- 4 days ago
Indecomm CEO Rajan Nair joins HousingWire's Allison LaForgia to discuss why trust, consistency and transparency are essential to AI-powered mortgage quality control. Nair explains the challenges of automating QC, the importance of repeatable audit findings and how Indecomm's AuditGenius and IDXGenius solutions are designed to help lenders improve audit efficiency while maintaining human oversight.
Category
🗞
NewsTranscript
00:12From Chicago, I'm Alison LaForgia with HousingWire, and today I am sitting down with the Indicom CEO, Rajan.
00:19Hey, Alison.
00:20It's nice to see you.
00:22Great to meet you, and thanks for having me here.
00:24Now, you've spent 25 years working across mortgage due diligence and technology.
00:30How has that experience shaped the way that you think about where I can and can't create value?
00:39That's a great question.
00:41It has been 25 years.
00:42It feels like 50, given what the mortgage industry throws at us.
00:48For me, the core for value is, does it make sense for the customer?
00:54And does it deliver enduring benefits and tangible and measurable benefits?
01:01So, we've always centered our – everything that we do has been centered around our customers.
01:07So, 25 years ago, there was an automation.
01:11I mean, there was paper in the industry.
01:12I mean, there was actual physical paper.
01:15And so, at that time, we spent a lot of effort in helping digitize the paper.
01:20So, our journey really starts from there.
01:23And so, there, too, it was about adding value.
01:26And as the technologies evolved and got better, we started incorporating them into our own operating models.
01:34And so, for us, our customer always – I mean, there's nothing more important for us than that.
01:40And that has been, I would say, the core reason for our success.
01:45And as far as can't, I mean, we have – you know, I don't believe that there's anything that can't
01:52be done.
01:53So, we have that attitude in our company.
01:56And I'm lucky to have a team that – where that attitude percolates.
02:00That is fantastic.
02:02Leaning into the opportunity to actually innovate is always a really interesting philosophy.
02:08What can't you do?
02:09You just go right through it.
02:12Yeah.
02:12I mean, it's – I mean, there are certain areas of the business.
02:16They're not our competencies.
02:17We stay away from them.
02:18I mean, like, for instance, contact centers.
02:21I mean, we don't do anything that involves customer contact or rather end customer contact.
02:26Our focus is really on the middle office and the back office.
02:29That's where we do really well.
02:31So, if there was – I mean, those are the only things that prevent us.
02:34I mean, this is a conscious decision that we've made.
02:36But within that, we have innovated.
02:38We have built a lot of technology, a lot of capabilities.
02:42And that, too, basically an execution model.
02:45So, it's not just the technology, but also where the people need to come into play.
02:49We have that as well into our value proposition.
02:52So, for us, it's a – you know, that's really what drives our innovation.
02:59And now, of course, we're on the cusp of AI.
03:02And, you know, that's where we arrived at.
03:03But, again, our technology goes back when, you know, when I said there was paper in the process.
03:09And so, we've lived through it all, essentially.
03:11Now, one of those really important processes that the whole industry looks at is QC.
03:18How is QC different from other areas of the mortgage process from an automation standpoint?
03:23And why has this area not been successfully automated as some of the other areas upstream in the process?
03:31QC – there have been attempts at automating QC.
03:34I mean, it isn't that there hasn't been one.
03:37We have been at it for over five to six years.
03:41QC – the main challenge with the automation is whatever automation solution is deployed in the world of QC, because
03:48QC comes in much after the origination process, the QC auditor is expected to be at a higher skill level
03:58than even the underwriter.
03:59QC – so, what's expected from them in terms of the result that's expected from them, that quality, essentially, is
04:05the key.
04:06And so, we've spent a lot of our time finding out what stops us from delivering that quality and what
04:13stops us from building that trust within the auditors.
04:17QC – so, that's the key, that can you get an automation solution that the auditors can trust, embrace, and
04:23adopt?
04:24QC – and there are many ways to go about the automation.
04:28And I think a lot of what is out there in the market has focused more on trying to automate
04:35a checklist.
04:36QC – we've gone away from that.
04:37QC – I mean, I think that's really been our focus in terms of trying to see what drives the
04:42QC itself and tying it back into the core requirements of the mortgage – of the loan process itself and
04:49tying that into the solution has been our focus.
04:52QC – so, that – and we have a huge QC organization within our company.
04:57QC – so, we have a place where we can test all of these things, and we don't have to
05:00wait for a customer to try it.
05:02QC – so, we try, we keep trying, and we've innovated, and we've kept adding new capabilities.
05:07QC – and now, finally, with where the frontier morals have reached with AI, we're seeing this immense opportunity to
05:15bring all that experience and that learning that we've had into now what we feel is a truly adaptable QC
05:24solution.
05:25QC – I want to stay on trust.
05:27QC – for a moment, if a lender runs the same loan file through an AI audit twice, why does
05:32it matter if they get the same result?
05:34QC – our industry is highly regulated.
05:37QC – I mean, it's highly regulated, investors have very strict criteria, guidelines, everything revolves around that.
