Podcast EP 68: Mortgage Market Outlook, AI Automation & the Future of Lending — Shashank Shekhar

[00:00:00] Speaker 1: I'm excited to welcome Shashank Shekhar, who is the CEO of InstaMortgage. Shashank is also a founder and CEO of InstaAI, InstaMortgage, InstaAI. All right, so that must be relevant. Focusing on Gen AI technologies to improve efficiency in sales and marketing for mortgage industry. Amazon number one bestselling author. Shashank was named 2022, 2023, 2024, three times in a row. Entrepreneur of the Year by Stevie Awards. Expert guest on TV and radio show Mortgage Matters. The author of widely acclaimed books, First Time Home Buying 101. I would probably buy that or recommend that for some people. Real Estate Unleashed. And latest, you have to tell me when did you publish the latest one? I knew about two. My First Home. So, Shashank, welcome to the podcast. [00:01:09] Shashank Shekhar: Thank you. And actually, that's not updated because my latest book is actually Let AI Be Your Coach. It's an international bestseller that got out in the market about four months back. So, I guess I have not updated my resume online. [00:01:23] Speaker 1: Let us touch on that during our podcast because, you know, it's going to be focused around AI. But before we jump into AI topic, specifically for the mortgages, because we both are elements or Lego blocks of the mortgage industry. And we'll probably see things similarly, but from different sides. I'm more from the vendor side and process side and you're more from the mortgage origination side. Let's start with, you know, we are past the quarter point increase in the Fed rates and the market is in shock. Everybody expected in the beginning of the year rates to drop. I just listened to Stewart's and First American Fidelity quarterly announcements. They're saying that the market will be pretty much frozen for the next two years. Survival mode. People are close to being, you know, giving up on the mortgage market. That's going to be bloodbath and very tough atmosphere for LOs, brokers, mortgage lenders. The home prices are falling in a lot of the areas, especially in Florida. The number of homes will be less than 4 million. That's all. That's the projections, right? Of everybody's sort of referring to next year. And when I visited the MBA annual conference, people were somewhat crying. Uh, saying that, um, it's a survival mode. So I wanted to pause here, so saying, do you see the same thing? It's, it's a gloomy atmosphere or, you know, you see something different? [00:03:10] Shashank Shekhar: Uh, it is a gloomy atmosphere. I'm not, you know me from before, Alex. I don't, uh, uh, I'm not a motivational speaker. I'm not here to tell you what is not true. What that means is that what some of the numbers that you mentioned is indeed correct. Uh, I mean, NAR pending home sales hit a record low in September. It was down 7.3% of the month, 11% on the year. Uh, Fannie Mae has raised its rate forecast to almost 7.4, 7.5 across all of 2027. And you're right. You're back from MBA, which is projecting almost 9% drop in mortgage origination next year. Uh, and even all actually all the way to through 2028. And if you are in, we were in the MBA and people were crying, it's because the, uh, the, the, the product, uh, the, uh, the margin on, on origination has dropped to what? 25 basis point from almost 90 that it was in, in say at the, just after COVID. So all these, all these numbers, uh, are indeed accurate projection. It's something that I've always taken with a bit of, bit of a salt, so to say is because it's one of the most difficult industries to make any projections on. There are so many factors that govern mortgage and real estate industry is that one thing that I will, I will not do is make predictions on it. And, and we have seen some of the biggest experts and economics get it wrong, not because they're bad at what they do is just because there are so many moving pieces to project, uh, mortgage numbers and real estate numbers is as, is that you could be the fed chairman of the board and you still don't have all the data point to project even the, even interest rates, something you're in charge of. Let alone everything else. So, uh, whether it's gloomy and, and it's, it's all doomed for the next two years or not is because again, there's no way to know that. I mean, just before Iran war broke out, we actually fell below 6% on mortgage rates and we were looking really good for 2026. If you asked economists on that day, uh, or the, or the expert predictor, so to say, they would be saying we're going to have finally a turnaround year for mortgage in real estate. We all know how that panned out. So because we are impacted by so much of geopolitical and macroeconomic factors that nobody knows, nobody can predict. I wouldn't go into the fact that all through 2028, it will be bad. Not because I know that it won't be bad because I know that there's no way anyone can with any amount of certainty predict that because all the predictions are based on today's numbers. [00:05:38] Speaker 1: Um, I hear you at the same time, um, our mutual acquaintance, Chris Whalen is now shorting all the banks. So he probably