10.19.26 Advertising Versus Marketing; AngelAi’s Pavan Agarwal on Automation; Bond Movement

Robbie ChrismanWelcome to the Chrisman Commentary, Daily Mortgage News Podcast. I'm your host, Robbie Chrisman. Topics on today's episode include advertising versus marketing. Last week showed you why you can't predict rates, and my interview with AngelAi's Pavan Agarwal on automating mortgage loan manufacturing from document review and income calculations to loan approvals and challenging traditional lenders to modernize their origination processes. Thanks to today's podcast sponsor, Experian. Experian Verify’s integration with Workday gives mortgage lenders broader access to reliable, up-to-date income and employment data, streamlining verification, enhancing efficiency, and delivering a smoother borrower experience from application to closing. Visit experian.com/business/products/experian-verify. There's a cohort of American males who believe that with one to two years of dedicated practice they could win the Master's Golf Tournament. As a marketer, it's important to understand that you can sell these people anything. Advertising is not the same as marketing, though it's a component of it, and artificial intelligence is a part of it, especially for lenders. Many will say that any AI agent should proclaim, I'm an AI agent. And anything from a bot should proclaim, I'm from a bot. Would you let an AI agent manage your money or move money in escrow? After the Muse launch from Meta, U.S. banks are saying hell no to AI agents moving money. Some U.S. banks are thankfully blocking their clients from letting Meta's Muse AI bot root around in their accounts. Are you scratching your head about what's going on? Me too. Last week delivered a perplexing combination of persistent inflation and weakening consumer resilience, yet it delivered a surprising Treasury rally. Inflation pressures intensified across nearly every report, with September CPI rising 0.6 percent month over month, PPI surging 1.3 percent versus 0.5 percent consensus estimates, and import prices surprising to the upside. Underlying demand appears less robust than headline retail sales suggest, with real wages down 0.7 percent year over year, and the Fed's Beige Book revealed growing consumer resistance to higher prices amid rising fuel, freight, and labor costs. The majority of that data reinforces the case for further Fed tightening, with the December rate hike nearly fully priced and an October move still possible. Treasuries rallied as technology sector weakness and concerns about AI infrastructure financing triggered a flight to safety and short covering. The 10-year yield swung from a high of 5.32% to a low of 5.04% over the course of the week. While the 2s10s curve bull flattened 9 basis points to 36 basis points by Friday's close, mortgage backed securities weathered the sharp Treasury volatility relatively well, with spreads widening only modestly despite significant intraday price swings. The Fannie Mae current MBS coupon fell 12 basis points to 6.26%, and the primary-secondary spread tightened slightly to 91 basis points. If you really want details, conventional MBS performance, backed primarily by Freddie Mac and Fannie Mae loans, remained more resilient than the broader rate volatility might have suggested. Speculative pay-ups strengthened across most coupons, led by sixes and six and a half, with the largest gains concentrated in lower loan balance pools, underscoring continued demand for favorable prepayment characteristics. Loan level pay-ups also increased broadly, particularly for smaller balances, while high balance loans continued to trade at substantial discounts. The rally offered near-term relief, but did little to resolve the underlying tension between sticky inflation and deteriorating growth prospects, appearing to reflect positioning adjustments and defensive flows rather than a fundamental improvement in the inflation outlook. Markets are increasingly caught between an inflation-driven case for higher rates and growing concerns that consumers, corporate investment, and equity market stability may not withstand sustained financial pressure. Consequently, elevated rate volatility and uncertainty around the Fed's next move are likely to remain central risks for both treasury and mortgage markets. Lenders are likely to continue consolidating costs, investing in technology, expanding wholesale and strategic partnership channels, and refining customer retention strategies to preserve production and profitability. Credit scoring reform and affordability initiatives could create meaningful opportunities, but only if implementation maintains consistent risk assessment and broad investor acceptance. For today's interview, I wanted to welcome to the show the creator of AngelAi, Pavan Agarwal, on the to talk about automating mortgage loan manufacturing from document review and income calculations to loan approvals and how challenging traditional lenders to modernize their origination processes could revolutionize the mortgage industry. By blending data processing with real interactive robotic presence, he has a unique view of how automated the mortgage loan manufacturing process can be. Robbie ChrismanLet's start with everybody's favorite subject here, Pavon, and that is reducing the cost to originate a loan. Robbie ChrismanAnd without throwing any of my preconceived notions on the proverbial dumpster heap here, the opportunity you see, maybe that's a positive way of phrasing it, the opportunity you see to reduce the cost to originate. Pavan AgarwalIf I take a loan that's originated by one of my retail uh district retail loan officers, they direct their borrowers to the AI. Okay, the borrower has a conversation and gets in a couple of minutes, the application is taken. And then from that point to the loan applicationroval, there's literally no human touch. Unless the borrower says, you know, I don't understand, call me and let me, you know, explain this to me. Okay. And that's when the loan officer uh or an assistant would jump in and explain it to the borrower. But every document the borrower sends, every question, and you know, quite frankly, the borrower could could ask every question to the AI and get an answer. So imagine like tax returns, W-2s, bank statements, the borrowers will just simply take a picture of those documents on their phone and give it to the AI, and it's immediately processed, reviewed. The bank statements read line by line and look for the large deposits. All of that stuff is analyzed. When the document is reviewed and analyzed and says, Yeah, this is acceptable, or the income is $10,000 a month or whatever, as those decisions are made by the AI, it's one and done, it's final. There's no human