05:44QC – so, if you cannot repeat any – by the way, I mean, that applies to pretty much any
05:50process, more so QC, because you're supposed to be coming in and you're supposed to be auditing
05:56QC – what's already somebody else's work. And if you cannot replicate that same loan, the same outcome, I mean,
06:04you basically just completely lost trust.
06:06QC – and the auditors are inherently suspicious people. I mean, I don't mean it in a negative sense. I
06:11mean, they're forensic in their attitude, right?
06:13QC – Exactly.
06:14QC – So they're now – so if you basically give them the scenario where they say, hey, you ran
06:19this loan, this was a result, and now it's a completely different loan, you basically, you've lost them.
06:24QC – And therefore, so it's more – it's obviously from a regulator perspective, and it's also from an auditor
06:30acceptance and an adoption perspective that it becomes so critical.
06:34QC – And it's not an easy thing. I mean, especially with the AI technology that's out there, the way
06:39– there are vendors that we know out there who are deploying the solution,
06:43QC – more from a perspective of just exposing a model, anyone, any of the foundational models, to a checklist
06:52and to the documents and say, now let's go automate this.
06:56QC – It works in a lab environment. But as you start releasing it, you don't see the repeatability and
07:02the consistency of the results.
07:03QC – And what we've done essentially is to be able to, with our subject matter experts, with our –
07:09we have a very deep expertise in that area.
07:12QC – Everything that drives the automation, we put it into a box, and we provide that scaffolding that's necessary
07:19for the automation to run.
07:20QC – So that once you deploy the AI models within that box, it performs amazingly.
07:26QC – But if you just let it run loose by the very nature of how those models work, you
07:31don't see the consistency and the repeatability of the solutions.
07:35QC – So let's tie a bow on this very specific use case. When lenders are being told to trust
07:40AI in their QC platforms, should they?
07:44QC – You would think they should. I mean, that's why – I mean, that's the investment that they've made.
07:49QC – But they need to ask a few questions.
07:53QC – One of the questions they have to ask is, first of all, they have to ask the question
07:56is, where is the data coming from?
07:58QC – The data is often the most ignored aspect of any automation in our industry.
08:05QC – The data is not uniform. There are many attempts to unify data models, data structures, but data resides
08:14in documents.
08:15QC – There's a ton of information that resides there. So you've got to get the data out.
08:18QC – So we lead in with that. And so because we have IDX Genius, which is our data extraction
08:25solution that we, again, using AI, it basically produces –
08:30QC – We extract over 17,000 data points for a single mortgage.
08:34QC – But that too is repeatable and predictable. And so that's the first thing that they've got to ask.
08:41QC – The second thing that they've got to ask is, how much AI is running real time, essentially?
08:47QC – You want to restrict how much the AI models can do. And this sounds like counterintuitive.
08:54QC – You want to do all the AI work up front for the authoring or the preparation of the
09:01system.
09:02QC – You want minimal AI running in production in live in real time.
09:07QC – Because once you let it loose, you cannot control it.
09:10QC – No matter what anyone says, it's not controllable. So you want to have that controlled environment.
09:15QC – So that's the other thing that we've got to ask. And the third thing, from a trust perspective,
09:19and this comes into play from a compliance, and we're seeing this a lot.
09:22QC – The moment you say there are AI models, then you have to prove how the AI models are
09:27not biased.
09:29QC – What's the confidence behind it? And that's built into our solutions, into our platform itself.
09:38QC – So that's where Audit Genius, which is our QC platform, Audit Genius AI, and we're very proud to
09:43launch that here.
09:45QC – And Audit Genius AI is our next version of Audit Genius, essentially. And Audit Genius has been in
09:51production for 20 years.
09:52QC – This is really the true, completely AI-first model, the business model that we've deployed.
09:59QC – And of course, there's IDX. So that's really what drives our new product suite, essentially.
10:04QC – A new approach, our AI-driven approach to QC.
10:09QC – Now, you made a very important clarification point about where AI should be in a process.
10:15QC – And all of the extensive work that you've done at Indicom to make sure that that makes sense.
10:22QC – I want to make sure that this gets called out to our audience.
10:25QC – When an auditor or investor asks how a finding was made, what really should a lender be able
10:32to show them?
10:34QC – They should be able to show them a glass box.
10:38QC – You look at the findings, and you need to be able to get down, right down to the
10:46document, to the page, to the field.
10:49QC – You need to be able to show them, this is what drove this particular decision or outcome.
10:54QC – And that's where, again, the data model, the underlying data model, that is where it's critical.
11:00QC – But more importantly, it has to follow a consistent path.
11:04QC – And when you present the findings on a loan.
11:07QC – So essentially, QC, what comes out of QC is findings, right?
11:10QC – There are these errors, these issues, and these issues, these are major issues, these are moderate issues.
11:16QC – But every issue that you point out, you need to be able to drill down.
11:19QC – But also not just issues.
11:21QC – What was cleared by the auditor?
11:23QC – Why was that cleared?
11:25QC – Were there false positives there, right?
11:27QC – Because false positives are a big factor in AI, in using AI models.
11:31QC – So every aspect of the QC that has been done, you must be able to drill down to
11:39the last minute data field, essentially.