hears something on the market that, that, uh, makes him believe that the forward projection really, you know, not, not great. It is. And at the same time, uh, the builders, they're forecasting terrible year. It is. So, uh, as, as, um, uh, one of the CEOs told me this morning, builders forecast, forecasts suck. So, so that's, that's his summary of, uh, of the next year. Yeah. [00:06:14] Shashank Shekhar: Both the consumer sentiment and home builder sentiment have, have fallen to multi-year lows. So, yeah, I mean, we're not looking good. All I'm saying is that, uh, there is, uh, again, I'm not doubting any of the projection. All I was trying to say is that there's no way to know for a hundred percent sure that what you're saying is, is accurate. And I could have said the same thing if, if everybody said that, uh, like next two years, we'll see unprecedented growth or something. Uh, even then I would have said the same thing is, how do you know? Uh, is because for example, nobody knew there, there, there is going to be this war that, that happened in April. So, um, I mean, we've brought everything down to, uh, with, uh, with it, so to say is that similarly, we can't see these things being in the mortgage and real estate industry. You don't know what, uh, what governments are trying to do. And, and we are not just impacted by what they say the U S governments do, what say Chinese government do, what, what say a Russian government does. All of that impacts mortgage and real estate, which is crazy to be an industry, which is, which is impacted so much by outside forces. Uh, even foreign forces, um, I mean, this morning, for example, we had had slightly better interest rates because what's happening in France. So I'll take that for a change, but, but what I'm saying is that, yeah, the impact is from all over the place and it's, it's a little hard to predict, uh, in a market, but everything that we are looking at, it does point to, uh, a stable, to a declining and a stable in a, in a way bad stable, because we are not in a good market, stable to a declining market over next two years. [00:07:41] Speaker 1: Do you see any, uh, bright spots? [00:07:45] Shashank Shekhar: Oh, the bright spot. One thing that I jokingly tell people is that the higher the, the rate remains for longer, the longer the time, the bigger the refinance boom will be. But when will that be something that we don't know? So, so that's why bright spot is the fact that, um, yeah, I mean, we are just, I mean, people, I mean, we are still, even in a bad year, we're still selling say 3.8 million homes, whatever we will end up selling at. Uh, 70% of that almost gets financed, 65, 70% of that. And if you're selling that at 7% today at someday, they will refinance. It's just that, that someday we have been hoping that will come for four years now and we're still waiting. [00:08:24] Speaker 1: So, so therefore, um, a couple of things to mention if, if the economy will be sour, certainly capital markets should be a bright spot with a lot of trades, uh, folks that are participating in that, um, will benefit. I just came back from ABS East conference and I heard that at least four large banks are putting together the new trade desks to acquire, um, um, not just distressed, but, uh, everything that's moving, uh, on the market today, uh, such as DSCR, non-QM, uh, HELOCs and closing seconds. So those still probably will be, um, you know, a bright spot in my mind just because nobody wants to trade, nobody wants to refinance a 3% mortgage. [00:09:08] Shashank Shekhar: - Right. - Yeah. - Yeah, that's true. And M&A market is extremely active in both the mortgage and real estate. We have seen some of the trades happening already. And if anything, it's probably just the, just the beginning of the, of the terminal effect. So to say, I mean, congratulations to you, your, uh, your couple of companies got acquired. InstaMortgage recently closed, uh, its own acquisition or being acquired by a NASDAQ listed public company. So, uh, we are both in the, in, in a similar boat in a, in a good way, but, but yeah, that's, that industry will be, will be very active in the coming months. [00:09:41] Speaker 1: - And I think the opportunities are there. Uh, a lot of people will just give up in my mind and, and, you know, they'll be ready to be acquired or, uh, they'll just, you know, there's, there's a sense of uncertainty and that always creates an opportunity. - Yes, it does. - So, um, coming back to how to survive in this market, right? - Yes. - Before we get to your book, um, you know, this is a good lead into the AI. Right? So if you didn't play with AI, you maintain the same staff, you grow this salary, uh, every year and your cost to manufacture a loan is really high. And when we met just a quick side story at the executive round table, you were the presenter to talk about what your company was able to achieve in AI. I was blown away. I was impressed. You said all the right things. - Thank you. - That, yeah, uh, that, that one entrepreneur would need to do to build the most efficient flow to manufacturing the loan. So, so you deploy the AI in the right spots, in the right places. You put the humans where