- 23 hours ago
Matt Dowd of ICE Mortgage Technology joins HousingWire to break down mortgage tech innovation, product strategy and operational strategy.
Category
🗞
NewsTranscript
00:11Welcome, everyone. I'm here today with Matt Dowd, the VP of Product Management at ICE,
00:16and we are filming this live in person at ICE Experience 26 in beautiful Las Vegas. So,
00:23Matt, thank you so much for sitting down with me.
00:26Thank you for having me, Sarah. I appreciate it. Look forward to the conversation.
00:29Absolutely. Welcome to Vegas.
00:30Yes, it's been so much fun. Okay, so you guys unveiled some pretty big things at this conference,
00:35obviously, and chief among them is this AI agentic call thing for servicers. So, please,
00:42you know, take that mangled description I just gave and tell me how you think about that.
00:48Yeah, we named it something different than mangled AI for servicers, but
00:52ICE Aurora is really our launch, right? And there's a lot of different use cases,
00:57and I'll get in a second with the ones you specifically mentioned. But the reason Aurora,
01:02and to understand it is important for you and anybody listening, is Aurora is an AI framework
01:08that's built across origination through servicing. Okay.
01:11Right? So, what it does, it provides this AI backbone that provides consistency, speed, precision,
01:17and a continuity of compliance and governance, rather, across the organization. So,
01:22as we look at Aurora, what it really is, it's AI built into the system of record. So,
01:28in servicing, it's built into MSP, for example, or servicing digital, which are the applications
01:32that people already use. And I bring that up because, as you mentioned, like chat and voice,
01:39those are really just use cases of Aurora, this bigger AI picture. So, in the case of chat,
01:44and I'll speak specifically to the borrower, because that's where one of the big use cases come.
01:50I said, we're at chat, AI, right? Chat available for servicing digital. It's available 24-7. It can
01:56help the borrower by allowing them to answer questions anytime they want. So, we all know what
02:01the most frequent calls are, right? That come into a call. You and I are consumers. If I want to
02:07understand why my payment didn't hit, or maybe my payment went up and it's escrow related,
02:11and it's nine o'clock at night because I'm busy all day, I don't necessarily want to wait till
02:15tomorrow. But those are easy enough to just type in a chat or call and have a voice respond to
02:21it.
02:21So, you know, the ability to, for the borrower to self-serve and ask questions about their loan
02:27at any given time, obviously a big win for the borrower. And then on the servicer side,
02:32we're reducing friction, we're reducing call volumes, obviously reducing training times,
02:37and really the dependency on human bandwidth that they incur. So, those are some of the
02:43follower benefits as you look at. And think of chat and voice as kind of the same thing,
02:47right? You're either entering it on a keyboard or it translates to voice. So, those are some exciting
02:52ones that we're launching in beta right now. Those are. And I also think,
02:57you know, something about the servicer relationship is you don't really pay attention to your servicer
03:01until you need it, right? I mean, if it's on auto pay, how often are you talking to your servicer?
03:05Maybe never. But then when you have this question, it could be intimidating for borrowers to figure
03:10out like, who do I call? But also like, I just have this one question or, you know, if they're
03:14in
03:14distress, it can be difficult. Like, they might want to find out that information, not through a human
03:20voice. That's right. You know, you bring up such a good point because in servicing, when it's a
03:25performing loan, you probably just have your mortgage taken out and you're the perfect customer.
03:29So, you're not interacting with your servicer on a frequent basis. Like in origination, when you're
03:35getting your loan, you're calling your loan officer or whatever all the time. Different in servicing.
