On today’s sponsored episode, Editor in Chief Sarah Wheeler talks with Stephanie Durflinger, Chief Product Officer at Dark Matter Technologies, about making smart decisions with AI, and how lenders are implementing this technology to get real ROI.
Related to this episode:
Dark Matter Technologies
https://dmatter.com/?utm_source=housingwire&utm_campaign=podcast&utm_medium=paid
Enjoy the episode!
The HousingWire Daily podcast brings the full picture of the most compelling stories in the housing market reported across HousingWire. Each morning, listen to editor in chief Sarah Wheeler talk to leading industry voices and get a deeper look behind the scenes of the top mortgage and real estate stories. Hosted and produced by the HousingWire Content Studio.
Related to this episode:
Dark Matter Technologies
https://dmatter.com/?utm_source=housingwire&utm_campaign=podcast&utm_medium=paid
Enjoy the episode!
The HousingWire Daily podcast brings the full picture of the most compelling stories in the housing market reported across HousingWire. Each morning, listen to editor in chief Sarah Wheeler talk to leading industry voices and get a deeper look behind the scenes of the top mortgage and real estate stories. Hosted and produced by the HousingWire Content Studio.
Category
🗞
NewsTranscript
00:00Welcome, everyone. My guest today for this sponsored episode is Stephanie Derflinger,
00:11Chief Product Officer at Dark Matter Technologies, to talk about making smart decisions with AI
00:17and how lenders can implement this technology, which is a little different than others.
00:21Stephanie, welcome to the podcast.
00:24Thanks, Sarah.
00:25Great to have you on here. I'm excited about this conversation.
00:29Let's talk about the implementation of new technology, specifically AI.
00:34What are the biggest obstacles you see that lenders need to avoid as they try to innovate?
00:41With AI on implementation, it's of particular importance that people are involved,
00:48all the people are involved. AI is scary for the workers, the worker bees.
00:55If you think about manufacturing alone, it is just that, it's manufacturing.
01:00You've got these different personas that are involved at different stages of the loan.
01:05People perceive that AI is going to take their job.
01:09That can be scary. That's one. Change is scary. This is number two.
01:15People, they've been doing something the same way for a very long time.
01:20So a lot of AI implementation is resistance, resistance brought, I should say.
01:30When you're implementing AI changes in particular, but all changes really, but
01:34AI changes in particular, it's important to have the three legs of the stool involved.
01:42One, the technologists obviously need to be involved.
01:47They're going to implement the technology. That's the easy one.
01:49Two, the business owners or the executive sponsor. It's important to have strong
01:55executive sponsorship. When I say executive sponsorship, I mean it at a couple of levels.
02:00I do mean it at the executive level because someone has got to be pushing from top down.
02:04Someone has got to insist because there is going to be resistance and there is going to be friction
02:09and there is going to be hearts and minds problems. The hearts and minds problems need to know
02:15that there is someone driving this and someone who is passionate about it.
02:19And then the second layer of the heart of the executive sponsorship should be the business
02:25line leaders. So you need that. And then the third leg of that stool are the people themselves that
02:31are going to be working with the technology. And one of the tricks that I like to do is if you have
02:40resistance and if you have a rather vocal person who's resistant at that line level,
02:47I usually recommend that you bring them on the implementation team and you sell them on what it
02:53is and you make them an advocate instead of a naysayer. And if you can do that, and I'm saying
03:00this as someone who is a technology person implementing one of our technologies, I would
03:05tell my customer, I highly suggest you bring Tammy on this team. Let's get her to love it.
03:12If Tammy loves it, then she's going to amplify that message out. So my best rate of success on
03:19implementation is taking the problems head on. The fear of job loss, address that head on.