you have to, um, and also I, I actually, I thought that we spoke about manufacturing. I didn't realize that you also put a front end on the Insta AI to, to, uh, increase the sales. So let's start the conversation with MBA tells us rights on the same conference. It, it's $12,000 to manufacture a loan, including sales and marketing. So how do you go from that number down to survive on today's market? [00:11:16] Shashank Shekhar: - Yeah. And that, that has been extremely sticky and stubborn. If you look at it, that number hasn't moved for, for almost a decade, other than say a few quarters here and there, of course, during COVID we produced so many loans. So it wasn't that the, the, the cost per loan was actually going down. It's just that we only had a fixed number of a staff. You couldn't hire more because there were no more people available and we closed almost two X of what we, what we usually close. So of course, just by sheer arithmetic and math, your per cost, per loan cost went down. And so that was temporary and, and that was, that was understood. But yeah, that number has remained above, above 10,000, uh, for, for a long period of time. And what is surprising is that if you, if you go back to say 1990s, early 2000, uh, there was so much of manual work in, in mortgages, right? I mean, you had to actually, um, say, put the documents together and, um, the, uh, you had to ship the documents, that stuff. And then when the origination cost was so much, it made sense, right? Everything was manual. The other fax stuff, all that. And there's so much of technology, right? We have point of sale. We have loan origination system. We have e-signing. We have the disclosure squad automatically. All of that put together, you will think all of that mortgage technology will, if not, nothing else will bring down the cost of origination, the cost of production by at least 30%, right? 25, 30, just commonsensical. I'm not even saying if I was not even from the industry, it was from outside. Like all these manual stuff that you guys were doing, everything is, is practically online now. So how does, how does that not reduce the cost? I think there's a lot of inefficiencies in the system, uh, which, which causes, uh, which causes this, this cost to remain stubborn as it, as it is, is that we have, we have taken technology, but we have not used the technology in terms of the way we have, right? There is, there's an automation that comes with the technology. And then there is, uh, there's AI and then there's IEA, right? Intelligent automation. We, we achieved a lot of, uh, operational efficiencies, not through using any advanced technology in the beginning, because of course there was none before AI came in. It was mostly through intelligent automation. It was just figuring out what can we automate. It wasn't like any serious technology work. We were not, I mean, our pockets were not that deep that we will go and spend say $10 million into building technology or something. But a lot of it comes from understanding your operations, building that, that intelligent automation, uh, so that the same human can achieve more, uh, or can do more from, from what they can, of course, in the world of AI that has, that has completely changed, especially in 2026. You heard me speak last year when AI has not even, uh, I would say reached the stage where it was, uh, operational at a scale that, that where it could be. Uh, this year has just seen tremendous, uh, evolution of what AI is capable of. And at the speed that's moving, uh, I'll be very surprised that, that, uh, by this time next year, some of the more forward companies in terms of AI adoption are not seeing 30, 40% cost reduction. I mean, I was, uh, I was speaking another, another conference and somebody said that, I mean, do you foresee a 15% cost reduction? And I, I, and I actually said that 15% would be for losers. We should be seeing 50% cost reduction because. Now cost reduction means employee reduction. Uh, it depends. If the volume remains where it is, then it does, does mean employee reduction. But if the volume goes up, then of course you can, you can do more volume with the same number of people, but, but I'm again, just like before, I'm not here to mince my, my words here or in the sense that, uh, uh, like some of the CEOs will tell you is that, oh, AI does not mean cost reduction, yet we look everywhere. And, and you have some of the, some of the world's biggest companies cutting, cutting people. They won't say that on the mic probably. [00:15:21] Speaker 1: So yeah, I, I'm, I, when AI boom started and that was a few, three years ago, let's say, um, you know, I was a strong believer that AI doesn't eliminate the jobs. It enhances the job. I'm a strong believer now that AI eliminates jobs. That's it. It's, it's, it's reality, uh, any clerical, any checklist, any validation, uh, without jobs, without any subject matter expertise that covers that 1% that AI cannot do or doesn't know how to do. So that's the, you need that human. But if you don't carry any subject matter expertise that you've been in the field for 20 years and that corner case pops up one out of thousand times, but I need you to be