03:38But when you do need them, like you need it to be there. And it can be intimidating. And think
03:43about
03:43as people potentially, certainly with, you know, payment questions, but end of the year, taxes are a big
03:49thing. Like, you need your tax statements. Do I want to like, hang on hold for time to say,
03:53where do I find my tax statement? Hold on, go to the website or where's my tax statement? Type it
03:57in
03:58or call and everybody's like, here it is. And it just displays on your device. So, I agree. I think
04:03from a borrower perspective and a service perspective, just huge wins all around, right? From both sides of
04:09it. So, we're super excited about this. And again, I'll just touch on one other thing. The importance of
04:15it is because we're using the services data in this case, which is if you're the borrower,
04:21it's your actual data. We can be extremely specific, right? Because it's the data that
04:27lives in MSP. It's not just an LLM that lives outside of there that has random data. So,
04:32the questions we're able to answer are going to be accurate and precise to that individual. And yet,
04:38we build very specific guardrails around our AI. An example would be, if I'm a borrower and I'm asking
04:44a question like, should I refinance my home? And at what rate, for example? Well, you don't want to
04:51have AI answering those questions. So, the guardrails provide the off-ramp. We're intelligent
04:57enough to know that this is a question that should not be answered by artificial intelligence. And then
05:02we'll hand that off to that customer service agent so that they can actually answer the question
05:07you know, as they should. And that's a safeguard for the servicer to feel like, you know, this is
05:12completely, I can trust this and we'll get it right and somebody's not hallucinating to my clients.
05:18That's right. That's right. I hear, obviously, there's a lot of discussion around AI. Like,
05:23it's everywhere. And what I'm seeing a shift in, three, six months ago, there was a lot of discussion
05:31around, you know, what type of functionality do you have? Can you answer questions, right? Is it voice?
05:36Is it chat? Think of it like feature-based. There's been really a tantamount shift. Much of the
05:43discussions I'm having are around exactly what you mentioned, Sarah. It's around compliance. It's
05:47around governance. You know, is it auditable, right? Can you explain it? Like, that's where
05:53everybody's focused right now because that's where the potential pitfall comes in. Can you imagine a
05:58scenario where AI gave, like, bad escrow information or miscalculated a payment? And then you're in front of
06:05the court and you're like, our black box said it. So we are so guarded with, you know, those guardrails
06:12as we use and the governments around it because it is, in my opinion, the data and that governance
06:17around it, that's what's so key to success. To protect, right? We've been protecting services for
06:24decades, right? Keeping them out of headlines and we do not want to put them at any risk. So
06:29we are aggressively building AI, but on the flip side, we are very guarded and cautious about it.
06:36I think that's a really interesting point. Tell me about the value of having it embedded in the
06:40workflow. Yeah. So when you think about it, like having this governed AI framework that's embedded
06:48into your system of record and into the workflow, you no longer have to stitch it together or bolt it
06:54down. It's not something that's being, you know, incrementally added on. And by not doing so, you
06:59think about the controls you now have around it. So obviously role-based permissions are real
07:05important, right? What you can access, I might not be able to access. Well, you know, within MSP,
07:11whatever roles are assigned, those are assigned. It's one time. We have access to all the regulations
07:17and the rules that you've already implemented. So again, we're ingesting those to define the guardrails.
07:22You don't have to do it in two or three different places and then make sure that everybody,
07:27that everything's sinking. Like that can get very dangerous, in my opinion. And then as you tie it
07:34into workflow, I love this question because servicers can then look at their workflow and decide
07:40where should AI automate and where should a human decide? And that's our big, you know, our thesis is
07:47AI is always going to inform, but ultimately let the human decide, right? So it's reducing the work
07:52applied by creating exceptions or risk and saying, okay, Sarah, this is what we came up with. You
07:58should decide what's best to do. So you can stream, having it built into that system of record allows
08:03you to configure where you want to streamline your workflow, but still maintain that human control
08:08or human in the loop. So interesting. The fact that you've seen that sort of change in the last
08:13three to six months would seem to me that, I mean, you know, you guys are responding to a need
08:17that's out there. So it's not just people are like, tell me what this can do. But it's like,
08:21they're assuming like it can do that. Now tell me how, you know, it stays compliant. So I think
08:25that's kind of a big change. It's a very big change. I think, yeah, if you and I were to,
08:30you know, sit down at this, at the same event last year and talk about AI, it would have been
08:35a
08:35completely different discussion, right? People were, I think, extremely guard and still trying to figure
08:41things out. But now most companies you talk to, like you talk about AI, it's like, do you use
08:47Microsoft? Sure. Well, you know what Copilot is, you use it in PowerPoint and Word and email and
08:52everything else. So there's this comfort level now, right, where people are comfortable with the
08:55technology to a degree. And that's how we're positioning Aurora. And that's what, as we talk
09:01about building into the system of record, think about it, wherever you are in the servicing life
09:06cycle, whether you're an MSP, whether you're doing customer service, whether you're doing loss
09:11mitigation, default services, whatever it may be. That's where Aurora can present himself and help
09:17solve that, that, that workflow problem. But the guardrails come in is like allowing that
09:24servicer to decide, do I want to incorporate AI in here? And then how much of it? So yeah.
