- 5 months ago
On today’s sponsored episode, Editor in Chief Sarah Wheeler talks with Raj Nair, CEO of Indecomm Global Services, about how lenders are deploying gen AI while still keeping important guardrails in place.
Related to this episode:
Indecomm Global Services
https://indecomm.com/
HousingWire | YouTube
https://www.youtube.com/channel/UCXDD_3y3LvU60vac7eki-6Q
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:
Indecomm Global Services
https://indecomm.com/
HousingWire | YouTube
https://www.youtube.com/channel/UCXDD_3y3LvU60vac7eki-6Q
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 is Raj Nair, CEO of Indicom Global Services, to talk about
00:12Gen.ai and the exciting ways Indicom is helping lenders deploy it, as well as the guardrails
00:18they're making sure to implement. Raj, welcome to the podcast.
00:22Thank you, Sarah. Thank you for having me here. Excited. Looking forward to it.
00:26Thank you. Thank you for being here. So I have to say something about Indicom really, really
00:30fast. I've been on the committee that looks at our Tech 100 and picks those award winners
00:38every year for since we started it, actually. I think I was here when they started it. You
00:44guys have been on that so many times. I think this is your fifth year in a row, fourth year
00:48in a row, something like that, but year after year. So just a shout out to that. That's a
00:53really incredible thing to get on that list year after year. So congratulations on that.
01:00Thank you. And I mean, it's always an honor to be on that list. But, you know, I think
01:05five years, I think, and that's probably in a row. But that actually doesn't tell the whole
01:12story. It's over 20 years. I mean, we're probably in business for 23 years, but like 18 years
01:19ago is really when we started a tech journey. So it took us a while to get maturity. And
01:24that's part of our story. It's just we don't rush things. We get it done right. So I guess
01:29it took us 13 years to make that list.
01:33Well, it's impressive. And let's dig into the tech part a little bit. So, you know, AI,
01:38especially generative AI right now, of course, it's the big buzzword, has been for the last
01:42two years, in the mortgage industry, especially, you know, really seeing a lot of people want
01:48to make the best use of this. What are some of the most common misconceptions you're seeing
01:53from lenders or executives when it comes to Gen AI?
01:57Oh, well, I mean, first of all, I think just trying and understanding the real, you know,
02:04real potential of Gen AI, that is usually the first challenge.
02:09executives are bombarded with this so much noise, right? I mean, you switch on, I mean,
02:17you can't escape that, right? And then plus also their own lived experience. So many of
02:22them have made this foray into ChatGPT. We all use ChatGPT or something or the other, right?
02:28So we have, you know, so we've tried giving it instructions, it produces results, it looks
02:34great. And it is great. I mean, so I'm not here to, I mean, I think the key thing is to understand
02:40that Gen AI and that whole technology framework is for real. I mean, it is, and it is evolving so
02:49rapidly. That is another challenge is that how do you keep up, right? I mean, how do you make sure
02:54that your investments that you make don't get obsolete even before you start adopting them,
02:59because something new comes in? So that is, that is kind of one of the things that, and that's
03:05probably an adoption level, right? But from a, from misconceptions perspective, it's this thing
03:11that, that, that Gen AI can handle the complexity of the mortgage process, the mortgage documents,
03:19mortgage data, because ultimately it all, it comes down to the data that's resident in different
03:25systems or in the documents to be able to bring all of that together and then provide the, you
03:34know, provide the outputs that are needed to drive the mortgage process. And, you know, we, we work
03:39a lot on the, on the, on that middle office, on the production of the mortgage widget, essentially,
03:45right? So, and there, and we know from the MBA numbers, you know, it's $30,000 to produce a
03:51mortgage. So, about $7,000 of it is to just produce the widget. So, so in there, I think
03:57everyone's looking at how can we get that done faster, better? How can we make that experience
04:02for the customer better? I think that's, that's what drives the drive, drives the automation
04:06journey. So now Gen AI comes into the mix, right? And then the answers, like in the, the, the impression
04:12that there is that that's given out is, oh, we can just push documents through it. It'll just
04:17produce the whole mortgage. We don't need humans in the loop. We just, you know, we
04:21can, you know, it just comes down and there are some really fantastical exaggerations that
04:26are made along those lines. And I think that's really where we have to, you know, we have to
04:32tackle that with, with, especially at an executive level that what you try with your, say, you
04:38upload your own utility bill and you get, you write prompts against it and the outputs that
04:43you get, it's very different from taking an entire mortgage package, you know, a 10 or
04:49three, an appraisal, a credit report, an AUS. I mean, all, these are all the real documents
04:55that drive the mortgage decision. And then have the, Jenny, had the LLM spit out a decision.