that subject matter expertise. When it pops up, that's the need for those people. But if you are just carrying the paper from one side to another, your job is, is at risk. [00:16:23] Shashank Shekhar: Yeah. So this time last year, Alex, I was, I was of the belief that if anything, AI does, uh, does enhance, uh, enhance the work that, that you're doing. Um, because what we built, for example, at Insta AI, it was to improve the productivity of loan officers. How, how do you create content quickly? How do you get coached on your, on your sales stuff quickly? How do you get your guidelines in seconds instead of researching it for minutes? It was more, it was built for productivity gain. It was not built for anything else because it was not capable of doing anything else. It was not capable of, for example, to replace loan officers and, and, and do all of that stuff. But that's, that's why I was saying the, what I've seen in the AI world in the last seven months, I would say February onwards, when Claude, Claude launched, uh, launched co-work and everything that has happened after that. I'm increasingly of the belief that it's going to be very hard not to replace a job with, with what they're becoming capable of. And so if you, if you want to survive in the industry, unless it requires a physical human to do physical stuff, like, like real estate agents are probably, they have to be in the local market showing you homes and all of, even though technology exists, a lot of people want to go with a real estate agent, look around the house, all of that stuff. But on the Mogad side, it's crazier. I was, I was having conversation with, with, uh, with my board yesterday and I was, that's what I was saying is that, look, a loan officer could be sitting in Seattle and can do the loan from anywhere. So one loan office, if you think about it, Alex, look, if most of the work that loan officer does can be done in the back end through AI agents. So what do you do as a loan officer, right? Let's, let's review it. Let's, let's skip the client acquisition part. So let's, let's say if you go and meet real estate agents, have coffee with them, lunch with them, that's not replaceable. That's something you're doing, but everything that happens during production of the loan, right? A loan application comes in, who takes the loan application? The point of sale does, right? Point of sale system. All the document uploaded, uploads through, happens through point of sale. And now you can build an AI layer, AI layer on it. So you don't have to review it. It can review if a certain pay stub is missing. If it can review, if a bank statement is missing, it can tell you what's there, what's not. And it can do an intelligent analysis while you are uploading the document to figure out, say what the processing of the document looks like. Then once it has evaluated it, it can throw in multiple interest rate options to you, can explain to you what's the difference between 30 and 15 or a fixed in arm. All of that stuff goes into production. It can draft an email on your behalf. Hey, Mr. and Mrs. Borrower, I just reviewed your documents. These five things are missing. And by the way, looking at the first level of underwriting that I've done, looks like you might have a problem and might need some gift funds, right? So, and it can do all the way until closing and can run an email campaign after post closing to stay in touch with you. So if you think about it, a loan officer only has to bring in the client to the door and a hundred of agent can be, so one loan officer could be closing a hundred deals simultaneously without actually working on any one of those until a problem arises. And then you have to pick up the phone and really talk to this borrower. So, and that's, is that an efficiency gain? Yes. But if I'm doing the job of 50 loan officers, then where, what are the other 49 loan officers doing unless the business increases by that much? And you and I both kind of established that right at the beginning of this podcast is that chances of that happening though is not zero. It is, it is very close to zero. And even if it increases, it's not going to increase two or three or four fold that we saw during COVID. If it increases, it will be 5%, 6%, 7%, not good enough from that perspective. So that's how I see it as that elimination of job from AI is, first of all, it's already happening. And we are not talking future. Now we are talking actually past, not even present. And two, it's inevitable. It's, it's just something that's, I mean, the people who say that, that we are just going to improve the efficiency and productivity and get more out of people. But how do you get so much more business to get more out of so many people that you have? [00:20:23] Speaker 1: I agree with you. And this is a fairytale when people start or, or maybe it's, it's a nice way of describing job loss or job elimination. Let's, let's touch on the places where you would insert humans, right? You, we talked about sales marketing, but in loan manufacturing, are you just having one human per loan clicking a button to fund the transaction? Or are there more jobs related to checklists, underwriting, e-signing? Um, obviously