- 5 months ago
On this week’s sponsored episode of Power House, Zeb Lowe, HousingWire’s Senior Director of the Content Studio, chats with Rudy Zabran, the COO of Consolidated Analytics, to do a deep dive into AI’s role in mortgage, specifically when it comes to automation and data normalization.
Rudy discusses his role in transforming Consolidated Analytics into a 700-employee powerhouse, the transition to a data-first mortgage approach, aligning technology with company culture, and insights on the build vs. buy dilemma in AI technology.
Here’s what you’ll learn:
How Rudy ended up in mortgage after a career in fashion
The role of fractional components in automation and AI
The specific challenges and opportunities present in adopting AI in mortgage
The importance of aligning AI tools with a company’s culture and goals
Rudy's hands-on approach to staying connected with industry challenges
Related to this episode:
Rudy Zabran | LinkedIn
https://www.linkedin.com/in/rudyzabran
Consolidated Analytics
https://www.consolidatedanalytics.com/
HousingWire | YouTube
https://www.youtube.com/channel/UCXDD_3y3LvU60vac7eki-6Q
The Power House podcast brings the biggest names in housing to answer hard-hitting questions about industry trends, operational and growth strategy, and leadership. Join HousingWire president Diego Sanchez every Thursday morning for candid conversations with industry leaders to learn how they’re differentiating themselves from the competition. Hosted and produced by the HousingWire Content Studio.
Rudy discusses his role in transforming Consolidated Analytics into a 700-employee powerhouse, the transition to a data-first mortgage approach, aligning technology with company culture, and insights on the build vs. buy dilemma in AI technology.
Here’s what you’ll learn:
How Rudy ended up in mortgage after a career in fashion
The role of fractional components in automation and AI
The specific challenges and opportunities present in adopting AI in mortgage
The importance of aligning AI tools with a company’s culture and goals
Rudy's hands-on approach to staying connected with industry challenges
Related to this episode:
Rudy Zabran | LinkedIn
https://www.linkedin.com/in/rudyzabran
Consolidated Analytics
https://www.consolidatedanalytics.com/
HousingWire | YouTube
https://www.youtube.com/channel/UCXDD_3y3LvU60vac7eki-6Q
The Power House podcast brings the biggest names in housing to answer hard-hitting questions about industry trends, operational and growth strategy, and leadership. Join HousingWire president Diego Sanchez every Thursday morning for candid conversations with industry leaders to learn how they’re differentiating themselves from the competition. Hosted and produced by the HousingWire Content Studio.
Category
🗞
NewsTranscript
00:00We're looking at it. We're trying to figure out, does it reduce friction?
00:04Can it unlock value? And can we trust it?
00:07Those are kind of the three factors that we're looking at when we're talking about building technology
00:13or partnering with a vendor ourselves to implement technology.
00:17Those are the critical components.
00:26All right, Rudy, thank you for joining me today.
00:27I'm happy to be here. It's a great event out here and what a great facility to be having it on the SMU campus
00:32at the George Bush Presidential Center, I guess is what it's called. A very nice venue.
00:37Yeah, so for any of our listeners that are not able to watch the video, we normally, you know,
00:41we do these remotely normally and we have the occasion to sit down face-to-face because we're here at the AI Summit.
00:48So on that topic, what's it been like here for you today?
00:51It's been great. I mean, you have a lot of good speakers out there, a lot of different mixed perspectives on things.
00:56There's been a lot of back and forth around, you know, what jobs are going to be disrupted in the mortgage industry.
01:00We always find that, you know, an interesting topic.
01:03We just listened to the CTO at Rocket kind of lay out their path, how they're thinking about automation in the business.
01:10I think that's interesting. So it's been good.
01:12A lot of different perspectives in the business out there to listen to.
01:15A lot of smart people on the floor.
01:16Yeah, for sure. I don't know a lot of what our conversation today is going to be.
01:21But before we do that, I wanted to ask you, we've actually, we've had a couple of conversations.
