Mortgage Valuation

Marvin Stone (00:01)
This is Stewart in the Studio, the podcast where mortgage professionals stay ahead of the curve with expert guidance from Stewart's thought leaders. I'm your host Marvin Stone. And each month we dive into trends, topics and tech to transform your business. Let's do this. Well, hey everyone, welcome to Stewart in the Studio. I'm Marvin Stone. And today we have both of our regulars in the room. Rich, why don't you kick us off with just 30 seconds on who you are and what you're working on here at Stewart.

Rich Kuegler (00:26) 
Hi Marvin, Rich Kuegler, I'm the Director of Client Success here at Stewart Lender Services, just coming off a very busy, busy conference season. And a lot of the key focus was really on helping lenders understand how they can plan for expanded loan volume, looking at capacity and efficiency in their productivity for the origination process, particularly in home equity, which is continuing to emerge as a big channel. 
  
Marvin Stone (00:53) 
Yeah, for sure. Thanks, Rich. T.J., let's do the same for you. 
  
T.J. Harrington (00:57) 
Good afternoon, T.J. Harrington, Product Strategy for Stewart Lender Services. As Rich was busy with conferences, we've been having conversations around both originations and servicing work, both performing and non-performing servicing. We see a lot of consumer pain in the market right now with default rates ticking up and trying to find the most efficient solutions to reduce servicing costs and drive outcomes. 
  
Marvin Stone (01:21) 
Yeah, sounds great. So Rich, I know you've been on the road or in the air lately, a number of conferences and several sort of high level meetings with some of the top IMBs in the market. Can you just share kind of the outlook and conversation, maybe a little bit about what they're thinking about technology and AI? 
  
Rich Kuegler (01:40) 
Certainly. There's not a time or not a meeting that goes by where AI is not a topic and not a topic of of large conversation and large interest. I think there's going to be a ton of investment in really exploring the potential for AI in a number of different facets of the business. I think the most common one right now is really agentic AI. So how can you help improve the customer experience by introducing models or leveraging that technology 
  
from a productivity standpoint? But there's also a number of back-office capabilities and really data capabilities that are really being discussed at very high levels that I think we're going to see a lot more of and probably talk a little bit about today if I'm not mistaken. 
  
Marvin Stone (02:25) 
You're good that way. T.J., I know like Rich, you've always got your finger on the pulse of the technology of the industry. But also, I know you talk to regulators, talk to people in compliance, all different levels. What are some of the challenges you're hearing? Or are you hearing challenges from lenders like what to do about AI? 
  
T.J. Harrington (02:45) 
Yeah, it's really interesting. There's something coming out of the administration now being floated as a test balloon around outlawing offshore call centers, which is pushing everybody in the world, as Rich just said, to agentic AI for call center work. And it's really, really interesting. I think the name of the game is really going to be the building the guardrails and the governance around AI to be successful, to take on more tasks and to make sure there's not bias or other regulatory issues that pop up as the models get more sophisticated. 
  
I think that's really what people are struggling with today, where they see the use cases to say, this AI application is well-suited to what I could do from a replacement or efficiency standpoint. But again, what are the rules of the road? What is the governance? How do we do compliance, QAQC testing? How do we do regulatory testing? How do we ensure that anything touches consumers, works the right way, and delivers the right outcomes? 
  
Marvin Stone (03:37) 
Well, those are the things you never hear about in this whole AI hype cycle. That's all the hard work that gets done. So that brings us to Andrew. So Andrew Komaromi, thank you so much for being our guest today. You're here at Stewart in the Stewart Valuation Group. And for those who aren't familiar, Stewart Valuation Intelligence is one of the top AMCs in the country and really an early adopter in the AI space, not just in terms of getting headlines and things like that, but really in terms of making 
  
a difference in the experience for lenders and internal staff as well. So Andrew, with that, give us just the 30-second high level about who you are and what you do here at Stewart. 
  
Andrew Komaromi (04:19) 
Yeah, excellent. Thanks. Happy to be here. Appreciate the invite and always fun chatting with you with you all here. So I've been a part of the organization in some capacity since 2009. I had up the senior director of data, AI and automation. So it's a really fun, cool job here. I get to solution and look for what we can do to help help all the humans here, facilitating all the orders that come in. You know what, what can what can our team build? What kind of tools can we roll out? 
  
We spend a lot of time thinking, a lot of time dreaming and a fair bit of time building. And it's nice to see when we executed. Hey, look at that. We got that much better quality or we got that much stronger in the current time or just made the borrower's life and the lender's life that much easier. 
  
