The mortgage servicing industry is shifting its mindset around compliance, recognizing it as a foundational infrastructure rather than a separate product or add-on. This approach reframes compliance as an integrated, always-on element of servicing operations, embedded within platforms and processes.
As AI and automation become increasingly central, they are not only enhancing operational efficiency but also redefining compliance management to enable real-time monitoring, predictive analytics, and seamless regulatory adherence.
This Leaders' episode with Matt Tully, Chief Compliance Officer and Shane Leonard, SVP, Chief Product Officer, at Sagent explores:
- How infrastructure-level compliance is reshaping servicing strategy
- Where AI embeds compliance – and the risks and rewards
- How real-time monitoring is changing regulator and borrower relationships
- The biggest servicer challenges in moving beyond bolt-on compliance
- What's next for AI in managing risk and regulatory change
Transcription:
Transcripts are generated using a combination of speech recognition software and human transcribers, and may contain errors. Please check the corresponding audio for the authoritative record.
Michael Moeser (00:15):
Hello everyone and thank you for joining us. I'm Michael Moser, senior content strategist at National Mortgage News and I'll be your host for today's leaders episode. We're diving into a topic that's rapidly redefining mortgage servicing landscape, the evolution of compliance from a standalone function into core infrastructure. Instead of being treated as a built on or afterthought, compliance is increasingly embedded directly into servicing platforms and process that are always on and always integrated. At the same time, advances in AI and automation are accelerating this shift. These technologies aren't just improving efficiency, they're enabling real-time monitoring, predictive insights, and a more proactive approach to managing risk. Joining me today to share their perspectives on this transformation are Matt Tully, Chief Compliance Officer and Shane Leonard, SVP Chief Product Officer at Sagent. Welcome, Matt and Shane.
Matt Tully (01:17):
Thanks for having us. Appreciate the opportunity.
Michael Moeser (01:20):
Now, before we dive into the core of our conversation, Matt and Shane, can you share a little bit about your roles and priorities that you're currently focused on?
Shane Leonard (01:31):
Well, in my role as chief product officer here at Sagent, it's incumbent within myself and the team to make sure we look forward and not look backwards to understand where we're going, understand where we've been, but really have a good grasp of the technology in front of us and the things that are going to change.
Matt Tully (01:52):
And from a compliance standpoint, my responsibility is to make sure that our platforms are compliant and up-to-date with all of the regulations, industry guidelines, rules, and other feedback we get from industry regulators to make sure that as our clients are out there in their day-to-day operations, we're giving them the tools and the resources they need to be compliant.
Michael Moeser (02:17):
Let's start with the current state of compliance today. It has traditionally been reactive and siloed. Where's that model breaking down most today and what's forcing servicers to rethink it now?
Shane Leonard (02:29):
Well, I think from a technology standpoint, we've always been handcuffed. We've had to use outside technologies to deal with what we refer to big iron technology, difficult things to connect. From a lineage standpoint, understanding something that took place 60 years ago in the documentation and then bringing that forward to today is next to impossible. With a modern tech stack and an understanding of how we can connect those immediately from the beginning, which is what Matt and I have strived to do with our two teams, working together hand in glove, as we say, to make sure that we understand from day one if we build this thing, why did we build it and what was the intention of the solve? And then to make sure that whether it's a regulatory body or a GSE, that we can trace that back easily without a lot of difficulty, without spreadsheets that it's natively built within the system and that we've documented in such a way that something changes, we can get right back to the center of the subject matter, quickly determine what the change actually needs to be as opposed to trying to guess at it, spending weeks on it.
(03:52):
We can do it in a matter of minutes.
Matt Tully (03:54):
And I'll hit on a couple themes there. Shane used a few words in particular that I think are really important to understanding the challenges of today versus where we want to go. Lineage being the first one. With DARA, what we have done is created a lineage between compliance requirements and servicing features and ted all of those together. Anyone that's in mortgage, especially mortgage servicing, lives by control matrices and other things that tell you what the requirement is and who's responsible for that. And when you have a legacy system, what you have really is years and years and years of code that has been built on top of itself because we all know a lot of the investor requirements and a lot of the regulatory requirements have grown and really expanded at an exponential rate in the last 20 years. And Shane talked about you've got 50, 60 years of code.