you need e-notary, right? You need an notary to babysit the transaction. I mean, what, what, what are the places that you foresee you will insert the human, because without the human AI can do some nasty things like wire more money or execute on behalf [00:21:08] Shashank Shekhar: of the borrower. Who knows? Yeah. Yeah. So in, in, in six to nine months, so we have AI in, in practically all the parts of the operations right now, uh, but not to the point where we have eliminated the entire department. So it's not as if we have no processors or as if we have no underwriters and definitely we are funders because as you said, we are not giving AI access to the bank account and I know I have the funds for, uh, for funding a loan or something, uh, or post closing. It's just that the, the efficiencies are so much more, but what we have started looking at as we are building the new layer of technology, which we are, uh, planning to, uh, really sometime in February or March, depending on how soon we, we clear it is that we are, uh, we are re-looking at every single thing that we do now from an AI first mindset, which is a mindset switch first and technology switch later. Um, and so we, uh, we just met with our, uh, developers, uh, just this last week. And that's my directive to them was that we need to re-look at every single thing that we do from this mindset of can the AI agent achieve this? And only if you're able to prove to me otherwise that we have done everything we can and AI agents can do it, then we will have the human in the loop from that perspective. Of course, anything that requires licensing, like a loan officer has to sign off on a 1003, uh, as a human, not as an AI or whatever, of course we will, we will adhere to all of that stuff. So we are, we have a completely different mindset towards it. And I would not have had that mindset again. I will be very honest with you sitting with you say this time last year is because I hadn't seen that the day I had evolved to a point where we can go in with that mindset, but we are going into 2027, uh, with that instruction to our engineering team saying, that's the how, that's the way we want to re-architect the entire operations and come March, let's see how much we can achieve. Uh, we, we are not going in with, I mean, we are not delusional here saying that a hundred percent of it will be achieved by, by AI by March, but just changing that mindset make us question everything that we are doing from a human perspective. Thankfully, we have a extremely experienced human team. So if anything, they can contribute towards this entire bill that we are doing. And we grew 60% year on year last year. We have, we have had a decent growth this year. So thankfully we don't have to eliminate people as the, as the growth is funding, or growth is taking care of the operational efficiencies we are getting. [00:23:34] Speaker 1: I think people should really use your business model as a template, but I want to double click on, on two things that you said. Uh, so, so, um, as you're rolling out AI, I, I, you with, with how you positioned it, uh, are you worried about talking maxing and, and really overspending on AI, doubling, tripling the bill just on the experimentation stage, because that I thought that was a big problem in the middle of this year, right? So people just said, have fun at it, uh, prove to me, like you said, or do whatever you can to, to create the product related to AI and then see if it works. And that caused a lot of the CEOs to roll back and say, Ooh, wait, my bill is just tripled. Even though it doesn't hurt me, it's still, you know, goes to my bottom line. Let's be more cognizant about how we approach this. [00:24:27] Shashank Shekhar: Yeah. There's a new term in, in finance called EBITDA with two T's earning before interest, taxes, depreciation, and tokens. So, so token cost is a real thing. Uh, and you have to be very careful how, how you spend it. So one, of course you run small pilots, you don't run it large scale. We also, uh, are working a lot of a small language models versus large language, language models. A lot of context that we are working on does not require open AI and anthropic integration. Uh, so you can use much smaller models that in, that reduces your, uh, token cost. Some of them are even, uh, open models. Um, so, uh, so you have to be very careful. Most of the product that I see out there by default gets built on open AI or anthropic models, uh, which then, and, and so it's, it's a, it's quite a dichotomy that most of the products get built on anthropic and open AI, while most of the solutions don't need large language models. I mean, you don't need a model trained on multiple tetrabytes when you're all only trying to solve, Hey, can you find my guideline quickly? For example, there's no reason why you should be plugging into an open AI model. Uh, and that's more say an engineering talk, so to say, even though, by the way, I'm not an engineer, just, just a disclaimer. Uh, but I know enough about AI and product build to, to speak that language. Uh, so yeah, we are, we are doing a lot of experiments with SLMs right now, uh, trying to figure out. And