05:02That is not how it works. I mean, you, and then you also have to take into account what
05:08happens when, what kind of guardrails do you put around it? Because we know, I mean,
05:14just the nature of the technology. I mean, and I think, I think just an understanding
05:17of what that, what an LLM is and what does it, how does it work, right? Putting it into
05:23a little bit of a layman terms without getting into neural networks and, and, you know, all
05:28of those things and, you know, you know, rag and all those things. Just being able to explain
05:33that is really the key is that ultimately it is, it is guessing the next word. I mean,
05:39you know, all said and done, it's, that's what it's doing, but how you, how you prompt
05:43it is really what drives this output, but you still have to put guardrails around it.
05:47So, so, so I would say it's this exaggerated view of its, its potential, a lack of understanding
05:57of how to put guardrails around it. And the third is how do you, how do you keep up with
06:02your investments, right? Once you adopt it, how do you keep up with it? So, you know,
06:07things are changing. I mean, we hear this all the time. So how do you, how do you keep
06:10up? I think those are the three areas that I think executives kind of, many of them, some
06:15of them are, I mean, they get it, but we run into who, who still need a little bit of that
06:20guidance and handholding. And unfortunately there are, there are vendors in our space who kind
06:26of take advantage of that, that lack of experience and kind of oversell. So that's really
06:31where, at least from my perspective, that's what I see.
06:36Now, I agree a hundred percent. When I started at HousingWare in 2013, just, it's been an ongoing
06:42theme from my perspective that you have a number of tech companies come into the space and be
06:48like, oh, we can fix all this. Oh, you guys have a problem. Let us fix that. But they don't
06:52understand the mortgage side. They're just coming at it from the tech side. It's one of the reasons I
06:57brought up the fact that you guys have won this award for us because you're applying the technology,
07:02but you're coming at it from that knowledge base of like, you understand how complicated
07:07the mortgage process is. And I feel like that's step one.
07:11Correct. Correct. I mean, you, you bring up such an important point. Oftentimes people don't want to
07:18hear, they want the simple soundbite, right? And the simple soundbites are possible. I mean, the people
07:27who really don't know the details, I mean, ignorance is bliss. So sometimes you can just say it, oh, yeah,
07:33it can get done, right? And I wish it was the case, right? There is a, there is a happy path in, in the
07:40mortgage process, but that path applies only to a very small percentage of the loan population.
07:46Right. So if you really have to make that impact, it has, you need a more universal use case. And
07:52that's where, I mean, our background, we started as, you know, and we still are, I mean, we run the
07:59back office for a lot of lenders, servicers, title companies, mortgage insurers, everybody in the
08:08mortgage space, if you may, investors. So we are, so we run the back office. So, so we are, so our
08:15familiarity with the mortgage documents, the process itself, controls, compliance, what are the
08:23pitfalls of, you know, of, you know, of, of the, of that each step in the process, that is something
08:29that we live. I mean, we live and we breathe. So a lot of our, all our technology platforms,
08:35really, either they started with us building it to use it to deliver our services, or we actually
08:41used our, we took some, you know, our experience in terms of challenge areas, and we built that out
08:48as a use case, because we felt like, for instance, income calculation is a, is a big challenge for in
08:56the underwriting process. It is probably the number one area that even experienced underwriters stumble.
09:03It is, there are so many different variations that you, that come into play. So we actually
09:11tackle that first from an automation perspective, and that's really income genius. And so we kind
09:15of, we really started there and that helped us cut our teeth in terms of just what, how do we go
09:22about implementing automation and drive adoption? When you start working with users, you need to be able
09:28to get them satisfied and convinced that you know their business. You know, you're, you're, I mean,
09:35it's one thing to have, you know, build a tech in a lab, and then come in and, you know, you can blow
09:41the executives away. But then when you start implementing it, the user's like, whoa, wait a
09:46minute, right? And some of it is just resistance, right? I mean, some of it is just resistance to
09:49change. It's not that, it's not a genuine technology concern, but you still have to address that.