11:42QC – Now, how quickly should a lender be able to answer that question?
11:47QC – And what does it actually take to get there, to get to that glass box that you mentioned?
11:52QC – Instantaneously.
11:53QC – I mean, they need to be able to…
11:54QC – Yeah, I mean, today, the technology is available.
11:57QC – And the technology, again, it's not running.
12:01QC – So a QC auditor is auditing a loan.
12:06QC – Everything that's presented to them, all the AI work that's done is already done.
12:12QC – It's not when they're opening up a loan that any of that's run.
12:16QC – Everything is almost staged, essentially.
12:18QC – And that staging is the scaffolding that I'm referring to.
12:22QC – So that behind the scenes, tracking down to every single investor requirements,
12:28QC – Lender overlays.
12:29QC – Mapping that to specific documents and fields within those.
12:34QC – And then mapping that back into the checklist.
12:37QC – So that when the auditor logs in to their system on that loan.
12:41QC – They see here, like here are the seven findings.
12:43QC – Here are 10 findings.
12:44QC – Here are all the things that were cleared.
12:46QC – They need to be able to say, okay, at this point, I'm going to be able to click
12:50all the way down to the document.
12:51QC – But the reason they're able to do it is because all the work that was done here.
12:55QC – And that's what we've done with our approach to AG.
12:59QC – We call it AGAI.
13:00QC – And so that's really what we've done.
13:02QC – Everything has been done here.
13:03QC – And again, I just have to reiterate this is that it all came about because we've been at
13:08this for years now.
13:09QC – And we've seen where all the pitfalls are, where the adoption challenges come in.
13:13QC – And we've painstakingly kind of really gone down to that level of detail.
13:18QC – And again, we are fortunate because we have a very large QC operations ourselves.
13:23QC – So that's really what's kind of what we've seen.
13:26QC – And that's really what we're now able to talk to lenders and show them actually it working.
13:31QC – Now, looking ahead, where do you see the greatest opportunity for AI to improve the mortgage process over
13:37the next few years?
13:38QC – And what gives you the confidence that we can actually get there?
13:42QC – I think it's all around.
13:44QC – I mean, I think the, when I look at, in fact, I think this week,
13:48QC – I saw a report from the NBA on AI and its future and its adoption.
13:53QC – And I thought it was a very thoughtful, I don't know who wrote it.
13:55QC – I saw it in a blog, but it was from the NBA.
14:00QC – It really came down to this conflict between what does a human do?
14:05QC – And what does the AI do?
14:06QC – Which is a source of anxiety for everyone, right?
14:10QC – I mean, I think all of us worry about that.
14:13QC – And what it's trending towards is that there is going to be a harmonious existence between the AI
14:23models and the human.
14:24QC – The human is going to be able to do a lot more than they were able to do
14:28before.
14:28QC – And in fact, now when I'm talking to clients, I mean, we talked to one very large client
14:32last week.
14:33QC – And their approach was like, I'm not looking to reduce my cost.
14:37QC – I'm trying to see if my people can do a lot more.
14:40QC – And so driving a lot more production, if you may, through that, with that headcount.
14:45QC – Putting it crudely, that's what it sounds like.
14:47QC – But what it comes down to is, these AI models are just going to keep getting more and
14:52more powerful.
14:53QC – They're going to be able to do things that humans would take like several minutes to do.
14:59QC – They'll take seconds.
15:00QC – But what the humans can do is like, look at that and basically now make decisions.
15:05QC – And make inferences.
15:07QC – And basically, you know, create new products that we could not do before.
15:12QC – The mortgage industry is constrained by its traditional technologies.
15:16QC – And it cannot deliver more products.
15:18QC – I mean, I think everyone will benefit eventually.
15:20QC – I think we need to just give it some time.
15:22QC – But that's really my fervent.
15:24QC – And I love to see that MBA report that came out, which kind of just kind of echoes
15:28that.
15:29QC – It's almost one of those things where you don't know what you don't know.
15:33QC – And an example that I have heard Tila Mathias give in her speeches has been,
15:42QC – You have to think about crossing the ocean.
15:46QC – It used to take weeks, months.
15:48QC – Then we had aircraft that could do it.
15:51QC – And suddenly there was a whole option of travel possibility that's expanded horizons for all sorts of people.
15:58QC – So you don't know what's possible until it's possible.
16:00QC – Yeah. And we still have a very successful cruise industry.
16:03QC – Exactly.
16:03QC – So people still go out on seas.
16:06QC – I've done one.
16:07QC – And it's, there's still a great, a great harmonious existence.
16:12QC – There is. And I think, I think much of this angst is understandable, but it's also being, unfortunately,
16:20being promoted by people who have vested interest.
16:22QC – I think, eventually, the technology will bear itself out, just like the internet did.
16:27QC – Rajan, thank you so much for joining me and talking me through how lenders should approach AI in
16:33their QC process and how Indicom is using AI to continue to create value.
16:38QC – Thank you, Alison. This was a wonderful conversation. I really enjoyed it.
16:42QC – Thank you.
16:43QC – Playing with.
16:44QC – Speaking with.
16:46QC – And, more or more or more.
16:47QC – Are you still there?