(04:45):
It's kind of the analogy I use is like Jangablox that are based on top one another and anyone who's played that game knows all you need to do is pull one out and the whole tower comes down. We know where things are, but every time there's a regulatory change or an investor change, it requires us to dig through code. And what we've tried to do with the new system is rather than keep piling things on top, we've built side by side and you have this ... Again, I want to emphasize what Shane said, this lineage, this connection between compliance requirements and servicing requirements that allow us to track the function and more importantly, track the change and make those updates faster than we're able to do in the legacy stacks that we have today.
Michael Moeser (05:29):
I like the analogy because I think as the Jenga Tower builds, that definitely becomes more complex and yes, you don't want it to fall down. Certainly appreciate that analogy. But now that we've talked about compliance becoming infrastructure, what does that actually look like inside a servicing platform and how is it different from what most servicers are doing today?
Shane Leonard (05:54):
We talked a little bit earlier, I mentioned the fact that with old tech stack and what you can and can't do, new tech stack is much different. You can connect to dots easier. Ultimately, it's about the user experience and what the servicer sees on the screen. And we've done some simple things to solve for some of these, the ability to get back to information. But probably the most consequential thing was making sure that as we went through the process, we documented it because while doing so, it allows us to use this new fangled technology called AI and we can look and know on a daily basis exactly what changed from the day before. We don't have to guess at it because one of the things if you train AI properly, it can do without any problem is say, it looked like this yesterday, it looks like this today and here's the difference.
(06:47):
There's no guessing at there. There's not a lot of things that can go wrong in that process. And then the ability to add the human in the loop through workflows and not just have the tool sitting out on its own, but have it native to the infrastructure that you have. None of that works if you don't start with the lineage from the beginning.
Michael Moeser (07:08):
I was going to say that sounds really powerful because I can imagine that an individual might be looking at what was different or what is different from yesterday and instead of that searching, you're almost saying, okay, here's what's different. And now having the human focus on the difference in the status, I guess.
Shane Leonard (07:30):
That's exactly what's happening. And I'd love for Matt to dive into that because his team is the human in the loop there. We can't ignore that. We're using the tool properly. We're not trying to replace a human. We're trying to supplement or augment what they are to push it forward faster to get to an answer that you're trying to get to anyway. And now the machine can actually do that math, so to speak, this plus this equals this or this minus this equals this and give that answer completely accurately. And now it's a matter for us to just determine what are the changes that need to be made within the system. And what we do between the two groups is exactly that an environment that's collaborative that says, is this just simply an enumeration change? Do we need another field, a data point to actually track said change or do we have it already?
(08:24):
And we can go back to the servicer and say, "You're covered." Matt, I don't know if you want to expound on that, but that's how our two teams work together and using tools that have been developed in a collaborative fashion to do that. Yeah,
Matt Tully (08:38):
I'll definitely double click on that. And Shane talked about the usage of artificial intelligence. It's important to emphasize as all of us are experimenting with AI, we've all seen the hallucinations and the high confidence answers that actually aren't really grounded in anything. And that's what happens when you just point AI at the internet and don't train it. We've taken a different approach. And before I get to the AI, let me tell you a little bit about how we built compliance's infrastructure into the DaraCore system. We spent a solid eight months going through every regulation investor guideline, insurer guideline, industry guideline, think of CDIA, Metro two credit reporting, and breaking those down literally line by line. I talked about control matrices previously. We took all of these and created a giant control matrix. The last count I had, we had 8,879 unique requirements that a mortgage servicer has to comply with at the federal level, the investor level, and even down to the state level.
(09:45):
And we took all of those and we put them in a database and that database sits side by side another database, which is the product features that Shane and his team have come up with that basically says, these are the things a mortgage servicer has to do day in and day out. And we've gone through this painstaking process of tying compliance requirements to servicing features and vice versa. So we have that lineage we talked about and that's really important in the context of AI because again, we're not just pointing it at the internet and saying, "Tell me how to build a compliance system." We built all of that infrastructure in and then you take AI and place it on top of that. So Fannie Mae or Freddie Mac or Ginny May or the CFPB updates a rule or a requirement, our tool can actually go out and not only look to see what the change is, but compare it to the compliance database and then get informed insights into where in the feature database we're going to have to make updates as a result.