03:27The expectation isn't that we're going to replace people. The expectation is that we're going to be
03:31ready to scale. Make sure those messages are delivered. Whatever the intention is,
03:38make sure the message is delivered consistently more than once. Bring on the folks that
03:45may have the highest level of resistance onto the team. Make sure you have the triad, the
03:51line level folks, the business executive sponsorship, and the system administrators
03:57or the technologists all on board to implement. I think if you do that, you're setting yourself up
04:05for good success. The second part of it is follow-up. The second part, you have to follow
04:12up. If you don't follow up, 30 days, 60 days, 90 days, six months. If you're not following up,
04:18then you have a chance of regression into old habits.
04:24It's interesting that when we're talking about technology, one of the biggest challenges
04:28is still people. It's the people part, right? And I feel like that's how did people come
04:34at the problem in the first place. Were they like, here's the problem we need to solve,
04:38here's the technology, or did they loop in the people who are part of that solution using the
04:43technology? Well, it's definitely hearts and minds. That's what I tell our customers. It's
04:48a hearts and minds thing. At the beginning, sometimes there's shiny things, and people
04:58get really interested in the newest technology. There's a lot that's technology for the sake of
05:04technology, and I think sometimes people can get wrapped up in that. That's the least successful
05:11reason to implement a technology. Understanding a problem that needs to be solved,
05:17that problem being identified either by a business value that is not being realized,
05:23or a friction point that's been identified by the people that are working with the manufacturing
05:30line, those are the things that will drive success with the technological solution,
05:36and that is absolutely people-driven. And so it's best, absolutely best, if it comes from
05:45the folks who are actually working with the whole process. If they are the initiators,
05:52you definitely have a better chance of success with it, and that's definitely around identifying
05:57problems and then chasing the technology that can help with the problem, not finding a technology
06:07and then trying to figure out what problem it solves. I love that. I mean, that seems obvious,
06:13but I have talked to a lot of technology leaders over the last couple years, and I know that it
06:18may seem obvious, but it doesn't mean that's the way it always goes. It isn't the way it always
06:23goes. I'll tell you that years ago, I was trying to get folks to move to the cloud, to be comfortable
06:32not hosting themselves and move to the cloud. And it was an uphill battle. CTOs didn't want to do
06:37that at that time. And then a Super Bowl commercial changed the landscape for mortgage technology,
06:46and everyone after that Super Bowl was interested in technology. And all of a sudden, my job became
06:52much easier trying to get folks to transition to the cloud. But my point in saying that is that
06:58it wasn't because there was a problem they were trying to solve. It's because there was a cool
07:02commercial that got folks interested in it. That happened to work out for my benefit. But it's
07:07kind of human nature where we see something that looks very innovative, and we want to find a way
07:14to take advantage of it. And it's my job, and it's the job of Dark Matter to look at those innovations
07:21that are making themselves available and innovate ourselves with the folks that we have on our team
07:28and translate those innovations into meaningful solutions to real problems that lenders are facing
07:34today. I love that. Let's talk about technology costs, because this is something very interesting
07:40to me. I got into the industry and covering the industry in 2013. And so much of what we reported
07:46on and what the industry was trying to do was really around compliance, right? Like three years
07:51out from that, Dodd-Frank really had a lot of things that people had to do. But it feels like
07:56now as an industry, the cost of compliance that was rising for so long is sort of replaced by the
08:04increase in the cost of technology, even with AI. And so why do you think, why are we failing to
08:10lower the origination costs when we're making this added investment?
08:15I think there's a few different levels. So, I want to say what pre-Dodd-Frank, we were somewhere
08:21around $3,000 a loan cost to originate, and then it shot up to around nine, and we're somewhere
08:27around 12, I think, right now. And that shoot up to nine had a lot to do with the compliance
08:32changes and the compliance cost. And now the shoot up to 12 has a lot to do with technology.
08:38I think we've leveled off on the compliance. I don't think we see huge continuing increases
08:43from compliance, and certainly there are increases in technology. It's twofold. One is that we're at
08:50the inception point of a lot of interesting technology, and to apply artificial intelligence
08:58into the technology space is expensive. It's not cheap. To realize the value takes a significant
09:07investment. And it's not just the technology investment, it's also the cost of change.