even there, uh, there are options now where you can route even within SLMs, uh, do active routing to the model that costs the less and, uh, and works efficiently. And that's something that I don't see most of the, uh, most of the products out there using. So yeah, you have to be very cognizant of the, uh, of the tokens being used because if you're improving efficiencies by 15% and, and spending the same amount of money in tokens, and then there's no point unless you're experimenting. And, and I'm, I'm a huge believer. I, I think I said this on housing, I submit a couple of years back is that even at our size where, where our pockets were definitely not deep at all, we were happy to experiment with no immediate ROI from day one, because I could see the, see the future and I would, I would have gotten it completely wrong. Again, it's not that every time I've, I've seen the future, that future actually happened, but I, I knew that this is a very powerful technology and we were willing to spend some dollars into experimenting with it until we saw a real ROI. But what that helped us with is that we were off the gate faster than most people, uh, because we are, we were not afraid to, to experiment with some of the things. So sometimes if you're treating it more as an R and D cost saying that, I mean, this is just the cost of building something beautiful tomorrow, uh, that's fine. Uh, but yeah, one thing, I mean, two things that I'll strongly recommend to anyone with injuring bent of mind or working on products is that look at SLMs, uh, small language models and look at automatic rerouters, routers, uh, to some of the models, which reduces cost. [00:27:25] Speaker 1: Uh, definitely. And probably it's, it's mentioned in, in, in your book, right? [00:27:28] Shashank Shekhar: I, I would think. It's not, it was not meant for, uh, meant for engineering, but it was meant for sales and marketing professional, but yeah, that's, uh, that is, uh, that is not in the book. That part is, and that's something by the way I've learned after I have published the book. That's a, that's the speed at which AI is, is progressing. So I did not get there. In fact, uh, the second [00:27:49] Speaker 1: point that I wanted to make in front of you is in my industry, in the title research and diligence, I'm not a believer yet. There are a few startups that, uh, deployed agents to do automatic title research, uh, in the States of Florida, Maryland, and Georgia, and then they get stuck. And, uh, the, um, title industry on the record keeping perspective is so segmented where, um, at least now it'll cost me, uh, anywhere from 20 to 30 million just to attempt to build a, uh, Claude agent, which I'll softly call a, a AI or SI researcher with a super intelligence in front of it. Um, for 3,700 jurisdictions in the United States. Right. So, so the, obviously it's possible if I would be Elon Musk, probably I would throw the money and see if it works, but, uh, one message or I, I hear the echo of the message from different large corporation CEOs, stop having fun, make money with AI. Yeah. While you're saying the opposite, you're saying, well, I'm throwing the money to see how AI can, can give me that, that, you know, that boost. The message from CEOs that I hear on the MBA conference, stop having fun. So no more experimentation, no more fun stuff. I want to see [00:29:24] Shashank Shekhar: that AI equals money. I'm, I'm not a believer of that philosophy. I mean, with any new technology, if you're not experimenting, if you don't set aside a certain budget for experimentation, R and D, it's, it's like, it's like pharma companies saying that only produce drugs that FDA will approve and, and will become multi-billion. That's not how it works. Uh, and, and especially for a technology that nobody has used in the past, it's something that you have, or you can, I mean, see, look, then you don't build stuff. Then you buy stuff, which is fine. I have nothing against people who just want to buy stuff and don't want to build. If you have that mindset where you're like every single dollar that I put in, I need to know where it's going. It's not performance and growth marketing. Like every single dollar that's spent on ads, can you show that we got a client? Okay. I mean, that's understandable from that perspective, but also understand that on day one, that would not have been possible on, even from that, because you're trying different ad creatives. You're trying to see what works, what does not work. Similarly here, it's, it's a brand new technology. You're still figuring out how to do it. And then you, if you, if you only have the mindset that show me a two X dollar on every dollar that I'm spending, then, um, all I can say is that, yes, I mean that then you are being very conservative with what the scope of the technology is, because if you want to, uh, figure out what does the pushing the envelope look like and what does it, will it look like in six months or nine months, then you have to be constantly experimenting, uh, something that I'm completely completely okay to do because I know that if we continue to experiment, then we will be ahead of [00:30:55] Speaker 1: the curve. I actually