01:26We did an interview with you, I don't know, six or seven months ago, I believe.
01:30And we dove straight into the topic at hand, which was AI and tech.
01:34And I did get a chance to ask you about your journey into the industry.
01:38So can you walk me through that?
01:39Yeah, sure. My journey in the industry was pretty unique.
01:42I was in the shoe business fashion industry for many years, worked for a family company called National Independent Retailers, a buying group.
01:50Spun that over, went into work for wholesale accounts at Candy Shoe Company and then Steve Madden during his famous Wolf of Wall Street debacle,
01:58which ended up, you know, took him to prison for a little while.
02:01After that, left and got an opportunity to learn how to underwrite with, at the time, it was American Mortgage Consultants, now CITUS AMC.
02:10And started there in 2005, left with a few partners and started, founded Opus Capital Markets Consultants,
02:17sold that business in 2014 and took an equity and operating stake at Consolidated Analytics,
02:21where we're kind of re-envisioning how mortgage services work.
02:26What were you doing at the shoe company? Like, what was your...
02:28Wholesale sales.
02:29So I was on the road traveling to conferences like this one and schlepping shoes around a season in advance,
02:36you know, selling the next season's line of shoes to major retailers across the country.
02:42All right. So you weren't in design or...
02:44Did a little bit of that.
02:45Oh, really?
02:46Yep. Did a few trips to Italy and Brazil, where we did some design work as well and a variety of other things.
02:51So, yeah, I've done a lot of interesting things.
02:53That's interesting.
02:54Yeah.
02:55So you've been at Consolidated Analytics for eight years, almost a decade?
02:58Since 2017, yeah.
02:59Okay.
02:59And both as CRO and COO.
03:03Yeah. When I came in in 2017, it was a much smaller shop.
03:07I came in actually as the chief operating officer at that time.
03:10I think we were maybe 15 people.
03:13It was a relatively small company.
03:15Shifted the focus over to sales and marketing.
03:17The company grew exponentially, both organically and through a variety of acquisitions that we did.
03:24And just this last year, we decided, hey, it's gotten pretty big.
03:28It's over 700 people now doing a significant amount of revenue.
03:32Offices across the country and an offshore facility in Chennai, India.
03:36And, you know, felt like it was time to step back into that operating role.
03:39So step back in the chief operating officer role.
03:42And I'm focused on that today.
03:44Interesting.
03:45Okay.
03:45So how did you...
03:46I mean, I know that you've transitioned over.
03:49Wearing both of those hats for a while.
03:50How did you make that work?
03:52Well, I think I've filled roles throughout my career in both sides.
03:55I've been on the operations side in the mortgage services space.
03:58I did sales in the fashion industry and then, you know, brought that over into mortgage services at Opus.
04:04So I've done all things.
04:05I think my sales acumen has generally come from subject matter expertise.
04:10I can crack a loan file today and still underwrite the income on a, you know, on a non-QM loan or a jumbo loan, those types of things.
04:16So I think my knowledge of the details have helped my sales career and vice versa.
04:21I got you.
04:22So the last time we spoke, I wanted to read this specifically.
04:25You mentioned that the future of automation may not lie in sweeping end-to-end platforms, but in bundling fractional components, which are small, effective niche technologies.
04:36So can you give me an example of what some of those fractional components are that lenders should be focusing on today?
04:42Yeah, I mean, I think there's a number of folks that are out there, you know, promoting sweeping automation, right?
04:47But in real life, that's not necessarily how technology adoption works.
04:51I think those fractional components, as I call them, those small elements that are like sharply focused and narrow, but that you can execute with expertise, those are the things that are kind of changing the landscape.
05:02For us within, like, lone DNA, it's very simple things like document classification and data extraction and, you know, all the data normalization and all those types of things that will allow, then, agents and other things to process your data and move it throughout systems and all those things effectively.
05:19So those fractional components are the bits and pieces that, you know, really will drive the automation of the future.