09:55You still have to bring them along the journey. So we see that as our biggest strength, is that we
10:01don't come in with this, this view that, oh, everything just, just once like wave of a wand and
10:07everything gets done, right? But we do have, we do have a methodology, right? A method to how we bring
10:13in that step by step so that, so that it's not a long drawn, expensive effort to implement the
10:20technology, but we focus on that adoption, the user comfort. And so, so it's really the use cases
10:25that we've built. And now that's transferring to Gen AI. So, so to be, I mean, I don't, I don't want
10:31to kind of jump ahead, but, you know, the Gen AI is just another technology, you know, you know,
10:39weapon in our armory, if you may, right? I mean, there are many others that we have used,
10:44and we see Gen AI kind of augment that on top of what we have.
10:48I think that perspective is so helpful and, you know, so important, especially right now. Let's,
10:54let's kind of dig down into some of those other use cases. You talked about income.
10:58What other real world applications do you see that are most important for lenders?
11:04Loan decisioning. I mean, I think just, that's just the whole loan decisioning.
11:08When you look at the lenders universe, I mean, they, they, they, that loan decision,
11:16A, it is just that they rely, I mean, they're still reliant on experienced underwriters.
11:21And when the, and that's the area that has the maximum volatility with the volumes, right?
11:27Correlation of the volume volatility. So it's like if the volumes, when the volumes drop,
11:33you have to let people go and they, I mean, they have to hire underwriters.
11:36It's a mess. I mean, we've dealt with it because we live in that world.
11:40So, so that, that taking that loan decision and breaking it into different components,
11:46automating that, automating each of those components so that when the underwriter goes into the file,
11:53like 60% of what they usually would do is already done for them.
11:57Maybe sometimes 70%, right? I mean, it's, it's, it's done. It's not, none of the stare and compare,
12:02go check this, go check that, which is what, it's a lot of what they do.
12:06They open a spreadsheet, they start calculating income, all of that's taken away and everybody's
12:10doing it the same way. It's consistency, right? So, and then from there on, the underwriter now
12:16takes over. So that's like, we're not at a point where you can just completely eliminate the
12:22underwriter. Right.
12:23And we wouldn't recommend that. That is a risky proposition. I don't know any lender who,
12:29who is willing to do that, right? I mean, it's, it's great to, it's great to hear that soundbite
12:33is great, makes great headlines, not a good, not what doesn't work in practice, but you want to
12:38make them more productive. You want to give them more information at their fingertips so that they
12:43don't have to go search the mortgage loan file. They don't have to do any of that. So that's like
12:47one use case. I mean, that also drives, you know, better borrower experience because you are now
12:53having less than number of touches on the loan, right? I mean, you're giving all the information
12:57ahead of time so that the loan officer can interact with the borrower. You make that a much smoother
13:01experience. Another use case, which, which you would, I mean, for the longest time, I thought that
13:07that is a use case that has been addressed. A business, a problem that's been addressed is,
13:12it's just the act of indexing a loan file into the right bookmarks and identifying the different
13:21versions of a same, of a same document. And for example, a 1003, the loan application, I mean,
13:27there can be five versions of it. There can be one for each borrower. So maybe two borrowers. So now
13:31you're talking about 10, you're talking about 10, 1003s. Each one is, each page, each one is 15 pages.
