A.I. in Finances and Mortgages ft. Pavan Agarwal | Coffeez for Closers with Joe Shalaby Ep. 46

0:03What's up, everybody, and welcome to another episode of Coffees for Closers. Joining us today is a true pioneer in the artificial intelligence space. As CEO of Celligence International and SunWest Mortgage Company, he has not only transformed financial services through technology, but also championed equality. Under his leadership, SunWest has dramatically reduced loan decline rates for African-American applicants, aligning them with national averages, a testament to his dedication to fair lending and equality. But our guest's impact doesn't stop there. His creation, Angel AI, revolutionizes AI with the ability to make empathetic, multi-million dollar decisions at lightning speed.

0:52A first in the industry. This, combined with his engineering prowess, boasting numerous patents in AI and blockchain, sets him apart as a leader whose innovations are truly making a difference. Join me in welcoming the one, the only CEO of SunWest, Pavan Agarwal. Thank you, Pavan, for joining us today. Thank you, Joe. And it's an honor to be here. We've been friends and business partners for years. So thank you for having me on the show. Yeah, I'm really, truly blessed to have someone of your stature, a person who really has impacted the mortgage industry, a person I've been following since I entered the mortgage space.

1:37And you've really been a mentor and someone who I really aspired to follow. So I thank you for coming to the show. And we're going to make some magic happen today. So I like to start the show with a little icebreaker just to get people introduced to who you are. So the first thing I like to ask someone is, what's your morning routine? What's your daily morning regimen? That's a good question. I'm usually up around 7. and the first thing I do is I start, while I'm brushing my teeth, I'm posting, checking social media, I'm posting, you know, my thoughts for the day and also I notify my team of this is what I'm going to do today, this is what the plan is and then I get ready, I actually have a couple of small dumbbells in my office and I have a little short workout routine and the idea is just to get the heart rate up for about 10-15 minutes and that's pretty much all you need if you just do that every day you can stay in decent shape there was a time in my life when I was going to the gym hard with a trainer 2 hours, 3 hours a day doing all the heavy weight lifting and all that and then I was like, you know what?

2:59I'm never going to be Iron Man or something like that. Never going to be a bodybuilder. I kind of went off that phase and then just my trainer taught me this very simple 10-minute, 15-minute routine. It's like if you just do that and eat right, you're going to be in good shape forever. I love it. I love it. Now, you've been the CEO of SunWest Mortgage Company. How long has SunWest been around? 1980. My dad started it in 1980. And he met my dad once a couple years ago. Sweet, man. God rest his soul. Yeah. So, yeah, 1980. I've been working with him and my mom since the 70s, since we immigrated to the country.

3:40So pretty much, you know, as a typical immigrant family, everyone works together, right? And together, put food on the table somehow. Same with our company here. We're immigrants. You know, I came from Egypt. My brother works here. My sister works here. My partner, his brother, works here. So it's very similar here at eMortgage Capital as well. Yeah, yeah. And when did you become CEO of SunWest? Around 2008, 2009, right after the big short. The big short, nice. And then SunWest Mortgage Company, SunWest Bank. So there's two variations, right? No, there is a SunWest Bank. That's not us. That's a different company.

4:27Oh, okay. That's Eric and Carson. I know them well. They're nice people. He's actually running for the owner of SunWest Bank. He's actually running for Senate in Wisconsin on the Republican ticket. Oh, really? Yeah. Solid guy. Nice. Yeah, SunWest has been – we do a lot of business with SunWest. We've been working with you guys since we started the company. And I don't even know if you checked, but I think we can boast that. I think we're either number one or number two broker with SunWest in the whole country. Yeah, I believe after we spoke last, I think we're number two for eMortgage. Okay. Yeah.

5:01Yeah, yeah. You guys are our number two. I think we're your number one. I'm not sure. Probably. Yeah. Probably, yeah. But yeah, it's an honor. It's an honor. Now, I want to talk about, because everybody's talking about AI, and SunWest has been on the cusp of artificial intelligence since the before the word artificial intelligence was a trend and you know what i have a couple questions because this this is going to be a very ai centric um conversation but first off what inspired you in the very beginning to start an an empathetic AI model to help loan decision making? Well, it started from my dad because growing up, you know, since the 80s, he would tell me, look, this business is so simple, just automate it, find ways of automating it.