05:26And at its core, it's a lot of just the data normalization.
05:30But how do you get there? Document classification, data extraction, and those types of things when you're looking at a loan post-close, anyways.
05:36Right. And so I know there's been a lot of talk for the past several years that seems to only escalate about AI disrupting the mortgage industry, but there's not a lot of talk about how that's actually working in practice.
05:49So where are you currently seeing the most measurable return on investment in AI implementation, and what do those solutions look like?
05:59Yeah, I think if you were to ask a mortgage originator, their answer might be slightly different than mine.
06:04So we're a mortgage services company.
06:05So the way that, you know, we provide services is normally with data and documents.
06:11Where we see the ROI most effectively today is what I just mentioned, which is the ability to get data, identify documents, get data from documents, and execute rules and things on those data sets.
06:23We have had an exponential lift in our productivity from, you know, going from a world where it was stare and compare, somebody's looking at a screen and then populating data on another screen and applying a compliance test to it or a credit test to it to now a lot of that data can be, you know, populated by the system, by the LLMs, if you will.
06:45Originally, it was OCR technology, right, and it's kind of antiquated, and now that we've used LLMs to identify data from documents, we're seeing a tremendous amount of lift in terms of the quality of the extraction.
06:59We still keep human in the loop.
07:00It's still totally necessary to have a human in the loop verifying the data for those instances where you get a low confidence score or something around a data element.
07:09It's still, you know, imperative that it's there, but the basis for all the testing that ultimately gets done in a due diligence capacity or an appraisal that's already been completed is the ability to get the data off the document, normalize it so that you can run tests against it.
07:25Where do you see solutions still falling short?
07:28I think anything that's, like, judgment-based or contextual in nature, any area where normalization of data hasn't taken place, I think a lot of people struggle with that, right?
07:41You want to go on this AI journey, this automation journey, but you don't have your data structured and it's not normalized.
07:46It's going to be very difficult, right?
07:48The model context protocols, these MCP layers that are coming into play now, I think that should help a lot of the vendors in the space and also the lenders, you know, kind of normalize some of the data.
07:57That's not without its challenges either, but I think that's, you know, that's the tough one is, hey, I've got to have a data environment that's standardized enough where these tools can actually work.
08:07Can you, MCP, model context?
08:09Model context protocol.
08:11It's essentially like a data normalization layer that sits between your system and your agents, if you will, or your vendors.
08:19So, it's a, you know, it's a way to translate data, like a common language, if you will, for data that's coming in and going out.
08:28Okay.
08:28So, and you told me that doc-driven workflows were basically, they were living on borrowed time and it was an inconvenient truth, but truth nonetheless.
08:36So, from your perspective, how close are we to a truly data-first mortgage experience?
08:41Yeah, I mean, I said it before, I'll say it again.
08:43I think they are living on borrowed time.
08:45You know, the mortgage industry is, for a variety of reasons, still addicted to paper, right?
08:50But those that have the edge are the ones that have, you know, standardized data and, you know, are able to, you know, push it through processes without an actual document behind it.
09:00How long?
09:01I think that's a great question.
09:03There is a investor requirement for transparency and tracking that, you know, requires documents today.
09:09There are legal ramifications and requirements for documents today.
09:14The mortgage industry is addicted to PDF documents or, you know, so how quickly it goes away, I'm not sure.
09:20There's a lot of players at the table that have to make decisions to move away from it.
09:23I think the people that are starting that process right now and have the data environment are making progress already.
09:30You can have the PDF document, but having the data standardized and normalized behind the scenes will allow you to work with the document in ways that folks that are just working with documents alone are unable to.
09:41So I know that there's a lot getting in the way.
09:42What would you say is the biggest roadblock to that data-first mortgage experience, the biggest one?
09:49A lot of it's legacy systems and processes, right, and then just the regulatory environment.