Marvin Stone (05:05) 
Yeah, I like that “helping the humans.” I think that that gets put on a T-shirt or something somewhere. 
  
So we'll preview the conversation to kind of talk about where we're going today to sort of set the expectations. So first we'll start, you know, kind of talking about why SVI looked at AI, you know, why did the organization really look at AI as a possible solution or really more of an opportunity, I would say. And then really kind of get down to the core tenets of the SVI approach. Like, you know, I've heard a lot about vibe coding. Do you just pick up and start coding something or what does that really look like? Exactly. And then really talk about your Sage 
  
solution is as sort of the right seat or since you're a commercial pilot, maybe the co-pilot. I know that's an often used term, but then really, I really want to dig into the responsible AI piece. Every large organization talks about responsible AI. So that's always where if the conversation doesn't start there, it ends there. So let's start with the big picture. So most everybody's heard about vibe coding and AI can write software, write code for you now. So why did you guys 
  
take sort of a, I think you took sort of a deliberate approach. Tell me a little bit about your approach and kind of why that, why you took that approach. 
  
Andrew Komaromi (06:17) 
Yeah, absolutely. So I mean, so we've been doing AI and honestly in some capacity since at least 2010 in some form for another. So I mean all of this started with data and being able to kind of augment our team with with insights. So I mean we I came in, you know, looked at the processes again. This is going back to 2009, 2010, 2011 and we have so much data out there we can really look at. 
  
some patterns, start looking for pattern recognition, start looking for some predictive analytics there. And all of this is just kind of meant, like you said, as a co-pilot, we're trying to augment the person who is actually either managing the order, assigning, finding the best crazier for this assignment, looking at the quality control on this, right? An order comes in, what do we really have to look at to have this be a clean, high-quality review? Because you can't inspect quality to something, so 
  
our team is there, they're making those judgment calls. They're experts, they're licensed appraisers, but can we put tools in their hands there? A lot of that started back with the IMD. So that's the integrated market data engine that we've rolled out years ago at this point. And that whole format and that whole formula really built the foundation for us and kind of laid that machine learning. It's not quotes, that's actually 
  
talking about the machine learning groundwork there. You know, it comes 2018, 2019, 2020, OpenAI, ChatGPT, everyone starts really using these models. And you suddenly kind of took this sort of, sort of science-y kind of backrooms behind the scenes approach, and everybody has access to it now. So now there's this potential for a lot more to do with it kind of democratized it. And so that, that comes with obviously some risks, which I know we'll talk about responsible AI. 
  
Also comes with huge opportunities. So mean, looking at the foundation we'd already built, it seemed like a natural shoe-in to start looking at these models and think, well, this is the next step. You know, we are consistently and continuously looking for how we can guide the, help guide our human team, help the humans. So if there's a tool out there, we want to explore, we want to build it and we want to iterate on it. So, I mean, that's, that's really where, where kind of it started. It wasn't even like, why are we doing AI? 
  
Andrew Komaromi (08:40) 
We've kind of always been doing it. And this just gave us that many more tools, right? You know, we've got your big ones out there, Gemini, Anthropic, Claude, ChatGPT, and a host of others. So really, was just a natural stepping stone for us. And I'm very excited about what we built and where we are looking to keep on keep developing that, right? I mean, it just comes down to the lender borrowed how we make your lives easier. 
  
Marvin Stone (09:04) 
For sure. So I guess I've got, you know, in a conversation with T.J. and Rich in the past, we were talking about like, what does AI do for you that the traditional coding approach can't do? And then really now the commercial models are so strong. You mentioned Claude and ChatGPT and these others, they're so strong and they have billions invested in them. So what does AI do for you that you couldn't do in either the traditional approach or just an off-the-shelf AI tool? 
  
Andrew Komaromi (09:35) 
Yeah, I appreciate that question. AI kind of really helps us us dream, right? So you have that creativity there. You've got the ideas you're talking. In my role I get to talk to a little bit of everybody at SVI. That's the assignment team. That's the client relations team is to QC team. It's the sales folks. It's everybody. And so you know as you as you hear these you know, roadblocks people are trying to solve and boy, if I can only do this we would 
  
get this done faster, if only I had this. Now you have the ability here, and I almost prefer someone else coined this here, so I can't take full credit for it, but I've kind of mainstreamed it, but co-develop. I think vibe coding implies you're kind of just in the back seat there. But if you're co-developing and back to the co-pilot, right seat kind of expert there, it's having like a super capable person next to you 
  