09:30And then making it easy for them to make that decision, right?
09:32When we just keep our little Aurora icon floating there. So just like everybody's used to, it's
09:37like, there it is. You know, they will learn that, oh, that's, that's my tool tip. That's
09:41my help. I have a question. Okay. And, you know, for us in Aurora, Aurora is really based
09:48on rules, right? It's not based on instruction. So it's going to be a lot of input, output, and
09:53then it's going to be, you know, simplifying steps in the process. So people will just start
09:59seeing, oh, whatever screen or page that they're on, there's Aurora icon, and that's the use case.
10:04So a little bit different than, and this has been the big shift that we've seen from six
10:09months ago or 12 months ago, where people were looking at it as like, what's the solution going
10:14to do? Is it chat? Like, is it voice? Is it accessing operational metrics? All those things we do,
10:22but we're just trying to say, just look at it. Is this omnipresent solution? And then the workflows
10:27can kind of help themselves. Does that make sense? Yeah, no, it does. When you have something
10:32like this, technology is so great unless people don't adopt it. Yeah. Yeah. So that's the,
10:37that's the big one. That's the big one. So as you're thinking about this, as you're developing
10:40this product, how do you, you know, in this specific case, how do you make sure that this
10:45becomes indispensable to people? Yeah. Great. Really great question. We put a ton of thought
10:49around this. Okay. Um, so you mentioned like chat and voice. So we're, we're educating people
10:55start here because it's not extremely high risk. Consumers and users are generally used
11:02to that, right? It's not that much of a, of a stretch for them. Borrowers of course are
11:06all familiar with it. Uh, in customer service, it's easy for you to explain to your customer
11:12service agent, Hey, when Matt Dow calls in and you go to your screen, there's going to be
11:16three predictive elements that say why he's calling with a conference score, probably
11:21payment, right? The next might be something different. So I click on payment because that's
11:25what's mostly what I'm calling for based on the history. We have a note summarization that
11:30comes up, right? That says, okay, well, Matt called in, here's the times he's called in
11:34over the last 30, 60, 90, 120 days. This is what you talked about. So like as a user, oh,
11:41great. Uh, so when someone calls in, I'm going to hit this button. Okay. If they're calling
11:45in on a payment, here's everything we talked about. So, you know, a very, very transparent
11:50way to do it before you get to the, you know, real complex where, Hey, we're going to let,
11:55um, autonomous agents take over multi-step workflows. That'll take a little bit more change
12:00management because it's going to be asking people change your behavior. So we're starting
12:05up front. Another example would be with our ice BI that we've incorporated in that, that,
12:10that helps with automation. And you know, one of my favorite use cases is around FEMA disaster.
12:17Okay. Right. And I'll tell you why, uh, just as a human, right? Imagine the news is coming
12:23in. There's going to be a big hurricane or whatever it may be, or it actually happens tragedy, right?
12:27That's like a horrible with ice BI. You're able to immediately scan your portfolio, identify all
12:34those borrowers that are in that area and then reach out to them. Right. I know if I lost my
12:38home,
12:39like the first thing I wouldn't be thinking of is I better call my servicer. Absolutely.