(10:49):
But there's an important but here and Shane alluded to it, you still need the human in the loop. We're not relying on AI to just do the work for us, but we are relying on it as a tool that can allow us to more quickly analyze regulatory change and more effectively and efficiently pinpoint where in our system we need to make updates as a result of that change. And that's a very powerful combination that doesn't exist anywhere else in the industry today because you're relying on legacy stacks that don't have the ability, they don't have that lineage or that roadmap, if you would, of compliance regulations to servicing features. And that's really what makes the ability we have to deploy the AI in a way that is going to help us really speed up how we deliver change to this industry.
Michael Moeser (11:40):
That sounds really powerful. I mean, 8,879 pieces of requirement and there'll be more. I'm sure there are. And I can imagine as these regulatory bodies continue to issue new rulings and requirements and having your AI be able to look at those specific requirements and call out, these are the 32 that need some kind of attention and then having a human that direct them or direct the energy to solve those issues, that's very powerful. Now what are the non-negotiable building blocks to make compliance truly always on? Are we talking data architecture, workflow orchestration, credibility, or something else entirely?
Shane Leonard (12:28):
I think it's a combination of things. Just to be competitive with Matt, we have 3,296 data elements in the system at this moment. Typically in servicing, that's around 2,200. Connecting the dots directly between those data elements, the user guides, the features and those 8,000 plus requirements that Matt has looked at is what we've done. You can't do that without a lot of work and you can't do that without technology. So there's various technologies involved to pull that all together, but we've done it in a way that it's extremely manageable and part of it, you mentioned the word workflow, part of it is workflow. The ability to serve it up in a way that is usable so that a servicer along the way can actually see that as well so that now they can actually connect the dots between those 8,000 and the 3,000 and the user guides in a very concise manner.
(13:31):
There's no guesswork. And what we found is it's also made us more efficient internally on how to get to the answer because we're building a new system. We're working on Dara daily at a very fast pace while those things are changing. So in order for Matt's team to be able to get back and say, okay, something changed on us and Freddie and Fannie are making changes as we speak, there was no way for us to say, "Okay, here's what we're going to go build," and we just stop because what we were building is in effect, those goalposts are moving as we go. And then what we've done is we thoughtfully went through the process and said, "What tool do we need for what element in this and how do we serve up that information using those tools so that the human on the other end can act quickly, concisely, less mistakes, less issues, and we can get ahead of the curve constantly.
(14:30):
We're actually practicing in real time what our servicers are going to be doing with DARA in the future.
Michael Moeser (14:38):
Matt, any thoughts on non-negotiables?
Matt Tully (14:41):
Well, I'll give you a non-technical answer, Michael, which Shane did a really good job laying out the technical side of it. The non-negotiable from my perspective is industry relationships. We're not just relying on the AI to do the work for us. We talked to Fannie, Freddie, Jenny, FHA, VA, USDA, all the regulators. We're plugged into the Mortgage Bankers Association, the Housing Policy Council. You have to have a pulse on what's going on. You need to understand what the regulators are working on, what the investors are working on. So as the technology is doing its work behind the scenes, really the human in the loop is understanding what's the intent of where the investor or the regulator's trying to drive this. And that's what helps us then build the future functionality in the system. And from my perspective, that's a non-negotiable because you can't go to the destination that the investor or regulator wants without knowing what their intent is and what outcomes they're trying to drive.
(15:39):
And we really pride ourselves on our connectivity in the industry to understand what those things are and then also have feedback from our clients to say, what's your use case and what are the ways you need to use the system in order to be successful?
Michael Moeser (15:51):
So we've been talking quite a bit about AI and obviously any conversation we're having around this technology, there are a few issues here. Where's it actually delivering value today and where is it creating more risk than it solves?
Matt Tully (16:07):
Yeah, we need to be really careful about not letting AI solve the problem without thinking about what are we asking it? Where is it getting that information from? And then doing the analysis of that information. And as I alluded to a moment ago, we've deployed AI, but we've put guardrails around what it's looking at. We're not looking at the internet writ large when we talk about our compliance database, we have it just trained on the database itself. Now we are able to point it out at the internet when it comes to regulatory change, but we've pointed it in very specific directions at the CFPB, at the OCC, at Fannie Mae, Freddie Mac, et cetera, et cetera. And you have to be really careful. Obviously there's a lot of things that the regulators and the investors are putting out at any given time. So you still have that human in the loop to say, what is it?