that you could say, I've got this idea. I want to build a platform that can do this and they can do that. And it will kind of just do it for you. It's, it's absolutely not in its native sense, production ready, but it helps you kind of sus out these ideas. So something that would have taken, you know, our group one, three, six months, maybe a year or two to sort of brainstorm and test and build and whatever. You can get that done now realistically 
  
in a matter of, mean, in some cases in a matter of hours, you know, we're looking for a new, Hey, something that can do this. Let's type it up and see if it has feet, if it works. I mean, that's, that's the whole co-development aspect of it. And, you know, again, they're, only as good as, as how you prompt these things and the care you take with it. So it's by themselves. They're, they're not, you know, they're not the end all be all. So I think it's, it's, you know, something that's been a really fun 
  
fun process with our team is just that a whole whole code development and what's the right context? What's the right prompt? And then like I said, take something we have a we have a nice kind of test development environment here. We can host it, post it up, share with our group and and you know, kind of rapidly iterated, see if something has legs or not. You we can sort of try 10 to 15, 20 ideas in the span of a week and see which one which one might land. 
  
Marvin Stone (11:51) 
But really, you've got such an interesting place, obviously, Stewart 
  
is a very large publicly traded organization, large IT operations, many IT operations and development operations throughout the company. And yours is dedicated specifically to Stewart Valuation Intelligence. But lending is highly, and T.J., I'm going to ask you to kind of weigh in on this. Lending is highly regulated regardless of what type of lender you are, whether you're a commercial bank or IMB, what have you, or that's on one side. And then on the other side, you know, appraisal, that's obviously 
  
the place where SVI sits. How do you see the intersection of that when it comes to guardrails? I mean, how in tune are lenders really with the guardrails needed in something like this? 
  
T.J. Harrington (12:37) 
It's getting more and more attention, Marvin. And I think to what Andrew said, you know, the idea of driving your frontline folks to be able to innovate and say, what problems are you facing today? And here's a tool that can help you figure that out. That's fantastic. And the idea of AI of making your most productive people even more productive, you know, makes your best people better. So that from like an employee enablement perspective is, I think, low hanging fruit. What you get into trouble or challenges with from a governance perspective is when you're 
  
taking wholesale tasking off rather than just enabling a human, you're taking a wholesale loan processing, underwriting, having conversations about it. The conversation really around AI, like maybe two years ago, was really, really interesting because you saw real estate agents using these public model AIs in lieu of providing real estate services. Some of those things were wrong. There was ghosts in the systems. There were wrong answers. There was mismeasurement. 
  
And the departments that regulate real estate agents were like, well, who's neck do we wring? We want to have a responsible party and you go, it's Claude, Claude. It's Grok. It's Grok's fault. And so really you can't blame the model. And ultimately regulators are saying, we need a responsible party. Who's the licensed party that's going to be the one who holds the risk associated with the use of AI and lenders and other constituents in the market are very sensitive to it. 
  
And so they're building out the routines and guardrails now. They're hiring the consultancies to say, hey, what is the playbook? What are our policies and procedures for AI governance? How do we manage it like we do for standard development today? What is the testing? What does it look like? What is the QAQC routines? Just like we do call samples for call centers, what are we doing for AI? What are we doing to make sure that the way that we're touching consumers works the right way? 
  
So I think there's a whole infrastructure and industry being built around the governance side that's, I think, evolving as fast as AI is evolving our businesses today. 
  
Marvin Stone (14:37) 
Yeah, really well said. 
  
Rich Kuegler (14:39) 
I think you could almost, excuse me, Marvin, I think I was going to say you can almost look at it like the implementation of any new technology into the mortgage process, where there might be a great promise of, oh there's a ton of efficiency to be gained. There's a ton of advances to be made. But at the same time, all of that needs to be metered against the requirements of the market, against the regulatory framework that the market works within, and then also the needs of investors and that community as well 
  
to make sure we've got something that's sound, or I should say that takes our sound principles and can deploy them better through some type of a technology advancement. And that's, think, where the challenge is right now, is the rush to implement all these efficiency tools to Andrew's point. And at the same time, make sure you're not running yourself into a spot where you run afoul of any regulatory requirement and put yourself in a potentially riskier situation than you were before. 
  
T.J. Harrington (15:36) 
And at the risk of calling somebody out in the market right now, what we see in the servicing side of the world is the rise of Valen, which is an AI native servicing application, kind of challenging MSP for hegemony in the servicing space. And they're still working and rolling out their default modules. They're not default-enabled today. And I think part of that is the need to be so careful where you're touching at-risk consumers in the default space. 
  