12:43Boy, how nice is it if they reached out to you, you know, obviously with whatever letter and informed
12:47you of the next step. So when things did settle down, you knew exactly what to do. And that's really
12:53the power of our ecosystem, right? So it's, it's not just with Aurora and AI being built in, but it's
12:59also with the other tools we have bringing in the MSP so you can create this automated process.
13:05Well, and for me, just as a human, again, like this is the path I'm looking for with technology.
13:10It's like, I, if I, if you're my servicer and I call in or whatever, I want you to be
13:14up to speed
13:14on everything about me. You should know all those things because you're my servicer and not just my
13:19service. I mean, when I call in on car insurance, when I call in on anything, it's like, if I'm
13:23your
13:23customer and I've given you all this information, please be up to speed. So I don't have to go through
13:28it. And so to me, that's a huge borrower benefit. If that guy's like, Hey, last time we talked,
13:32this, what you said, here's the steps. And you know, you're not going through the red tape. I
13:36think that's a huge benefit. It's huge. And I'm like you, like I'm your customer, know me. And
13:42as I touched on earlier, having this type of technology, and we think of AI is just another
13:47tool we've been automating for decades, right? Whether it's workflow configuration, like, all
13:52right, here's another tool. And maybe six years from now, or two months from now, there's some other
13:57technology that everybody's all excited about. We will continue to bring these tools into our
14:01application. But because we have the data that's about you, Sarah, right? Like, that's what makes
14:09it unique. So when you call in, it can be down to your payment, like down to the penny, this
14:13is your
14:14payment. And that's when you when you talk about AI, you still want it to be, you know, human at
14:21the end
14:21of the day. So if it's just finger in the air and guessing, you're still going to have the same,
14:26you know,
14:26the same experience as a customer, like, why don't you have all this in front of me? And that's what
14:31makes Aurora unique. Because we can leverage it off the, you know, regular LLMs, which we do, but then
14:37we layer on top that unique, you know, the unique customer information so that their employees can
14:45spend time, you know, what they should be doing, which is making judgment calls, which is showing
14:49empathy, right, and driving outcomes. Like, if all your people are doing that, then I think
14:54everybody's happier, right? Because that that employees like, finally, I'm not doing these,
14:59you know, redundant, routine tasks that don't require a lot of skill, like, let Aurora handle
15:04that noise. And let me as a human focus on empathy, focus on judgment, I want to do, right? Well,
15:10and
15:10if the more you make that relationship good between the servicer, the lender and the borrower,
15:15the more you're gonna have adoption, because that's what everybody wants. We're all trying to be like,
15:19let us let us get that relationship better. Yeah, it's not for us, we look at it,
15:23we are not necessarily saying, let's replace humans, or let's take over the thinking, right?
15:30What we want to do is we want to elevate humans to do the work that they can only do
15:34your employees.
15:35And I think everybody would be more happy. So is, you know, change management going to be regardless
15:40of the industry, this is going to be the hardest thing, because people's natural reaction is like,
15:44oh, it's going to take my job. Right. But if you show them that no, it's not at all,
15:48it's removing all the things like, look, if there's an AI agent to empty my dishwasher,
15:53I'd be like, this is fantastic, right? Sign me up.
15:56My wife wouldn't look at me and be like, you're not emptying the dishwasher. Anyway,
15:59probably a terrible example. But we want people to focus on the things that they get excited about
16:04and enthused about and then they can do. So that's what this is about, right? We're looking at this.
16:10I think when I talk to people at AI, it's very easy to get out into the future 12 months,
16:1618 months,
16:17but we try and break it down and say, okay, that's great. But can we talk about four months from
16:22now?
16:23Because like my happiest day is when AI isn't all this hype and headlines when it's like, yeah,
16:27I just use it. I just use it, right? And it just makes my life easier. And I'm comfortable with
16:32it.
16:32I engage when I need to engage. Isn't that what we all want? Get rid of the mundane daily stuff
16:37so you can focus on what's really interesting or why you started in that profession anyway.
16:42I love that perspective. Matt, thank you so much for sitting down with me.
16:45Thank you so much. I appreciate it. Enjoyed the conversation.
16:48Yeah, yeah.
Comments