09:11So, there's more than the technology itself. There's a process change. There's the hearts
09:17and minds change. It's got tendrils in it. And the cost of technology by itself, the innovation
09:25of the technology and the actual hardware of the technology, the data science IP, the people
09:33that are creating the artificial, the AI solutions, the algorithms and the models,
09:41and the modeling, they're not inexpensive resources. They're talented people. So,
09:46all of that is a high price tag. But over time, you start to realize the benefit of that,
09:53and then that ends up helping to reduce the cost. That's one. That's one part of it.
09:58The other part of it is, I don't know that you'll start to see the benefits,
10:03you know, realize the cost reducing in your per loan. But I also think that it's a shift
10:12in the amount that you're paying for the different parts of the cost to originate.
10:18And what I mean by that now is that where your bulk of that $12,000 may be in the cost of human
10:28capital, that may shift over to the cost of the technology. You know, not to an equal part,
10:34you'll have some savings, but a percentage of it is going to be higher on the technology space as
10:39you start to implement those technologies. And you see a reduction on perhaps, you know,
10:44the people that are necessary to produce the results. I think that's twofold.
10:51No, it is twofold. Okay, so we've talked about the cost of technology. But what you touched on
10:55and I want to go deeper is like the value of the technology. And we'd love to know from you,
11:00like, where have you seen material benefit, like specifically improvement in metrics
11:06from your clients who are the most successful in their adoption of technology?
11:11Well, if we talk about AI specifically, it's in the workflow, it's in the manufacturing process,
11:16and it's the workflow automation. So there's a few things that have come into play,
11:21there has been for a little bit, you know, business rules, you know, and there's some
11:27automation where you can automate like service ordering, that's nothing new. That's been around
11:32for, you know, a fairly decent amount of time, and folks have been taking advantage of that.
11:36The things that are a little bit newer, where we see our customers taking advantage and applying
11:42some of the machine learning and the automated learning has more to do with this, I call it the
11:49so what, you know, you have extracted data from documents, you have other incoming data sources,
11:56you have a rule set, you also, you know, are able to automate ordering, and you have all this
12:02information coming in. And you have, you know, dynamic points in time, and things are changing.
12:09And so you can then as data points change, move that loan into different states, order new services.
12:19condition loans, satisfy conditions of loans, recondition loans, and only surface things to a
12:25person that are outside of a standard parameter. And that that is where folks are seeing the
12:32efficiency gains when a human doesn't need to touch something. Because, you know, we understand
12:37what to do with it. And beyond a rule set, we understand what to do with it. We can move it
12:44forward, we can perform all the automated services with that, and only have it presented to a person
12:50when a decision needs to be made. That's where they're seeing the lift.
12:54It feels like that sort of ROI might be harder to calculate, just because it's not just like,
13:00we plug this in, this is what happened. But it I mean, when you're talking about workflow,
13:04there's a lot of things involved there. So how do you measure that?
13:08So ROI is interesting. There are, there are calculators and most companies out there, you know,
13:13us included, would be able to say to you, on an average basis, you know, this is a standard,
13:19and you implement this, and we see a reduction in, you know, time or cost on an average.
13:25But that isn't enough. I suggest that that is not enough for a lender to really measure ROI.
13:32For a lender to measure their ROI, and I think this, this is a, this is something they should
13:37take away for anything that they're doing, but especially for evaluating a technology.
13:42There are so many technologies available. There's, there's so much opportunity.
13:48Things are just coming into the market quickly. So what should you do, right?
13:53What's important to your business? First of all, you know, what value, what business values are
13:59you trying to drive? And what are you, what are you trying to deliver? In the second place,
14:04where do you see the friction points? Is there, what's either, what's missing in, in delivering
14:11that value? What can enhance getting that value delivered, you know, in a more efficient way?
14:17Or what friction points are there? What's going to move the needle for you? You have to know that.