13:38There's 150 pages right there, right? If all of that is lumped in one bookmark, imagine now the
13:44underwriter or the closer or the processor or anyone in downstream in the, I mean, even in capital
13:50markets, even, I mean, just think about it. There's so many people who are dependent on the 1003,
13:56they have to open that bookmark. They're going to go find the latest version, the first version
14:00for borrower one, the last version. I mean, you know, imagine it's just, this is this,
14:04it sounds very mundane, banal, but it matters when it comes to the effectiveness of decision
14:11making at each step of the mortgage process. So it's this, it's this, this indexing, making
14:17sure that the docs go into the right folders, that it's versioned properly, they can actually
14:22see it and they can get to it quickly. That is a use case that I thought, frankly, it was
14:30addressed. Now we, in order to do the loan decisioning, which is that a product is Decision
14:37Genius, we need to be able to extract data from documents. So we already built, so we built an
14:43enabling technology, I would call it, called IDX, which really gets to that. It's, it's, and it's
14:48not, it's not so unique that nobody, everybody, a lot of people, other people have it. What we've
14:53done differently is we take on the responsibility of making sure that the docs are in the right
14:59folders and we don't push that responsibility onto the lender, because that's what the other
15:04techs do. They basically get it done, they get it up to 80%. The remaining 20%, they say,
15:09okay, you go do it. Now that 20% is not exactly 20% of effort, because you don't know what is not
15:15in the right folders, right? And so now it's like, it's, you got to go search for that. So the 20%
15:20of the docs may translate to 50% of the effort. So that is something that we have taken up because
15:26we said, tell our customers, you don't have to worry about it. You're going to get it because
15:29what we, because we have this capability of, because we do the actual back office work,
15:35we can, whatever the automation misses, I mean, because it's going to miss. I mean,
15:39there are things that it is never a hundred percent. We, we commit, we make that commitment
15:43and we put those into the right folders. Now that is a use case that has been extremely well
15:49received for us in the industry. In fact, we started marketing that as a standalone service
15:54only about a year, year and a half ago. And we're just like inundated with these custom large
16:00customers who are still facing that problem. So hopefully that gives you a flavor. I mean,
16:04it's these kinds of things that, you know, on one end, it's this low decisioning, you know,
16:08very esoteric, very like heavy duty stuff. Right. And it's just like making sure the docs
16:14are in the right folders. It's kind of the entire spectrum. And all in those areas, I mean,
16:20all the, the entire repertoire of automation technologies comes into play, machine learning,
16:26LLMs, you know, that the, you know, just in just a pattern matching, maybe use some,
16:32I mean, we use some fuzzy logic, we use things like AGEX, regular expression. I mean, there's
16:36a plethora of technologies that you've got to use. I mean, there's a rules engine that drives behind
16:42it. So you've got to use all of these things to make sure that you get to the data, you get the
16:46docs in the right folders. So, so, so the underlying technologies are all the same, but the
16:50use cases start now multiplying. So earlier you talked about guardrails, right? You mentioned
16:56guardrails, now you have to have them. What, what kind of guardrails should lenders have in place to
17:00ensure, you know, that AI tools are reliable, they're accurate, all, you know, this is so important and
17:07we know it's sort of a black box. So how do you, how do you approach that? I mean, that's a good
17:13question. I mean, do you, it is, it is something that I think every automation solution has to
17:21contend with. You just have to, first of all, it's at the, at the, at the lowest level, when
17:29you look at the whole automation stack, I mean, the highest level is the end user, right? I mean,
17:33the end user is reviewing the results. At the lower most stack, it's really when the data is being
17:39obtained and extracted from different sources. You, there are techniques that, that can be built
17:46in, use deriving confidence scores. I mean, you can use, there's this bunch of different techniques
17:51that then you can augment it with, with wherever you say you're not that confident, you can actually
17:57have a human verify it, right? And the, and the nice thing about that is when the human verifies that,
18:02that is a feedback loop back to the model, right? It's giving the model saying the next time you
18:08encounter something like this, don't push it into the human. You know how to handle, this is how you
18:13got to handle it. Sometimes it takes like 15, 20 of those cases for the model to learn, say, okay, I
18:19got it. Now I'm not going to push this. So over a period of time, you eliminate those kind of edge
18:24cases, right? So now at the lowest level, it's important because everything that you do about that
18:29depends on the data that you're getting from you, right? And the data is wrong, it's garbage in,
18:33it's garbage out. So you got to get that fixed. So you, so that is one kind of, if you may,
18:39kind of a check and balance. As you then start building out the use case around it, and as you
18:45start, you know, using a combination of rules, and you could be using Gen AI, I'm using a series of
18:50prompts, you know, you basically are like a given, I mean, some very simple example. In the, in the,
18:57in the world of underwriting, one of the things you have to verify is if there are, if there are gift
19:02funds, right? Funds will come, came in through a gift that somebody gave the borrower, right? I mean,
19:06a family member, a friend to put down the down, you know, down payment. I mean, Fannie requires
19:11you to disclose that. I mean, find out. And usually that is accompanied by a gift letter.