agree with you. So the CEOs of the large corporations are mindful of the poor economy. They watch every dollar and I think they will switch to the mode of we want to lease or we want to, um, buy the technology rather than build. Um, two years ago, I heard the opposite story. Oh, AI is great. You know, build versus buy. And I hear even GSCs are experimenting. I think they got burned. They got burned badly. Millions of dollars went to the toilet and they just said, you know, [00:31:30] Shashank Shekhar: let's be mindful now. Yeah, it is. And that, I think that's the story until, uh, until a technology gets, gets stable, you will have the winners and the losers. Uh, and that is not because the people who were running it were winners or losers. It's just because sometime you pick a project, it works. So, I mean, we went through, I don't even know how many failures with it on trying on multiple things. Then we realized the AI is not there yet. We came back to it nine months later. Then it, some of it worked. Some of it did not work. So it's, it's an experimentation on daily basis. So of course, when, when you buy something, uh, you take that risk off your plate, the R and D risk goes completely off your plate, uh, which is great for some companies. That's perfect. We want to be an AI native AI first company. And if you want to be that, then that's just, uh, that's the cost of being in the game, [00:32:17] Speaker 1: I guess. So, um, we just mentioned that willingness to experiment is, uh, one of the key progress elements in the AI deployment. What other factors will stall or stop AI in deploying in the mortgage [00:32:35] Shashank Shekhar: industry? Do you see anything else, uh, in your world? The biggest factor is the human factor. I mean, the, the, the, the factor of not even, I mean, experimenting is because if you don't want to experiment, you can still buy something. That's, that's not a problem. Problem is adoption. It's the human factor, which is the problem. Uh, you can, uh, you can create stuff, you can buy stuff, you can do whatever, whichever way you acquire technology. Uh, but the mindset shift becomes a huge problem. And I'm not just talking mortgage industry, which is, uh, uh, which has traditionally been behind in terms of technology adoption. I know people in, in some very large companies and, and some of those are, will probably even be called tech companies, uh, because most of their operations happen through tech. They've been doing machine learning in the past. So these are really advanced companies where people come in and they're like, we don't want to do this. We don't want to have AI do this because they realize the potential of this, I guess. And they realize that the more they work on AI, the more the chances that themselves are creating or being eliminated. And that becomes a problem. I mean, if you were to build a fully automated underwriting engine tomorrow, do you think the underwriters would like to adopt it? They won't. And I don't blame them. If I was, I was in their shoes, why would I do that? It's like hiring an assistant who, you know, is getting trained to replace you tomorrow. Why would I be so keen to, to train that person? If I won't be keen to train that person, why will I be keen to train a technology to, to potentially replace me? So there is a human element and that happens with any kind of change, but this is, uh, a very seismic shift in terms of how the, the businesses will work. And that's a difficult one to tackle for anyone because, and if anything, I heard a, uh, heard a funny story out the other day from someone who works at Microsoft. Uh, they went, this person went actually on the other tangent because she was like, initially they were pushing us to use technology, like why aren't, aren't you using co-pilot? Why aren't you? And now the, the token costed so much. They were like, why are you using so much of co-pilot? So, so there it's, it's so funny is that you hear all kinds of a story these days. Oh, my people will not use AI. Now I'm hearing a story. My people are using too much AI and, and that's resulting into too much token cost. So that's why it's, there's nothing good or bad or wrong or right about it. It's just that, uh, any technology to any new technology and especially a technology of, of, of this magnitude will take a few years before it stabilizes the society, the work environment, the, the people's mindset towards it. And so you will hear all kinds of stories in favor and against in somewhere in between, uh, the potential it will kill the human race. It won't kill the human race. We have all kinds of opinions that we hear and, and who knows which of those will be true or not. [00:35:25] Speaker 1: My favorite story, just to, to add to your story is, um, we were employed to do AI deployments for the loan originator and, uh, we deployed the AI, uh, the management presents in front of me a timeline of how much time the AI saved them in processing. So without AI, one hour with AI, one hour. Mm-hmm . I ask a question that doesn't make any sense at all. Can you show me how the processor uses AI? So they use AI and then they scroll PDF files. So the same exactly. Yeah. And