So I think they're getting there. What they've rolled out is incredibly impressive. And from a speed of development, change management, et cetera, they're doing it the right way. But one of the benefits of MSP has been that it's had the tires kicked and every regulator knows how it works and are comfortable with it. And it's 1970s best technology with Cobalt. And so you know that the mainframe works and the data works and it's predictable. And so part of this, I think you see innovation at the edge cases first. 
  
Where maybe you see IMBs and others role in technology, this new innovation, and as it gets to the more core banking, as the regulators get savvy to it, as it develops and the governance and the infrastructure develops, you begin seeing bigger players adopt the more sophisticated system. 
  
Marvin Stone (16:43) 
Yeah. 
  
Andrew Komaromi (16:43) 
I really like your point on human enablement because that's exactly what we're trying to drive our team toward. We want to build the tools like, hey, I need this tool that can do this. You show up with the hammer to job site, it's not going to do the job for you. It's going to make it a lot easier. You're going try to hammer that nail in with a piece of wood or whatever. It's very easy to say, just have AI do it. You can't. It's not possible. That's where you run afoul. 
  
We're winning by inches, right? 
  
T.J. Harrington (17:14) 
And it's the thing you guys are in SVI, you guys are evolving enablement for your for your teams, which means you do more with the same people that the route items are sucked out of the daily life of a worker. They're they're concentrated in highest and best use of their skills and abilities. And I think that the business tools for AI, that's a no brainer. That's the lowest hanging fruit is human enablement. And I think when you move beyond that to whole processes or having AI. 
  
run the show in certain places where you begin to run into some of the concerns. 
  
Marvin Stone (17:45) 
So that sets the stage for everything SVIs building. Basically, we're going to come back after a short break and we'll dig into Sage a little deeper and we'll talk about really the core tenets of SVIs approach to AI, the unifying intelligence layer, Andrew, how you bring it all together. We're going to talk about where those decisions come into play and how they matter. And then we're going to talk about how you've built this, not just vibe coding, but really built it for scale and control. So we'll be right back after this message. 
  
Nationwide Appraisal Network (18:14)
We are proud to welcome Nationwide Appraisal Network, now part of Stewart Valuation Intelligence. 
  
Nationwide Appraisal Network brings over 20 years of experience and a reputation built on trust, service, and deep market knowledge. The people behind that legacy and how they serve clients are vital. Stewart brings advanced valuation intelligence, AI-powered insights, and workflows that drive speed, consistency, and smarter decisioning. 
  
Together, we bring proven expertise and expanded national capacity across appraisal management and valuation solutions. That means the same trusted service, now enhanced with data-driven intelligence and capabilities to handle high volume and complexity with confidence. Nationwide Appraisal Network, now part of SVI. One partner, ready for what's next. 
  
Marvin Stone (19:12) 
Well, welcome back everybody. So Andrew, let's dive in on some of those, really the core tenants here. So let's talk about how AI is used specifically to connect people, data, processes, so you create more consistency across that valuation lifecycle. 
  
Andrew Komaromi (19:30) 
Yeah, absolutely. So again, the core tenant behind all of this, I just sum it up to one kind of statement, we are keeping that human there in the driver's seat. People, they have that qualitative experience, that expertise. So you're staying in the driver's seat. wonder what we can do with our team here is, like you said, unify the data, the processes, all these kind of disjointed things that might live. Then you click here. 
  
Now go there and now email this person. Now check this document. That's where we can build a model which will lead into Sage that seeks to unify all those things. You've got the L1 you can touch that you could chat with and in the background that's doing a lot of these fancy automations to assist the person. So what may have taken 10, 15, 20 clicks to go through, check X, check Y, go here and there. 
  
A lot of this is just happening in the background. So your next task has already been nicely polished up. It's still up to you to take it across the finish line. But you're not having to go out in the field and grow the wheat, mill it down and all that stuff. You can do a nice, pretty piled-up set of ingredients. Maybe your water is already boiling. It's all there for you to get it done. 
  
T.J. Harrington (20:50) 
So you're saying that is your sous chef at SVI. 
  
Andrew Komaromi (20:55) 
You know, I do have a model that I was building. Like here's what's in my fridge and you know, it kind of knows the foods we like, so it's definitely some type of (indistinct). 
  