5:57So I've been writing code since I was a kid. And, you know, the first AI program I wrote was back in 1985. So really when no one was talking about AI other than the nerds like me. So his vision was, hey, the mortgage business, it's all paperwork. It's all calculations. There's patterns to everything. He's like, there should be a way so that this could all be done. by complete automation, right? He didn't use the word AI, but I took it to mean AI. And so after I wrote my first AI program, it was just a small little, you know, little thing. I don't even remember exactly what it did at this point now.

6:49But the, you know, the bottom line I discovered is it takes lots of data. You know, it takes lots of data, a lot of computing power. And I said, okay, we don't have the data. We don't have the ability to store the data. Back in those days, 10 megabyte hard drive cost$1 ,000, right? So I said, well, at least let me at least put an architecture in place so that we can start collecting data and start planning for this. And it always seemed like AI was just a couple of years away. So one more, the next chip from Intel or the next hard drive will solve the problem. Um, so, so because of that, cause it always seemed just around the corner, we, I laid an architecture and a foundation so that we can continue to build on this.

7:41And so that became part of our DNA, uh, of, of our engineering team since, you know, the last 40 years almost. Wow. So you guys are on the cusp of so many different technological platforms at SunWest. You also created the platform The Processor to really streamline the processing at SunWest. Talk to me a little bit about that. Yeah, so it's just another way for people to understand Angel AI. So we created another brand called The Processor. so because the way angel ai works is you you as a loan officer you don't have to literally you don't have to do anything other than build a relationship with your customer and which is really the number one responsibility of a loan officer doing like what we're doing right now this interaction and getting looking into each other's eyes and understanding where we're coming from.

8:42So if a loan officer can build that relationship and build that bond with his customer, why should he be so stressed out over the paperwork, over the details, right? So take, you know, Joe, I take your loan application. I say, give me your W-2 pay stubs, bank statements, you know, whatever, and you pull those out, you give it to me. I take my phone, take a picture or scan it on my Xerox machine or whatever, drop it into Angel AI, and then I'm done. Yeah. More or less. Angel AI will pull the rest of the missing documents. Yes. Angel AI will automatically read the documents and scrape the data off it.

9:26Complete the 1003. Okay. Okay, and if there are any, and it would analyze documents, let's say the bank statement had a large deposit, then it would automatically tag that we need an explanation for this large deposit or source of funds. Or if there was, if the pay stub showed, you know, variation in overtime, then it's going to ask for a BOE or ask for more information on that. um so it so and it looks at every single data point on the on the documents and and because of that it's it also catches fraud right because it redoes all the calculations so if someone tried to make up a w-2 or make up a make up a pay stub right and you know numbers won't tie together properly okay so so it's very good at just catching compliance and and fraud and and it also figures out what's missing what's needed and automatically so literally it underwrites on the spot automatically adds all the conditions and then after that we have a team that we call angel east us so it's a large team is a massive call center of bilingual agents that are will call your your customer and say, okay, you gave me all your documents.

10:47Thank you. But the April bank statement had a large deposit of$50 ,000. Can you please explain where that's coming from? Can you send us the backup information? So they're reaching out and having that kind of human conversation. So who calls them? These are, we have a call center in Puerto Rico, so they're all bilingual. Oh, nice. Okay. And they make the call. They follow the instructions that the AI has generated. And they make the call and they say and explain to the borrower, this is what is missing. So each account, so each loan officer, basically, whether it's through our wholesale channel, like with your loan officers, who gets assigned one dedicated angelista.

11:35Okay, and so they then reach out to the customer or if they need to reach out to the realtor or the title company, whoever, whatever document is needed, they reach out and they explain to that person. They follow, they basically, they're just, they're kind of like, I hate to say this, they're kind of like robots. The AI tells them what to do. They follow the instructions. They've been coached and trained really well. I don't know if you know Bill Pipes. No. He's like the, I'd argue the number one. Some people would say he's number three behind Tom Ferry. In my books, he's number one. The number one real estate coach in the country.