(16:57):
If it's a selling guide update, for instance, well, that's not going to be relevant to mortgage servicing, so we can ignore that. But it doesn't relieve us of the responsibility to A, know what the update is and then B, really make sure we're doing the analysis when we think about these are these servicing changes. We have multiple products in the Dara Suite. What products is it impacting? It could be the core, it could be foreclosure, it could be claims. We have to challenge ourselves in terms of the analysis. Yes, it can allow us to do the analysis much more quickly and much more accurately, which is great, but constantly reminding ourselves that it doesn't relieve us of the responsibility to do our job.
Michael Moeser (17:35):
Any additional thoge?
Shane Leonard (17:39):
Several actually. So one of the things with AI, there's a term that's used out there, it's called hallucination. So I think the number one thing you have to do is understanding the tools that you have and employing them properly is non-negotiable to the last point. There's a country song that I happened like and there's a verse that says, I'm what I am and I'm what I'm not. Understanding that what I'm not part of this is the important part. And I think that if you go about it and think that AI is going to solve all your problems, then we're probably right back to where we were with blockchain is going to solve the world problems as well. And we know how that works. So they're good tools when employed properly and that's how we think about everything. Make sure we have a screw and we're hammering on it and wondering why it's not going in the board.
(18:41):
We're very cognizant about that. So we think through that ahead of time, experiment a lot with it without fear because this is the time to do it and then make sure that we determine at the end that we're all in agreement, this is the right place, the right time to do that and know ahead of time what can go wrong and then make sure that you put the framework in place to make sure that you know when it's going to happen. It's just like any process that you put together. It's people, process and systems and that is still true, which is why we serve this stuff up. Here's an application, it's connected to this, this data's connected to that, and here's the output. Somebody take a look at it and make sure that again, Matt's 8,000 can connect to my 3,300 and what's in between that is the processing that has to take place to make that work.
(19:39):
And I think we've done, I wouldn't say adequate, I think we've done a stellar job at making sure that that's true.
Michael Moeser (19:47):
Definitely see how the human in the loop there can help mitigate any of those hallucinations from really going very far. Now, but when you embed in compliance into workflows, you're also embedding risk decisions. How do you prevent automation from becoming a blind spot?
Shane Leonard (20:09):
A lot of that has to do with what I just suggested in a moment is understanding the tool, what it can and can't do. Making sure that if you're going to hit a button that removes PMI, that it has all the data points needed natively in the system to do it. We've gone very wide and deep on what we've included within the system and that's purposeful because the more information that whether it's AI or the human for that matter, and we've struggled with that within the industry of having systems all over the place, all trying to connect to each other, keep those things in sync sometimes in real time, most of the time overnight and that entire concept has changed. We're in real time or near real time. We've placed all the fields that you need to do your job in one place with processes.
(20:59):
By completing that puzzle, you have less and less risk as you go. We did some analysis on some of our internal systems that we use today with LoanServe and realized that the connection points are 65% of the actual workload in those systems, in our default systems, just keeping the two systems in sync. That means only a third of the time we spend building a solution is actually providing value on that solution. So now not only did we erase that, we just became more efficient because now we can focus our efforts where they should be and that's making sure that solution does what it's supposed to do. I think that's the biggest differentiator here on the environment that we've created for ourselves and are going to deploy to shortly share to the servicing industry. And I think that the world becomes your oyster as a servicer because those data points exist, because you have processes that make sense and that they're all in one place.
Michael Moeser (22:08):
If compliance becomes real-time and continuous, how does that change the dynamic with regulators? Does it actually improve trust with borrowers as well or just your own internal efficiency?
Matt Tully (22:20):
It's all of the above in my view. If we think back to the experience we had six years ago during the pandemic, if you recall, the CARES Act passed March 26th or 27th, somewhere in that timeframe saying you could go on forbearance effective April one, everyone flooded the gates and asked for forbearance and overnight investors had to update codes so that they could track the borrowers that were impacted by forbearance. There were downstream impacts around things like credit bureau reporting. And I remember the mentality amongst regulators was, this is easy. It's technology. You just do the updates and not recognizing that the legacy stacks, including our own, sometimes have limitations around how quickly they can do things. And there is then that trust that not only can we do it quickly, but we can do it accurately and we can serve it up to you.