09:53The expectation is that there is going to be a document that's going to be signed and, you know, it's going to have somebody's signature,
09:59and I'm going to be able to compare that signature across all the documents in the file, and, you know, those documents are going to go to a custodian.
10:06They're going to get, you know, recorded at the local level, and, you know, we'll be able to track the paper flow from there.
10:11So systemically, a number of things have to change.
10:15I think there are, you know, there are vendors and agencies that are set up today to start working through some of those things,
10:21but we're still quite a ways off, I think.
10:23Okay.
10:25So as more lenders build or adopt AI-powered income and credit analysis tools,
10:29what advice would you give institutions deciding whether to build in-house or partner with a vendor?
10:34Yeah, that's the typical, like, build versus vibe versus blend question.
10:39You know, I would say that, you know, you should be doing the things that are important to your culture.
10:45AI shouldn't change your culture.
10:47It should amplify it.
10:48So if I'm an organization that is focused on, you know, my client experience,
10:52then I should be doing things with automation, whether it's income calculation or otherwise,
10:57that enhance the consumer's experience or the client's experience.
11:00If I am an organization that really, you know, has set myself apart by my operational workflows,
11:07then I should be utilizing AI or finding ways to utilize AI to eliminate waste and create efficiency in that process.
11:12So I think to the extent that, you know, if you're one of those organizations,
11:17does it make sense to set up a tech shop overnight?
11:21Maybe not, right?
11:22Go partner with vendors, vendors that, you know, align to your interest and vendors that understand your business
11:29and ones that ultimately can integrate with your tech stack as well.
11:32Because if they can't integrate with the tech stack, you may be building more problems than are ultimately necessary.
11:39So it's important to identify, you know, the solutions by understanding who your vendors are,
11:43what they can do, and how they can integrate with you.
11:46So how do you approach that decision internally with your team?
11:50What do those conversations look like?
11:53Yeah, I mean, for us, when we're looking at it, we're, you know, trying to figure out, you know,
11:57does it reduce friction?
11:58Can it unlock value?
12:00And can we trust it, right?
12:01Those are, you know, kind of the three factors that we're looking at when we're talking about building technology
12:07or partnering with a vendor ourselves to implement technology.
12:11Those are the critical components.
12:12We want frictionless, as frictionless as possible.
12:15We need to be able to trust it.
12:17We need our clients to be able to trust these things.
12:19So that's how we're analyzing every bit of technology that we're building or, you know, buying.
12:25One of the things that I've seen a lot of these conversations that I've been a part of
12:31or just been a fly on the wall is the, especially at the C-suite,
12:36where executives could really get whipped up into a fever to implement something for the sake of implementing it
12:43or because everybody else is doing it, because you know that it's the next step.
12:49And from any of my conversations with you, I never really, I never got that impression from you.
12:54You seem a little more reserved or measured in making that decision one way or the other.
13:01Is that something, is that you've had to train yourself to not succumb to,
13:05or is that more of a, do you naturally not, are you naturally not predisposed to wanting to jump
13:11because everybody else is doing it?
13:13Well, I mean, I think, you know, the approach to technology now is pretty iterative, right?
13:17We'll get out there. We play with things. We have kind of a sandbox environment where we go and test things.
13:22So it's not that we're not actively trying. We're actively trying all the time.
13:26We're just deploying things that are ultimately working for us.
13:29So there was an instance a few weeks ago where we had a need within one of our business lines
13:33to create a tool that would create some efficiency around some borrower screening.
13:39And we said, okay, we talked to the tech team.
13:41They said, hey, we are, you know, our project list is deep right now.
13:45This is going to have to wait a little bit. And we said, okay, well, let's go try to vibe code it, right?
13:48So we went in, we vibe coded it, and lo and behold, it actually worked, right?
13:53So we're not opposed to doing these types of things. I think we do them.
13:57We're always looking at, you know, in that instance, does it unlock value, right?
14:01So, you know, can it create efficiency for our organization?
14:04Does it enhance our customer experience?