19:16That gift letter is a non-standard document. I mean, everybody can write whatever letter they
19:19want, right? So you basically, so, so you can actually have, you know, through Gen AI have a very,
19:25I mean, very sophisticated prompt that actually reviews that and be, and deduces that there is a gift.
19:33The gift is at this amount, it is given from this person to this person, and you, you package that
19:39entire information into a description, a condition in the LOS, and you push it into the LOS, right?
19:46That's what we do. Now, the underwriter who's reviewing it, the, their job instruction has to be
19:52like, don't just take it as, for face value. You have all this information, just open the gift letter,
19:57which, by the way, has been indexed and put into the right folder, right? I mean, so you can,
20:00you don't have to go search for it. It's there. Just make sure it's okay, right? I mean,
20:04you basically still want your eyes to be on it, to just make, I mean, that is, I mean,
20:09where we feel that because there was a lot of non-standard information in that gift letter,
20:13that's why we're asking you to do it. But when it comes to standard documents,
20:17like a 1003, you know, an AUS, we, you know, we, any condition that's driven from that,
20:23the underwriter, I mean, the underwriter can be instructed, just don't, don't, don't worry about it.
20:27Just, you can trust it, you can move forward, because we have these layers of, you know,
20:32checks and balances that we've built in to make sure that the data is correct, and the condition,
20:37and the, and the, and the prompt, or the, or rule that ran on top of it is, it was,
20:43was executed properly. So, so it's, it's that kind of combination, because sometimes what you see
20:48is that, I mean, you can train the model, the LLM, because the LLM thing is that,
20:53we're not developing our own large language model. We're using what's out there, right?
20:59And then that's the other part that we do. We kind of triage between, I think, use multiple,
21:05you know, multiple LLMs to be able to get to the right results based on what we're doing,
21:11which task at hand. But even, even then, I mean, it could just subsistate. So, so you,
21:17you know, what do you do? So, so we identify the conditions that have been generated through
21:21that LLM through these inferences, as we call it, we call it inferences. And then those are the
21:26ones that the underwriter can, can make sure that, yeah, I mean, it looks good. I mean,
21:31it matches up with everything else. Because once they're in the file, they know everything about
21:35that file. They just need, they just need additional information at their fingertips that
21:40they don't have to go search. And we see that that drives up productivity enormously. I mean, just
21:45because you're not replacing the experienced underwriter. I mean, yes, you may, you may be able to go
21:49with a lesser experienced person, not somebody with 25, 30 years experience, but that underwriter
21:54knows how to underwrite that file. So, so that's kind of how we, you know, how we kind of build all
21:59these guardrails, checks and balances, so that, so that there is no risk of, you know, mass scale
22:05errors that, that have a cascading effect. So, yeah, I think this is still evolving. I mean,
22:11we're kind of coming up with new checks and balances. How can we build those guardrails? But
22:15building guardrails has to be a central part of your, you know, AI automation strategy.
22:24Absolutely.
22:26I think that that's really interesting, because my next question was going to be about how do you keep
22:31the humans in the loop? But it feels like you guys have already, that in some ways that's organic
22:36for you, because as you're looking at the whole process, as you're looking at the operation, you
22:40know where the human needs to be.
22:43Correct. I mean, we basically look at it and say, I mean, how can you make the humans more
22:48productive, far more productive? So you, I mean, yeah, you need lesser of them. I mean, let's put
22:54it that way. There's definitely, because there has to be a cost advantage. I mean, you cannot carry the
22:58same team. You cannot carry the same capacity to, as your volumes go up, because that is one of the
23:06challenges in this industry, right? And that's never going to go away, because we are driven by
23:10factors that are beyond our control, like whether it's interest rates, whether it's the state of
23:15the economy, inflation. I mean, we don't control any of that. So you're going to have to build a
23:21business model that's predicated around that volatility. And so it's how do you build an
23:29operating model that allows you to bring down that cost, but that cost without sacrificing the
23:35quality, effectiveness, turn times, none of that. So you probably need, you definitely need less
23:41of other people, but you're not, you're not, your goal shouldn't be to just, oh, I'm going to get rid
23:46of everybody. Definitely in some of the, I would call it the task-based functions, like
23:53indexing, data entry, I mean, things that are still done today. I mean, that you can completely
24:02eliminate, because you, because that's, at least from the way we go about it, we tell our customers,
24:07you don't need to have staff on your end at all doing that. But when it comes to underwriters,
24:12loan reviewers, auditors, you know, just those, those, you know, people who do the loan, the QC,
24:18you still, you, you're making them far, far more productive than they are. And you're bringing about
24:24a consistency and a standard in way in how they do their tasks and their functions.