(23:18):
I talked about earlier the lineage, the connection between compliance requirements to servicing features. We want to be able to serve that up away to our customers that when we're doing regulatory change and we're especially having to move in this dynamic environment, we're going to tell you in specified detail what features of the system we're updating in order to help you comply. No more digging through release notes and trying to figure it out after the fact, but serve that all upfront. And that helps them with their regulators when they come in and they do exams and try to understand who did what, when, where, and how. But more importantly, you talked about that borrower experience that there's not that hesitation. There's not that issue of what's going on in my loan. The information I'm seeing on the portal isn't accurate. The service representative can't answer my questions because something on their screen is not correct.
(24:11):
And the whole ecosystem benefits if we can do it more accurately and we can serve it up in a transparent way, which is what we're going to be doing with DARA.
Michael Moeser (24:20):
What actually breaks when servicers try to move from bolt-on compliance to embedded infrastructure? Where do most transformations stall?
Shane Leonard (24:30):
Pretty easy answer there is as systems were built over the years, they were never really built with an intention of connecting to each other. So things don't match perfectly. So you spend, I don't care whether it's just moving an MSR trade or actually moving and changing systems or building something internally, you end up with the same problem over and over again, which is we're trying to overcome the fact that the core doesn't do what it needs to do. So then we build all this stuff on the outside and then we expect it to work perfectly with something on the inside and that doesn't happen. They mismatch. They always mismatch no matter what our great intentions are. And then what you end up then is this disparate set of data
(25:17):
That you got to go untangle that mess and it happens constantly. I think servicers, as they look at it, as they look at any MSR trade, just moving loans from servicer to servicer, you see it without even changing the system, they're just simple problems like that. Our objective was to not allow that to happen, at least to the extent that we can control it internally. So again, wide and deep, making sure that things make sense, retiring some of the older things that we do. And going back to your prior question to Matt, the consumer ultimately benefits because now we could correct a mistake quickly in a real-time environment, that's one thing. But then also the accuracy of the data which causes those problems means less phone calls to the servicer, less phone calls from that borrower in that cycle that they get in and my escrow is messed up and I've experienced it with a large servicer myself recently and back to back years, that servicer made the exact same mistake because the data that transferred to them from the original servicer, the originator that did my loan on my house that I'm sitting in right now didn't give them the proper information and it went through two cycles for them to figure that out.
(26:43):
Now as a servicing guy, I sat back and I'll give them some forgiveness and some grace there, but I understood exactly what I was looking at when I saw it because those things are pervasive within the industry and all starts with that. And I believe very strongly that what we're doing is great for servicing. It's better for their customer base, much better. And having experienced it recently, I can only hope that my loan ends up on what we're building
Michael Moeser (27:16):
As we come to close for this session, Shane and Matt, any final thoughts?
Shane Leonard (27:21):
Oh, kick that one off. I think that we're in an adapt or die situation within servicing. The rest of the world is changing around us. Other financial industries are moving at pace and have made changes around this. We've mentioned AI a bunch of times, but it's not just AI. It's just the basic cloud tech stacks and working in real time and all of those things. So I think there's a moment, an inflection moment here that says, "Hey, are you going to be part of it or not? Are you not? " I would encourage servicers to think about that. If you're not doing anything, you are falling behind if you're not thinking about it and looking at this realistically. Again, AI is not the magic bullet. It's not a silver bullet. It's going to change everything, but it does effectively change the speed at which we can deliver solutions and you got to be part of it or you're not, and that's coming.
(28:18):
The future's here, whether we like it or not.
Matt Tully (28:21):
And I'll add to that quickly. We're really excited about what we've built. We're really excited about our ability to embed compliance, not just existing rules today, but we know the changes that are going to come because this industry isn't static. We know it evolves. We know regulators and investors are constantly working on things. And to Shane's point, it is adapt or die because it's not going to stop. We can wish it for it to be otherwise, but it's simply not the case. And you have to adapt into this paradigm by utilizing modern technology, by deploying AI responsibly to be able to deliver compliant outcomes for your regulators, for your investors, and ultimately for your consumers. And we're really excited to share it with you. So we encourage the audience to reach out. We're happy to do a demo of our Dara system and show you how we've embedded compliance and how we're making it real for our clients today.
Michael Moeser (29:15):
Super. Thank you, Matt and Shane for sharing your perspectives today and thank you to our audience for joining in. Goodbye.