Marvin Stone (21:05) 
So Andrew, the way I understand it with Sage is Sage is not only something of basically very targeted chat experience that your employees can use to get very specific information about any rule, any transaction, basically anything in the whole SVI ecosystem, and it's embedded into existing workflows. It does work behind the scenes. Is that accurate? 
  
Andrew Komaromi (21:32) 
That's exactly right. So on the front side, we wanted to cut down on, you know, making an educated guess or having to, you know, like most of us, honestly, a lot of us are remote. So it's not so easy to just ping the person at desk next to you and then they're shooting the teams, they're shooting them, whatever message. So if I want to check with some expert source on, hey, this house has X gallons per minute flow rate on a well, is this okay? 
  
You know, I'm no longer having to do like the control F, the peck and hunt on Google, open up some huge PDF. He just has the model and it's going to tell me yes or no. And it's going to give me that reference. So if I, don't believe it or whatever, I want to have the actual source for whatever reason, it can take me down to the exact regulation. The exact internal policy and procedure or whatever precedent it has to them. So it's, it's not using general knowledge. It's using all of our own. 
  
policies, procedures, and all the data we fed. And it can get out of the specifics of I'm working on order one, two, three, and knows everything about that order. And you can ask it for details around that. So it's, it's a step beyond just like a kind of a customized chat for, for a lot for better word. Um, that's only a small part of it. So the whole brain, it's what's driving a lot of our automations. So you mentioned the agentic actions, a lot of these things coming in. 
  
You know, somebody reaches out and they maybe need, you know, some updated instructions for getting out to the property. There was a massive storm and the roads flooded. need two more days to get out there. All of these things, we can trigger several events already. So when it comes up to that assigned person's desk or the person managing it, they can, they've already found a lot of that done. They can just click two buttons, double check and verify it and, and make it happen. They're out having to kind of do all that, all that leg work 
  
themselves. They just take it over the finish line.
  
Marvin Stone (23:26) 
Yeah, no, that's great. know, Fred Eppinger, our CEO across Stewart, often says, you know, the goal is to be the premier premier company as a service provider and a partner in this industry. And one of the ways we do that, I think, is just by having the right answer faster. And, you know, when you talk about having to look through documentation and control F to find something where we have all the information. 
  
It's just finding it is challenging in any organization. It doesn't matter what industry. So I think that's where a lot of it, connecting the policies, connecting the processes, providing the context, all those controls in there. That's where it seems like SVI Sage plays almost like this hidden hero that you don't really see necessarily, but it's just making everything more accurate and faster across the board. So you have better decisions earlier in the process, fewer downstream 
  
issues and surprises. mean, just, it sounds like it's just better all the way around. 
  
Andrew Komaromi (24:26) 
That's exactly right. And then that's true to your assignment, to your QC. So I guess not just the channel tool. Is this flow rate good or is this easement okay? It's the actual specific orders itself. It's what we've learned from our lenders, right? Every time we get some feedback back in an order, that's part of the model now and know what to look for in the future. It’s all in there. 
  
Marvin Stone (24:46) 
So let me ask you a question kind of going back to comments that Rich and T.J. have made about the guard rails and things like that. You know, if I'm a lender and I have no, if I'm a lender, I don't really know your system. But I can, I'm assuming this goes through all the governance of Stewart, all the security where, know, in New York Stock Exchange traded company. So the security controls are very tight. But one of the things that's really been making the news lately with all these new models coming out on the commercial side is 
  
things change. So a model that may work today, like I think Claude Opus 4.6 was very intuitive and then they released 4.7 and it's very literal. So how do you protect against changes like that? That, you know, I mean, your technology, I know it has to go through all these checks and controls, but what happens there? What do you do with that? 
  
Andrew Komaromi (25:40) 
Yeah, and that's that is something we are keenly aware of. And when I hear people say things like, oh, it's the black box that makes me cringe because it absolutely what we're building absolutely is not right. So explainability is the core foundation here. So I wanted we have a trace back. Soall of our this, for example, you ask the tool something you don't hate again, go back to the flow rate of the well, for instance, we can go back into logs and see exactly what 
  
path that went down to return that answer. So yeah, we have the reference, that's great. We can actually see what sources of ping or when and how confident it was in each source. So all that's in there. We do consistently look at the accuracy of it. So there's, without getting too into weeds, the models have the capability to kind of assess their confidence and the answer it's giving you. 
  