24:30That, that strategic use of like, here's how it's going to operate. Here's where you need a human.
24:36All of that is, is so important. Let's talk about adoption a little bit, right? Because what is a
24:42phased, realistic roadmap look like for most lenders? Like if they're looking at like, okay,
24:47what's my next step? What does that look like? I mean, the adoption, as I mentioned, that's where,
24:53because you're, wherever there, wherever there is, there is, there are humans remaining in the
24:58process, right? I mean, there are some functions, stare and compare, indexing. I mean, any type
25:04involved, anything that we call, you know, just stare and compare, right? Right. It's gone. You don't
25:09need it. You don't need any human in the loop. It's just, but where there is humans, you want to be
25:15able to, the biggest challenge that we see is really undoing what, you know, years and years
25:23of habits, right? There's a certain way they do things. So how do you, how do we get them to see
25:28that, that, how do you embrace the technology as being something that helps improve their, improve
25:35their output and their productivity? I mean, I think that's really what you've got to get to. So,
25:39so you, it's, it's got to be, it's got the information has to be presented to them in a
25:45consistent, clear, concise manner, you know, and, and, and in case of a load positioning
25:52automation, it takes the form of conditions in the, in the, in the LOS, because, you know, again,
25:59you're not doing away with the LOS. I mean, the LOS is still the, the heart of the, the, the heart of
26:05the, the, you know, the, the, the process and we're, we're pushing stuff into it. So it's,
26:11it's, so it's very seamless. It is, I mean, it's quick. I mean, that's the other piece. It's quick.
26:18They don't have to wait for it. They don't have to wait for hours. They don't have to do it
26:21themselves. A lot of it is ready for them. And I think, I think just, just get bringing them into
26:28the fold. And we basically have, you know, we advocate that when you implement something new,
26:34you, you create like a core focus group. We tell our lender clients, create a group
26:39of adopters who, you know, are open to change. I mean, you know, not, you know, who they are in
26:45your organization. They're saying, oh, I'll take this on, right? You create that group. You want to
26:49create some internal advocates. And, and, and once that happens, I mean, then, then it's like,
26:55it's amazing when they hear from their own colleagues, then from a vendor or from their
27:01management, actually. It's, it's amazing how, how, how they kind of just quickly embrace it. So,
27:07so we, we, that's, that's, that's the other piece that we advocate. You want to make it,
27:12the key is, even if there is, even if there are things that they have to do in addition to what the
27:19automation does, you want to make it consistent. You don't want something to work one day. And I mean,
27:25it's like one way, and then the next, next loan, I mean, like a month later, they see another loan.
27:30It's not working consistently. So having that consistency is the key. I mean, these are things
27:35that just, just drive that, the human behavior of saying, yeah, I can trust this. I, you know, I,
27:41I can, so it's, it's building that trust. And we basically say trust, but verify, right? I mean,
27:45trust, but verify, trust, but verify, right? And then once you get to a point, it looks good,
27:49you can trust it. I mean, you don't need to verify at that point. So, so it's, it's that,
27:53it's just that, and that's why we say, I mean, the automation journey, and especially with Gen AI,
27:59I mean, we kind of, I keep telling my client, this is not a sprint. I mean, this is a marathon,
28:04right? I mean, you gotta, you gotta, you gotta, you gotta be prepared for that. You gotta work
28:08through that. I mean, and, and that's really our, our proposition, right? We're not coming in and we're
28:15not in and out. I mean, we, we, we build that long-term relationship. We, we keep enhancing,
28:21we make huge investments in our technology platforms. Um, I mean, our, you know, that,
28:28that's what we've been doing for so many years. I mean, 18 years, as I mentioned. Um, and, and we
28:33just, and, and so that's, that's, that's what our, you know, clients, the, that's the assurance
28:38that we give them, that we are, we are here with you through this long race.