that person was fired. That person was fired because either they were resisting deployment AI and they just, they did understand that their job is at stake, but they didn't jump to a new position or new training early enough to avoid this. So that's unfortunate. But, but as somebody working for the mortgage company or bank or a finance company, uh, you really have to think through, is my job safe? Right. Uh, can AI replace my functions altogether? So, so that's, that's the story that, that really, you know, makes me laugh, but I feel bad for that person. [00:36:39] Shashank Shekhar: Yeah. It's, it's a, it's a human story. It's a human challenge. It's, it's got nothing to do with [00:36:44] Speaker 1: technology. Do you see anything related to red tape? Because I do. Deploying AI and red tape just, just puts a stop everywhere. GSC Fannie Mae guidelines, or just, you know, afraid of AI and people don't want to touch AI and, and they're just making it so hard to get AI vendors into, uh, our industry. PII data, cyber security risks, uh, SAC one, SAC two, and now SAC three, um, being fed ramped. So there's all kinds of things they throw at AI because they don't understand it and they just stole it, stole the progress. You don't see that on, on your farm, obviously, but do you see it around? [00:37:29] Shashank Shekhar: We, we are building technology for, for our own purposes. We are not, not in AI vendors. Of course, we, we don't go through, uh, go through some of the stuff that you're talking about. Uh, so yeah, probably not my expertise, uh, in terms of, uh, talking about how, how much of a red tape that you're seeing. Um, but I mean, people do realize that, um, and not, not to, not exactly on the AI space, uh, from, from a vendor perspective, but, uh, while AI tools can help you be, uh, cyber secure faster, uh, breaking that cyber security also has become faster from, from AI perspective. So it's, it is, it is a big challenge. It is, but having said that, that, that race has always been there is that it's, can the good guys be faster than the bad guys with whatever technology that existed. So when people say that AI has made it easier, um, especially, I think Jamie Dimon was on the stage and saying that Anthropik's, uh, mythos model has made it 10 times easier for, for, say, uh, cyber security criminals to, to do that, uh, understand that it can also help you, help you resolve something like that. So it's a, it, that, that problem will always be there, uh, that, that race will always be there because both the good and the bad have the same, uh, same army, so to say, same sword, same ammunition, so to say, but, but which one performs better? [00:38:50] Speaker 1: Um, agreed. Um, so, so just to, to wrap up our podcast, um, any final words to people, uh, just looking at us and saying, well, I'm probably too late in AI. I probably need to order your book on Amazon to read about it, but what can I do in my position today, uh, to catch up? [00:39:13] Shashank Shekhar: Yeah, no, you're not, not too late on AI. It's, it's a very new technology. I mean, if it's, it's like somebody who said, oh, I mean, I, I figured out industrial revolution three years after it happened. I mean, it stays there for next two centuries, so to say, if, if not more. Uh, so no, you're not too late. I mean, it's, it's like, did, should, should you have done it yesterday? Yes. But, but today is not that late, but if you put it out for say another six months or nine months, that will become a problem. And that's why I'm saying until late last year, I wasn't thinking that people are really, really late to it. And because yeah, I did a few things, but it was not capable enough to, to kind of start doing stuff that's a human. This was not truly artificial intelligence, so to say, at that point in time, but it has started moving way faster this year. I mean, so much so that Anthropics founder, Dario himself went out on, on one limb saying we should be regulated because we are moving way too fast. This is technology is moving faster than all of us imagined at even in, even last year. So, so you're not late, but if you keep putting it aside, even for a few months now, you could actually end up being, being very late. It's always better to get on ground zero and then go up the stairs as, as the building gets built versus trying to get into floor five, for example, that becomes very, very hard. So, and that's not just true for individuals, but organizations as well, is that if you're still just thinking, sitting on the bench about AI, do something about it. I'm not saying to spend millions of dollars into it, but just what happens is that the culture changes internally. People start talking the same lingo. People start understanding stuff, even with small little things, change can happen. But don't be completely aloof of it and being completely out of it because then it will probably be too late by the time you get into [00:40:57] Speaker 1: the game. Those are very wise words. Shashank, I wanted to thank you. I will see you in a couple of weeks in New York. Yes. At the Executive Roundtable. And I really appreciate your time today. Thanks, Alex. It's such a pleasure being here.