14:24And once you know what's going to move the needle for you, then you can start analyzing how you can
14:31get that accomplished. And understanding even the intangibles, you know, it's saying that people will
14:40be better, they're happier with their jobs, or a borrower is going to be more satisfied, or,
14:46you know, some of that can be difficult, you know, to, to qualify. What does that mean to my business?
14:52But if you can assign some kind of empirical evaluation, even if it's relative,
14:56this is more important than that. And then you, and then you create an equation that says,
15:03you know, this is how much I value these, these different metrics, then you have something that
15:08you can work with. And that has to be applied to your calculation as to whether you pick
15:14this technology to pursue, or technology B, you know, which is going to give you the lift,
15:21you can't boil the ocean, you can't implement them all at one time. So to pick to determine
15:28that ROI, you have to have a good understanding of what moves the needle for your business.
15:35And sadly, you know, we see oftentimes that people haven't done that, that they are just
15:42kind of going with the marketing evaluation, which is generic, it's generic, it's, you know,
15:50we don't, people don't know the business, they know generically, how things are. So you got to
15:56understand that for your business. And so that that is one thing I'd recommend, because you don't want
16:00to invest in technology and then not have it succeed on implementation. And a good ROI is one
16:06reason it doesn't succeed. And I'm going to lead into another reason it doesn't succeed, which goes
16:11back to that hearts and minds, which is adoption of the technology. So you know, one leads to the
16:17other, the ROI for your business is one thing. And then there's the ROI for the individuals who
16:23are using the technology. And that, you know, that's, that's the hearts and minds, you may have
16:29it implemented, it may be something of value for the business. If your team hasn't adopted the
16:36mentality, that's a value for the business. And there's a little friction in or change, because
16:42change is hard for folks. If they haven't, you know, embraced that change, the necessity of the
16:50change, they may at first start to work through it, they'll fall back, you know, at the first part of
16:56resistance, they'll probably fall back, it's human nature into their old patterns. And if there isn't,
17:03you know, an executive sponsor, if there aren't folks that are taking a look and, and constantly
17:10asking the questions about how they're, you know, using the technology, is it is it helping and
17:16working through that, you're going to find yourself with the technology implemented that is not
17:21delivering on the promise that you expected.
17:25When I think especially in a workflow situation, like, if something's too hard, or you, you know,
17:31if you haven't totally bought in, it's just easier to do it the way you've always done it, right.
17:36And that's where you get all those holes in the system where it's like, why isn't this performing?
17:40And you realize that people are doing workarounds.
17:44They do. And, you know, in our organization, we have a lot of products that you can, you know,
17:51add into, you know, we have a loan origination software, but we have other products that you can
17:55supplement in that system. And one of the criteria that we measure as a product organization is how
18:02well is it adopted? It isn't enough that it's been sold. It's how is it being utilized? Is it
18:10two things? Is it adopted? Meaning are the people that have purchased this actually using it in
18:15production? That's one. And then secondly, are they using it as as it was intended to be used,
18:20you know, or, or not, you know, so, so we understand, are they getting the full value out of
18:27what they've, you know, contracted or purchased from us? And, you know, can we educate them or
18:32adjust or, you know, how can we help them realize that full value? Very important that customers
18:37get their full value and that, you know, that they're not just implementing technology and,
18:42and it's all for not, because that's also increases that cost of technology. When you
18:47talk about it in the cost of origination, all these technologies people implement and don't use
18:52another factor. The most expensive technology is the one you don't use, right? Yeah, that's right.