28:43Amazing. And, and great stuff. Okay. I have one last question that is looking ahead. Let's look
28:50to the future. How is Indicom looking at the future?
28:53Many, many, many different, uh, dimensions to it. Um, but I'll, I'll just say that where we want to
29:01be is, or at least I'll, I'll, I'll, I'll speak from the Indicom wants to be. And then I think just
29:07from the industry perspective, maybe let me start with the industry, right? I think the industry
29:11industry is still grap coming to grips with, it has come to grips with the fact that the market is,
29:19you know, is, is, is what it is, right? I think, I think it, you know, after the COVID years,
29:2522, 23 was really hard, uh, 24, 25. I mean, this is stabilized. And I think it's waiting for that
29:33next, when, when does the next boom come around, right? I don't think it's coming around that quickly.
29:38It's not, it's, but it's at the same time, it's, it's growing, it's growing, it's steady growth,
29:44right? And that's a growth that most people can put their arms around. Uh, but I think, I think
29:48just where they are, where they're still grappling with is how do you, what's the best use of
29:53technology? Um, there are so many different options that are around there and, and, and the people in
30:00decision-making, um, the decision-making chair, um, I have to, I mean, they're just, just sifting
30:06through all those options and making, and making the right decisions. I think that's
30:11their, that's the number one challenge that they have as far as the technology adoption
30:14is concerned. Um, I think, I think the industry is still figuring it out, but I feel about where
30:20I see it go is, um, there is going, there are going to be some really practical use cases
30:27of AI, both chain AI, things like agentic AI to, to be able to, you know, improve building
30:35the agentic AI components into the LOS, for instance. I think the LOS vendors are going
30:41to be tasked and challenged to do that. Some of the more established ones are going to have
30:45it harder to do it because even though they have more resources, I mean, I shall not name
30:50anyone because we are partners with all of them. They're going to have challenges simply
30:54because they, they, they have made existing investments and their technology is rigid in
30:59that sense, right? Because it's been built and not any slam on their design and things
31:03like that, but that's how they, that's how things are. Once something is in, it's harder
31:07to build, kind of completely transform it, which is where I think we're going to see some
31:12new, the new age LOS, uh, solutions that will have a great opportunity. They, they will
31:20basically be able to build some of this things like agentic AI, you know, chatbots, gen AI. I mean,
31:27all of that into their, into the process, right? So that you don't have, you know, so you can, you can,
31:33you can start now imagining and visualizing the mortgage process very differently. Um, I think,
31:39I think, um, uh, this whole thing of, you know, completely touchless lending, zero touch,
31:44human touch. I mean, I don't know if, I don't know when that's going to happen in our lifetime.
31:49Uh, this is going to probably happen for a small percentage of loans, but there is going to be a
31:53large percentage of loans that are still the humans involved, but how do you make best use?
31:58And I think your LOS and your other technology that, that surrounds it, like what we provide
32:02is really what's going to drive it. So, which is where, which is where, where, where I see Indicom
32:07being is staying on that leading edge. I won't say cutting edge, right? I mean, because you cannot,
32:13I mean, and we have a whole R and D function that happens that's, we are constantly evaluating
32:18new stuff. We bring it in very responsibly because you can't just like, okay, oh, there's a new thing.
32:23Let's just incorporate it. It's very disruptive and it's not proven. So, so that's what we, we want to
32:29be that, uh, we want to be that, um, the, the marathon coach, if you may, uh, you know, the, the,
32:36that's with the, with the, uh, with the lender as they're running through that long race, right?
32:41So we are there every step of the way. Um, and, and we're, and we're making the investments and
32:47we're staying ahead of the curve as much as we can and, and driving that, you know, being part of
32:53that automation journey that they're inevitably, uh, they're, they're going to go down. So, um,
32:58so I think that's where I see it both for the industry and as for us.
33:02Raj, this has been a great conversation. I could talk all day about this. Um, we are out of time now,
33:07but it's an exciting time for the industry and it's an exciting time for Indicom. So thank you so
33:12much for being on. Thank you, Sarah. I mean, I, and I can go on all day. So if you don't have,
33:19no, it was great. I, I so appreciate it. And hopefully we'll talk again soon.
33:23Yeah. Likewise. Thank you. Thank you for having me. And it was wonderful talking to you.
Be the first to comment