And again, without getting too deep into it, a lot of these models, they don't want to just tell you they don't know. So if you're filling out a multiple choice exam, for instance, you don't really know the answer. You're not going to leave it blank. You're still going to make your best guess on, what option is it on multiple choice? Models can do that too. So we're keenly aware of that. And we actively use very specific, like context window prompts and all these things so that it would rather default to saying, I actually don't know this. 
  
Or, this, can't help automate this process. This has got to go fully manually because we don't want it making that guess unless it's very, very, very sure of that guess. We'd rather just leave that blank, have it go to the person, they manage it than just, I think it's C and maybe it isn't, right? Because there was a lot at stake here. So we monitor the same data, our team is keenly aware of it. And we do have a broader AI council within Stewart that we run through all the responsible standard AI document everything. 
  
You know, all the GitHub controls, everything is well documented. 
  
Marvin Stone (27:36) 
Yeah, for sure. So your AI doesn't tell me what a great idea I had or, you know, all of that that we get with some of the commercial providers out there. 
  
Andrew Komaromi (27:47) 
We've tried to dial that down. 
  
T.J. Harrington (27:50) 
Yeah. 
  
Marvin Stone (27:51) 
So one of the things you mentioned that I think is critically important is the AI council across Stewart. John Hamm, know, hats off to our CIO who put that in place very early on as others were still learning about AI. He put the entire council together very much across discipline exercise across the entire company. And so we get a lot of insight both nationally and internationally how AI is being used across the entire business. And all those controls are in place, which is great. 
  
So we always know that those guardrails are there. So Rich, T.J., I'm gonna bring it back to you guys to close us out. So we've talked about SVI Sage. Rich, mean, one of the things that you're always being asked about is speed. How do we get faster but still maintain control? I mean, what's your take on everything that Andrew gave us today? 
  
Rich Kuegler (28:44) 
I think it's a great example and real testament to Stewart's overall commitment to the industry and Stewart's commitment to being the premier services provider. We are in a services business, whether it's in our title operations or in our appraisal operations. Really, the focus on reducing friction in that process or either of those processes is key because our originators and our servicers and our capital markets investors 
  
They need to have a better experience. They need to have the confidence that Stewart is a provider that's there for the long term. And I think the commitment to developing this AI technology in a very favorable and also guarded and appropriate method is really helpful in demonstrating that to our clients and also making us a better partner for our clients going forward. 
  
Marvin Stone (29:36) 
Yeah, no, very well said. T.J., any final thoughts before we close out here? 
  
T.J. Harrington (29:41) 
I think Rich nailed it. I think that we've seen a lot of talk about investment in technology. last 15 years, we've seen every newfangled process technology thing that's going to come out and change the world. Last time it was blockchain, a lot of hype. I think AI is one of the ones that's catching fire because there are use cases that are changing the world. So hats off to SVI for making that investment. And I think our clients will see the difference in the quality of 
  
the appraisals, the speed, the turn times, the cost to fulfill the whole shoot and match. I think that across the Stewart enterprise, we're making these investments to keep pace with the pace of change in our industry and looking forward to what tomorrow brings. 
  
Marvin Stone (30:24) 
Great. And Andrew, one thing we didn't discuss, I'll just give you one minute here to kind of close this out. What should we look forward to with the AI space or with Sage? Either one. 
  
Andrew Komaromi (30:36) 
So much I wouldn't even know where to start. have a lot of exciting things coming down the pipe, most interestingly without giving too much of it away. But we want to get more of the upfront potentially risks of the order known when it comes in. So we can start preparing for that and we're not finding out about it when it's time to deliver this thing. So we can arm ourselves better upfront with all the information, then that gives our assigned team 
  
The appraiser filling it out and our QC team that much more tools again in your tool bag when that order comes in that they're able to just pass it right through the client, but would have taken an extra three days a month ago is now three days less today. So that's what we're most focusing on at this point is that upfront risk detection. 
  
Marvin Stone (31:23) 
Great. 
  
Now that sounds exciting. I think everybody would like to eliminate more risk at the front of the transaction. Certainly rather than finding out the surprises later on down the line. So Andrew, thank you for what is probably the first zero-hype AI discussion on the internet. Rich T.J., thanks so much as always for co-hosting this. And that's it for this episode. If you'd like more information, please visit stewartvaluation.com and we'll look forward to having you join us again for next episode of Stewart in the Studio. 
  
That's it for Stewart in the Studio, where mortgage professionals turn for fresh thinking and real-world solutions. Find more episodes and insights at Stewart.com slash lender. We'll see you next time. 
  
Disclaimer (32:05) 
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.