18:57That's right. Exactly. Well, let's talk about AI and then, you know, gen AI, what's out there
19:03that's meaningfully improving the PNL of mortgage lenders? Like what's really, really making a
19:09difference? Well, the data extraction is doing nice things for folks, right? That's taking out
19:14a lot of the stare and compare and, and the, you know, allowing the data to be digitized,
19:21digitize, you know, digitizing what's on those documents and moving it through. So that's,
19:25that's a big deal. I think that it's the most meaningful thing, right? At the moment that
19:31people are actually using, you know, there's a lot that's going on that, that is on the cusp
19:37of being used, you know, mortgage lenders or the finance industry in general, we're not bleeding
19:43edge, you know, we're, we're dealing with, you know, a very regulated industry. And so things
19:48aren't exactly bleeding edge in terms of the technology. However, the predictive models are
19:55certainly, you know, starting to show themselves helping borrowers, for example, as they are
20:04entering an application, you know, pre-populating things and pre-filling and, you know, creating the
20:10letters of explanation on the fly, you know, while they're entering information, all of that is
20:16certainly, you know, things that are available today and things that move the needle for lenders.
20:22So they're able to get applications in faster, they're able to get to the approvals faster,
20:27they're able to automate the disclosures, you know, all those types of activities
20:30are definitely there. We're getting to a state where, you know, recommendations can be made on,
20:36you know, programs and, and pricing. And, you know, we're getting, we're getting to a place
20:42where those things are available, not yet widely adopted, I would say, but certainly,
20:48certainly available. I think that we're just short of a place where we would, you know, go to
20:56recommend things like, you know, an underwriting approval or things like that. And we're not,
21:02we're not at that level. We might, we'll point out when it fits a program, doesn't fit a program,
21:09but not, not yet qualify a borrower, for example. Right, exactly. That's, that's the one everyone's
21:15looking at. When we think about adoption of AI, I feel like, you know, it's always the biggest
21:20question for lenders is like, when do I pull the trigger and jump in on something? You don't want
21:24to be too early, you don't want to be too late, right? When it comes to AI, is there a different
21:30calculus with this than there has been in the past as far as like, when lenders should make that
21:35decision? When lenders should make the decision for adopting the AI? Yeah, that's definitely,
21:44that's definitely a lender call. You know, I don't know that I could give advice on that,
21:50except to say that we are in a highly regulated industry. And I, I individually would want to rely
21:59on proven, proven technologists, you know, trusted technologists when, when we are experimenting.
22:10I don't know if experimenting is the right word, but, you know, we wouldn't want to go with things
22:14that are black boxed. Let me put it that way. We should be able to evidence what it is that
22:20is, is being output. And, and that's, I think, for our industry going to be necessary.
22:29You know, at least in the short term, and there's a long road ahead. AI and the opportunities there
22:36are very interesting. And the way it changes, the way it has the potential to change
22:44our legal system, finance, you know, how we do everything is very interesting. You know,
22:52there's, I've attended a few seminars where there's talk about the legality of agents, and, and when
23:02they behave on behalf of a human, who's legally responsible, and, you know, an agent is like a bot,
23:12you could equate that to a bot. And so, you know, there's, there's talk of these things
23:18already, even though this isn't in the market. And it's not applicable to finance at this point.
23:25But, but this is, this is the direction, you know, this is the direction of the technology,
23:29and it will make its way at some point, probably a decade from now, into, you know, into our world.
23:38It's, you know, it's, it's not there yet, but it, but it will get there. And as finance,
23:45and as mortgage industry, I think that, you know, we'll be much slower to adopt
23:50those type of things, and we should be. I appreciate the caution, you know,
23:55the caution, even though the excitement tempered with a caution there. What should lenders be doing
24:00today, right? What, what can they do now to ensure they keep up with this as it's moving so fast?
24:08I think that the best thing to do is understand opportunity. So look to the industry,
24:15technologists to understand what is available, keep in touch with your own business needs,
24:23and understanding what the problems are that you want to solve. I think that's, you know,
24:28the critically important thing. And, you know, be curious, be curious, and look for technology
24:35solutions to solve your problems. And as a hobby, look at the neat, shiny things, but,
24:44you know, solve problems, solve real world problems. Another area of focus should definitely
24:49be on data. Data is going to feed the future. It's already feeding the future. The future is here now.