14:06Can it affect our margins or our pricing? All of these types of things.
14:09We're weighing all those things, and we weigh them quickly.
14:11It doesn't take us a lot of time to think about, you know, our business,
14:15and we understand our business really well.
14:17We've got a tremendous number of, like, key stakeholders within our business
14:20that are wicked smart and understand the industry very well.
14:25So, you know, I think a lot of it is having the right people, you know,
14:28at the table to be able to help make those decisions.
14:30Right, and kind of following up on that, to what degree do you think it's important
14:34that AI and tech adoption reflect a company's, you know, goals, value, culture?
14:42And how do you personally assess that?
14:46Yeah, I mean, I think I would reflect back on what I just alluded to
14:50in terms of what your company DNA is, right?
14:54So if your company DNA is client-centric, then, you know, if you're high-touch,
14:59then, you know, you might need chatbots and other things to get in front of customers more often
15:04and, you know, grease the skids as it relates to communication and those types of things.
15:09I think you have to assess, you know, who you are as an organization
15:11and what you ultimately do well, and then pick the technologies that align to those goals.
15:16You can't do it all overnight, but identifying the ones right now
15:20that can add the most amount of impact, I think, is what's important.
15:22And that's a personal journey, if you will, that each company has to go on.
15:26And, you know, a smaller company is going to have much different objectives
15:30than a much larger organization.
15:32But I think in any instance, there are options for your small broker shop
15:39to your, you know, big top five lender in terms of automation technologies
15:45that they can pursue, whether it's customer-facing, whether it's servicing-related or otherwise.
15:49There's a ton out there, but it truly does depend on your identity and your capabilities.
15:55You need to know the skill set internally, make sure you pick the vendors and otherwise.
15:59That can help you out.
16:00So what would you say is your most successful tech adoption decision?
16:07Yeah, I think right now, you know, as fundamental as it is,
16:10and maybe that's why it's the most successful is because it's so fundamental,
16:13it's our lone DNA document classification and data extraction engine.
16:19We can take hundreds of different core documents and thousands of variables off those documents
16:26and extract data at an extremely high confidence rate, generally.
16:31There are tons of LLM models behind it, still some legacy OCR models for documents where it works well,
16:36where the cost is a little bit cheaper to use OCR.
16:37Those will probably turn into LLM models, but those have increased, like, our efficiency tremendously.
16:43You're getting data off the documents, getting that data standardized so that it can be manipulated
16:47in whatever way necessary to process our, you know, whatever service we're providing to a customer has been a huge lift.
16:55Things like income calculation tools that we're working on right now that are, you know,
17:00through phase one and pilot and moving into phase two, those have increased our underwriter efficiency as well.
17:07Using the first piece, though, which is the document classification data extraction,
17:12now you get those things into an income calculation engine and you're just, you know, adding accelerant to it.
17:16Right, okay.
17:18Well, looking five years down the road, what do you think is going to define success in mortgage tech?
17:24I mean, what you hear everybody talk about is just the reduce in the cost of manufacturer loan file,
17:30and I think in large part that will stand to be true.
17:34But the guys that will win are utilizing it to, the originators that will win are the guys that are going to be utilizing it
17:40to create deeper relationships with their clients, serve the customer on the customer's terms as opposed to old terms.
17:47I think we've heard some of the speakers today talk about, you know, the benefit that they've seen in terms of lift and response and closing
17:57with borrowers that are utilizing text chat versus, you know, having to call in on a phone number.
18:03I think that the reality is I know how I operate.
18:06I don't want to jump on the phone all the time.
18:08I'm pretty busy.
18:09Emails are my email bin.
18:10I walk in in the morning, I'm 300 deep already, right?
18:13Like, you know, being able to just text this information back and forth with a lender, that type of world,
18:18I think that stuff works, and I think that wins eventually.
18:21You've got to meet the borrower or you've got to meet the customer on their terms,
18:24and I think a lot of the technology is going to allow that to happen.