12:16And he's very good at coaching customer service teams. So I hired him to coach the Angelistas. us so he coached so he's coached them and he does a regular he coaches them every week on on how to handle objections how to communicate so they're very very good on the phone um with the with the customer they they've been coached to have high high level of emotional iq customer emotional iq and so they're able to because they don't have to think about what's being asked and they have to just focus on delivering the message because the thinking of what's being asked and why it's needed has been done by the ai and the ai gives them detailed clear instructions on what what they need and why they need it wow okay so all they have to do is focus on understanding the customer building a relationship with the customer and making sure the message is communicated across that just saves so much time so much time for the lo's it's uh it's truly like an lo assistant It is a complete LO assistant.

13:21The processing, you know, the processor is the AI because within the hour. So if you as an LO take the application, right? And by taking the application, you don't even fill in a 1003. You just give me the documents. The 1003 gets filled from the documents. Okay. Wow. So it completes your 1003. It completes the 10-3 for you. I don't, because me as an L.O., when I was an L.O. back in the 80s and 90s, right, I hated completing the 10-3. I never did it, right? I would talk, I would interview the borrower. The first question is, so what do you want out of this transaction, right? What do you hope to achieve, right?

14:03And just build a relationship, understand what their goals are and why they're buying this house and so forth and understand their background. And based on that, I would ask them then for the documents, I need these documents. Okay, and I looked at the documents, okay, they're complete or relatively complete, and then my loan application was done. And I would give it to an assistant, a processor in the back of the day, and I'd say, okay, everything you need is in these documents. Go complete a 1003 and go get the disclosures out. I'm not going to waste my time. Complete a 1003 myself. There's no point.

14:40So now, instead of giving it to a human being to read documents and complete a 1003. We give it to a computer that reads the documents, completes the 1003, and, you know, its accuracy is going to be 100%, right? A human being can make a mistake. But these OCR systems and the cross-data validation stuff, it's super accurate. So within the hour of you taking those documents, dropping the AI, within the hour, the 1003 is completed, all the verifications are done, work number, you know, whatever you know um the 401 lookups all that happens automatically so it's integrated with with all the systems out there and and your uh disclosures are sent automatically to the to the customer so everything that that what normally an LOA or processor does is done within the first hour it's awesome yeah it's it's it's a great um time saver um yeah and when you know and the thing is like people get so confused by it because they're like, why is it free?

15:48Right. Because we have, we have, and we were talking about this earlier. And before we enter this, this, this, this podcast, like you were telling me you have, you have CEOs from competing mortgage brokers, right? I mean, competing, not really competing, but from other mortgage brokers that, that compete with eMortgage in the podcast with you. And we were just talking about there are no competitors in this business. It should all be collegial and work together. The market is so huge. So same thing. I make the angel AI completely free and accessible. To every company. To every company. Even whether you work for cross-country or fairway or freedom or whatever, or movement, it doesn't matter to me.

16:30If you're a loan officer, you can use it. Are the angel-ices free too? or that's only if the loan comes to SunWest if the loan is submitted to SunWest then the angelistas take over but there's something similar to angelistas which we can talk about in a minute but the point is so we have loan officers from other mortgage banks who would never submit a loan to SunWest and that's okay they use it non-stop because they have literally stopped reviewing income and assets anymore. So they take an application, right? And they'll just drop it into Angel and let Angel tell it what the qualifying income is and what is any problems or what the assets are.

17:19And their loan is basically set up for that. So I've got LOs calling me and thanking me, saying like, you save me hours every day. and I don't have to, you know, given where margins are and things are so compressed, like I was, I need another assistant, but I can't because I don't have the, I don't have the money to do it, money to do it. And now I've got Angel and, and it's a free assistant. Yes. Yes. This is free assistant. It's the best assistant I ever had because it's accurate every time. Exactly. Exactly. Yeah. LOs, make sure you get a free assistant with angel ai um so what are some of the biggest challenges that you face bringing angel ai to life oh uh she's i mean it's decade plus of work um so what ai is all about training right and and and so we've been running you know loans through it constantly non-stop and and it it learns a little bit of time and then it it it can make a mistake so in the early days it would make a mistake and then a human being has to get involved and deal with it, right?