24:54So making sure that your data is of quality, bad data in is going to equate to, you know,
25:01bad results down the line. So having a data governance plan in place now, and looking at
25:08your source data is, is a step you can definitely take to ensure whatever opportunity in the future
25:15presents itself, you'll be ready to take advantage. So two, two things,
25:20two things, you know, when I've been doing these interviews for, for years,
25:25just on different topics of technology, data always comes up data has been coming up for like
25:3010 years, you know, but yet I feel like there's still so much there. And we've been creating data
25:36all this time, right? lenders have been, you know, processing data, there's still this data.
25:40So it's just like the never ending data thing. It's it is the future. I mean, it's not it's not
25:47even the future. It's right now. And it's going to be connected at some point, it'll be connected
25:53from the primary through the secondary markets in a meaningful way. And having the data, you know,
25:59the quality data is going to reduce so much, you know, error, and fraud, and it'll, you know,
26:09add so much value into the system when we have quality data. I think it's, I think it's fantastic.
26:17And I look forward to that day, actually, it's going to give so much opportunity for us.
26:23I love that. Okay, looking at all of that, what do you think lenders are struggling with,
26:28as they look to adopt AI? What is the biggest struggle?
26:32Understanding? Okay, it's, it's nebulous. It's, it's not easy to understand. It's not easy to
26:38wrap your arms around. I mean, I just talked about agents, and bots and legality of a machine making
26:46a decision for a person. That's, it's not easy. It's not an easy thing to wrap your mind around.
26:52So I think that's probably the most difficult. The other is that it's expensive. You know,
26:56right now, it's, it's not cheap. So you get two things at play. It's not easy to understand. And
27:02if you want to understand it, it's gonna cost. It's a, it's a huge investment. And on top of that,
27:10it isn't, it isn't black and white. It isn't one plus one is two. And it's, and it's, it's,
27:19it predicts things. It's not definitive. You know, so that's hard. That's harder, especially
27:28for people in our business, because, you know, we're used to numbers, and we're used to things
27:34literally adding up. And when you have something that might give you something different when you
27:42ask it twice, and it's really predicting, that's hard to, it's hard to wrap your mind around.
27:51I didn't even think about that. Just that, you know, I mean, obviously, I know,
27:54we have to have things that up, right? Lenders have to have things that up. But just the,
27:58the, the nuance there with AI is it introduces a whole nother thing.
28:05It's a different way of thinking. It's a different way of thinking. And just even,
28:10you know, even in a technology company, you know, my background is mortgage lending.
28:17I have data scientists on my team, and leading a data scientist team,
28:22in the end, getting aligned with their mentality, you know, and their,
28:30the way they think is, is different, you know, and so I see it myself in bringing these things
28:39forward. And so, you know, I'm experiencing that firsthand. And so I'm kind of, you know,
28:44projecting the difficulties I have, because of my mindset, you know, would be similar to those
28:50that a mortgage banker will also have, when they are, you know, wrapping their head around,
28:55you know, the use of AI in their businesses. I think that is a very safe assumption.
29:03So what's next for dark matter technologies? What do you, what are you excited about?
29:08I am, believe it or not, excited about the AI. We are making use of the, of the AI in our
29:17ecosystem. And it's, we're proliferating it. So we are excited about that. It's not, it's not the
29:25predictive piece of it. It is the workflow automation. So the workflow automation is
29:29everywhere in our system right now. And so it's completely exception based. And we're pretty
29:34excited to bring that into the marketplace. And it is bringing, you know, we used to call it
29:40bringing, bringing the loan as far left as possible. And we're pretty far left on bringing
29:45that loan. And so that is exciting to me. We also just just are bringing out to market a servicing
29:52system. And so we'll be able to bring together the primary and the secondary market. So now
29:57bringing the product and into the secondary market also and tying it together. We're really
30:04excited about that as well. That is very exciting. We're excited to be along this journey with you
30:10and talking to you and learning more about this. So Stephanie, thank you so much for being on with
30:14me today. Thanks, Sarah. Appreciate the time.