18:27And where do you see yourself in consolidated analytics in five years?
18:31Oh, that's a great question and not one you prepped me for.
18:33You know, I think we're going to stay with our head down, continuing to invest in technologies.
18:39We like building niche technologies that ultimately will fit a wider ecosystem.
18:44Our loan DNA platform is built with, you know, specific point tech that can increase efficiency within the process,
18:54and as you bundle these things together with APIs and otherwise, we think we can add efficiency to, you know,
19:00a mortgage originator that's originating loans or somebody that's working in the post-closing space.
19:05There are a number of gaps there.
19:06We're doing really well with the document classification, data extraction, income calculation.
19:11We've got a hedging tool that we're rolling out, a rate lock tool that we're rolling out,
19:15an agentic underwriting tool that we're rolling out.
19:18You know, all these things that are in our product stack that are soon to be released,
19:25I think are, you know, going to create efficiency and transparency and, you know, probably some,
19:32there's a quality component to it as well, right?
19:34If I can not rely always on the underwriter to make the decision but back their decision with a, you know,
19:41an agent underwriter that can see the problems in data and call it out, red flag it for my underwriter,
19:46those things are going to add value.
19:49Excellent.
19:49I've got one more question, a follow-up from earlier, which I didn't prep before either,
19:53but it's a totally interesting question because you said if you had a loan file,
19:58you could crack it open and you could underwrite and decision.
20:00I think I could, yeah.
20:01I'm still licensed as an originator.
20:03I could not hop into Encompass or Calix or anything in an originator file because it's been several years.
20:09So is that something that you maybe do from time to time whenever you're bored?
20:13Yeah, no, actually just this last week we had a deal.
20:16It was a securitization transaction.
20:19There were some questions.
20:19These were pretty unique loans, very large balance loans to business purpose entities.
20:26And there were a lot of questions around them because they were kind of uniquely structured with cascading LLCs and a variety of other things.
20:32And to get to the bottom of the answers, you have to familiarize yourself with the loan file.
20:36So, yeah, I have no problem cracking a loan file from time to time, getting in, looking at the documents, calculating what needs to be calculating,
20:43and then sometimes educating our team with, you know, I've been doing this for a long time.
20:47It's been over 20 years, I think, at this point.
20:49So when I can get in and help the team out a little bit, I do enjoy doing that.
20:52I got to say, I mean, you've given me a lot of really interesting information,
20:56but a C-suite executive still sitting down and cracking open loan file underrated is really blowing my mind.
21:01Well, like I said, when we started, I think a lot of my success, and I think if you get too detached from the actual work,
21:09it's, you know, it's hard to understand what your people are dealing with on a day-to-day basis, right?
21:12So I like it from that perspective.
21:14I also, you know, I want to keep myself sharp.
21:17You know, what's happening in mortgage today, whether it's, you know, we've seen a big product shift.
21:22There's a huge focus on DSCR right now, and DSCR used to be like a private money lending vehicle, right, or loan.
21:29And now non-QM, everybody's in the DSCR space, and we're big in that space, and HELOCs and, you know,
21:35closed-end seconds and all the things that are driving volume right now in a big part of our business.
21:39So I want to be able to sit there and, you know, have a conversation with you or a client about, you know,
21:44what's nuanced in those files.
21:45And in order to do that, I kind of got to get my hands dirty.
21:47Sure.
21:48Yeah, I think one of the biggest issues that we have in the industry is that the disconnect from the executive level
21:54and what's happening on the ground.
21:57And so if there were more executives that had their hands kind of in the mud like you were able to,
22:04that might cut down on quite a bit of that miscommunication, I believe.
22:07I'll take that as a compliment.
22:08I'm not sure the team always loves it when I start digging around, but I still do it nonetheless.
22:12Rudy, it was a pleasure.
22:13Thank you so much.
22:14Yeah, thanks, Jeff.
Be the first to comment