18:31Just before we had the warranty. So just all that constant learning and adjusting and tuning, right, creates this, yeah, it's a big challenge. There's a lot of work to do. And that means that our mortgage process, right? So the way SunWest is organized, it's not organized like any other company because it's all centered around the AI. So our, you know, our people never worked in, I would say not never, but for a long time have not worked in a traditional way that you would normally expect processors and underwriters to function. So they're always supporting the AI. And if the AI kicked out an exception, it did something correctly.

19:13And then their job was to go in and correct it. So we had a very non-traditional workflow for the last 10, 15 years. um so so the challenge has been just just running enough scenarios through so since uh you know 2018 when we you know officially launched angel ai um and and a year later in 2019 we we we put in the warranty right it's where we said okay i've done enough transactions through this we've tested this so much that it's like, I'll, I'll say that if it got something wrong, I'll eat that loan. Okay. That means, that means if it got it wrong and let's say, um, you got to do you approval when it shouldn't have been approved because the income was wrong or whatever, right?

20:02I'll fund it anyways and I'll balance sheet it. Okay. So, so since, so since 2019 to today, I've had to balance sheet about four loans. Four? Yeah. That's it? That's it. That's an amazing precision. Yeah. Yes. Yes. And we've, since 2018 to today, we've been down over 200 ,000 transactions through Angel AI. So every deal that SunWest does is ran through Angel AI. Correct. So how much of a cost savings has that been for SunWest? Yeah, it's massive. I mean, I think MBA put out something recently, the average mortgage banker spends about$13 ,000 per closed loan. And I think that includes commissions.

20:43So let's just say half of that is commissions. So that's$6 ,000,$7 ,000 is the operating cost for an average mortgage banker. Okay, per loan. And you know what the margins are these days. Your operating costs, you're already underwater. Forget the commissions, right? Yeah. Right. So, you know, three years ago, we were at about$300. a closed loan and we've made it more efficient than that so we're under 200 probably i haven't checked lately but my guess would be between 100 and 150 a closed loan so it's all it costs you to close the loan now yeah right you should be making more money with some less uh yeah so the you know of course then you have the um out of pockets like credit reports and stuff yeah you know i'm not including them credit report these days is a hundred dollars alone or something.

21:33Fannie reissues, this, that, and the other. Right. Right. So they're cheap. Well, are they more? A hundred dollars. Oh, they're like 60 to 80. Oh, whatever it is. Yeah. And it's not a hundred yet. Stop adding inflation to it. We're not there yet. It might be by the end of the year, but 20 % inflation rates, maybe. Um, so you've had such a vast background in so many different, you know, so many different professions you, you've been like, expert engineer, mortgage professional, CEO of one of the largest mortgage bankers in the country. With all that experience, what do you think some of the needed traits are for a mortgage CEO or a tech CEO in this environment, in this rapidly changing environment?

22:22You know, that's a good question. I mean, we were talking about that again, another thing we were talking about before the podcast started, which is our kids, right? And, you know, we're talking about how kids these days are making so much money because they're on YouTube and whatever. And my son, 10 years old, and when he was nine, he went through this YouTube phase and he made some money because he got some viral videos. And at 10, he grew out of that phase. And now all he does is soccer and track and Ninja Warrior. Just sports. And that's it. That's all he could think about. And I would rather he's doing sports all day long than being a YouTube star.

23:00okay and because when you're doing sports especially uh competitive sports like soccer right you are learning the real lessons of life which is the real character right yeah when you get hit right and you lose or you win and knowing how to win gracefully and how to lose gracefully how to take a hit right and how to hit back and knowing what to hit back you know that understanding the boundaries of life, understanding the boundaries of interpersonal relationships, right? That you get with living your life and interacting with people, like not virtually interacting, not in the metaverse, but real interacting.

23:43Yeah. And ultimately, all of life and all success boils down to your ability to build relationships and create connections, right? And if you're fundamentally good at that, and if you're fundamentally good at working with the culture and the people around you and understanding cultures, especially, you know, I'm fortunate enough to be an immigrant where I have two cultures. I have my Indian culture and my American culture. And so, and I think that's an advantage to my kids because now they grew up in a multicultural environment. So they're able to understand that not everyone is the same, not everyone thinks the same way.

24:27And similarly, you grew up in multiple cultures as well. So when you have that kind of core understanding, then if you want to be a YouTube star or if you want to be a top loan originator or a realtor or a doctor or engineer or whatever, you understand, you have a better, deeper understanding of the nature of your customer. there's always a customer and if you don't understand the customer you're never going to be able to deliver a product that means something to them so you're how are you teaching your kids in this environment how do you teach them to understand that that philosophy outside of sports because you've got your one son he's in sports I think it's the kids learn the most by watching their parents so the most important thing is what i think the biggest lesson that the kids learn is to see how mom and dad treat each other so if the way mom and dad treat each other in front of them is the number one lesson and and if you start with that and you focus on the proper how to respect each other properly right that 99 of what they need to learn they've already learned right there so we we start from that level and then and then ultimately you know uh just you know like i work a lot i i uh you know it's you know what it takes to run run and build multiple businesses and i get them involved i show them what is dad doing right why is dad traveling so much and i'd like to take them with me wherever i travel as much as i can that's awesome right so let them see firsthand let them have these experiences and see that you know money just doesn't fall out of the the sky that you gotta work for that you and these are things you do and let them observe right let them be a part of your life um and automatically they will they will learn the right values yeah because something me and you have the the blessing of is like we grew up probably poor because we were immigrated yeah you know and uh one thing that our kids don't have the blessing of is they they're not poor so how do we make sure that they have that balance you know to to continue to maintain that same grit that we have yeah and that's something that you know i'm always like asking everybody everyone who comes on the show who has good kids you know well-raised kids you know it's a it's something that is a my kids are younger they're all 10 8 3 2 so i want to make sure you know you just that's my wife likes to look at our kids like she's like that's my life's work like that that's what i do for a living it's like that's them i'm like i do this and i'm good at this and she does that you know so so any and so she watches all the shows because she knows i always i always hit you with parenting advice

27:32um so uh so so that's awesome that you take your kids with you on you know to work travel you know they see you grinding they see they see you making it happen um what are some like what are some of the struggles that your family faced immigrating from india Your dad had obviously a much greater barrier with language like my dad. So what are some of the challenges when he started SunWest? What sort of issues did he face and how did he overcome them?

28:11Maybe because I was just a kid back then. I never saw the challenges. I just saw it as fun, as excitement. it. And, you know, like coming from the poverty we came from in India to America, right? And the little things you take for granted when you come to America, like Americans take for granted, which was like luxury for us. So as far as I was concerned, we're kings, you know, like, this is great. One bedroom apartment in the hood. But it was like, wow, you know, it's amazing. This is amazing, right? So kind of having that childlike wonderment over how good life is, I didn't see the challenge.

28:59Everything was always positive. Yeah. And I think that's probably one of the reasons why I'm an entrepreneur is because still, I like to keep that same sense of wonderment and positive. Positive mindset. Mindset about, you know, everything is so wonderful. Yeah. Right. And certainly when I think back and look at where we were when I was little to now, I was like, well, I lived through that. It doesn't matter what comes tomorrow. Whatever God is going to hit me with tomorrow doesn't really matter because I've already lived through that. Yeah. I was talking to someone else on the podcast the other day and said, as a CEO, I could take a big beating.

29:44Just like you could hit me with anything. and I'll still smile at you. You just can't hurt me. It doesn't matter what comes at us, but we're able to overcome tribulation. India, all very poor. I'm from Cairo, Egypt, but I'm from this place called Helmet Zetun, which is the poorest part of Cairo, because now Cairo, there's a lot of rich parts. Egypt's considered a third world country, as is India, but there's certain parts of India and Egypt that should be considered six world countries. They're like multiplied, you know, like exponentially worse. You know, like I'm from those parts where it's like really, really poor.

30:24Like you're on the seventh floor, no elevator. And a little one bedroom, you know, that one bedrooms are like half the size of the one bedrooms here. They're like 200 square feet, right? And no air conditioning, right? So they're really poor. So we're really, now it's like, It's just crazy to live in the environments that we live in now. Have you ever taken your kids back to India? Yeah, many times. But where you grew up? Not where I grew up. My wife grew up in Mumbai. Actually, we were just there last year over Christmas. So they get to see their grandparents and cousins and so forth. Nice. What's their interpretation of how the lifestyle is there?

31:11for them they don't see any difference right because my wife's come from from a wealthy family over there and um and you know it's like and you know remember when i i left india and in the 70s right when it was basically a socialist um country under you know almost soviet occupation you could say uh wasn't exactly uh behind the iron curtain so So between then and now, I mean, it's a different country. Now it's opened up. It's a free economy and it's booming. Love it. I want to get back to the AI and ask you this question because this has been a big debate in the AI world. And it's really like, what's your opinion on the balancing of innovation and ethical considerations, especially like in high stake environments, like the mortgage environment where they have to make big decisions and you're balancing empathy with it as well?

32:16Correct. So I'll get back to the empathy. That's a big topic on to itself. But as far as the ethics of making these big decisions, so I think it was the Federal Reserve last year, the FDIC sat a circular about almost exactly a year ago. That said, it's okay to use AI to make lending decisions or banking decisions or credit decisions. as long as you, but you cannot use, you cannot use that, oh, this is the AI's decision as the excuse. Okay, so you can use the AI to make a decision, but then you also have to be able to explain how the AI made that decision. Okay, so you have to be able to connect the dots.

33:08And there has to be a human doing that. No, it doesn't. They didn't say it has to be human. They just said the AI needs to be spit out kind of like a receipt of how I came to this decision. And so that's a, you know, traditional neural network is very hard for a neural network. I mean, it's impossible. You can't. The way backpropagation training works is you just you can't explain how it made the decision. It just it learned. It sort of becomes intuitive and it says, OK, you know, turn right or turn left. It makes that decision. If you ask it, how did you make the decision? We'll say, here are my weights.

33:43And that doesn't mean anything. It's not understandable. So if you, to basically, so they put the bar, by putting the bar there, that you have to be able to explain how you got to that decision. So if the AI says, I'm going to approve Joe's loan, it has to also explain why it approved Joe's loan. Okay, that's a high bar. Okay, and it's hard to do. okay so the way we build angel is and and it's not it's not built off off the shelf tools like large language models and i'm throwing out terms which i don't want to confuse the audience but essentially that's the technology that's behind chat gpt so it's not built on tools like that and and it's a custom built neural network that we built from the ground up so because we were building it for mortgages and financial decisions, for us to test it that is doing it right, we had to build it from scratch to be able to spit out the reasoning for its decisions along the way.

34:51Okay, so when the FDIC came out with that rule, we're like, okay, no problem. We already had that covered. Otherwise, there was no way for us to test our own product. We had to write the code so that it generated the logs automatically. That's awesome. So the ethical standard is, at least according to the FDIC, that if you make a decision, you better be able to explain why you made that decision. You better be able to explain if you said that the credit is unsatisfactory, you better be able to explain why is it unsatisfactory. If a human underwriter underwrites a loan file, obviously they can explain why, and their notes would would have some of that written down depending on how good notes the underwriter keeps but you could always question the underwriter and they'll explain why okay but uh out of the box ais can't do that unless you build it to to do it okay so that's how we know that it it delivers the right decisions that's how we we audited ourselves right i mean apart from the 200 ,000 loan files that we have put through, real loan files that we put through, we run test cases, tens of thousands of test cases every week through it.