underwriter coming in behind that, double checking. There's no quality control department that needs to double check that. The borrower and the loan officer can be completely sure that they can trust the AI because I put a warranty behind it. So if the AI got it wrong, if it miscalculated the income, if it missed a large deposit, if it didn't read the credit correctly, and I signed off on that, and the AI signed off on that condition or accepted the document, I'm going to fund that loan and I'll scratch a dent it. So that's how solid the process is. So can you imagine going from taking the loan applicationlication and getting a loan applicationroval in as little as an hour or less, and whatever you got is final. And then from there to funding, also, you know, the only time a human's involved is actually right there, double checking that the wire, because there's still money going out the door. So we we have some oversight at those certain points. When it's that automated and that fast, the costs go go way down. So the only time that we have that we incur expenses, human expenses, in this process, is if it's an FHA loan. And as you know, an FHA loan or VA loan, you got to sign a conditional commitment, and that requires a human underwriter, a DE underwriter, to review every document and then um certify that they reviewed it and then they sign off on it. That's the biggest expense, is on those loans. And that's what drives our cost up too. So right now we're we're averaging actually it's less than $125 uh a loan. Robbie ChrismanJust because borrowers can potentially get a loan without interacting a human doesn't mean that they want to. Robbie ChrismanAnd I'm I'm wondering, or maybe they don't care at this point. Is your experience that if a borrower can breeze through the process and doesn't have to interact with a human, they are fully ready for that? Pavan AgarwalUh they prefer it. And you're right, just because they they can, they don't they don't necessarily want to. Yeah, those situations happen. I mean, we we we have borrowers that like, no, I don't, I I don't want to listen to this. I don't want to give you my bank statement or whatever. Okay, and then the AI picks up on that and it automatically triggers an assistant. Uh we call them angelistas. Each originator, each broker, that whether it's TPO channel or retail channel, each originator is paired with an angelista, and then that person calls the borrower and coaches them through it, or the loan officer calls the borrower. Now, let's if you just break down a retail origination process, let's see where the costs are. Obviously, the number one cost is the loan officer. And you know, I say marry a loan officer, you're you really need that sales relationship builder. The second cost is loan processing, which is as much as you know, 25 to 30 basis points between the LOA and the loan processors. And then the third cost is sales managers whose job is mostly to push exceptions through underwriting and get you know problems resolved, resolutions through. And that's you know, any overrides of anywhere from another 20 to 30 basis points. So right here, we've we're already getting in the neighborhood of 75 basis points just on managing the sales process and managing exceptions. We've completely eliminated that sales manager process, we've eliminated loan processing, and we've eliminated loan officers assistance. So if you if you think about it, we're asking loan processors to do such a stressful job and to do the impossible, which is to get the documents and to double check it and to make sure this they spend a couple of weeks with reviewing documents and packaging it so that the underwriter can receive it and get the same day approval. Everyone wants that same day approval, so they put all this pressure on loan processors to be as good as underwriters and to try to outthink what the underwriter is going to do so that they don't have to go back to the borrow for another condition. Okay, so instead of that, if right up front the document is reviewed on the spot and a decision is made, and this is what we need or we don't need, think about how much you're saving. And then if it's a complex file, you don't need a sales manager involved to push it through or help structure it because the AI, anyway, as an expert, it's seen every situation possible. And so it runs through this algorithm and tries all the combinations and says, It says, this is how we're gonna solve this problem, this is how we're gonna calculate this income. I mean, we get it all the time where customers come into AngelAi, like deals that are declined elsewhere, they come into it and they and and they immediately get an approval. It's because it's got the knowledge of tens of thousands of experts and it's seen it all, and not only does it seen it all, but it also because AI can try out scenarios that never have been tried out before. Robbie ChrismanAt what point do companies driving down the cost to originate actually put traditional or legacy mortgage lenders out of business? Because there are still plenty of companies out there doing billions and billions of dollars a year that are paying traditional sales costs and costing uh toward $10,000 a loan to originate. Pavan AgarwalTraditional sales cost in the purchase business. You know, again, I say it once again, I say marry your loan officer because in the purchase business, realtors need a loan officer that's going to build relationships. And, you know, it's just that that little extra support at the open houses and all those things. Those processes are necessary. And that's what you can't get rid of that that part of it, especially in the in the purchase market. But you don't want your loan officer having to get involved in the manufacturing. I mean, like imagine you go to Toyota dealership and you talk to the car salesman. Does he go back? He says, Yeah, I'm gonna take the, you know, whatever, 2026 Corolla. Does he go back in the factory and starts building it for you? You know, he's gonna make sure the car gets washed and make you take you for a test drive. He's not actually involved in putting the paint on the car. But why is it that in in the loan business, we expect loan officers to get involved in the manufacturing? And if not loan officers themselves, then their immediate assistant get involved in manufacturing. Robbie ChrismanWhat does the future of loan officer look like to you? Should they be uh studying to enter a new field, or how do you see the role evolving over time? Pavan AgarwalNo, it's the future is amazing. I think loan officers are going to make more money than than ever before. I think a lot of them who have made their whole value proposition, their ability to get the income calculation right and to make sure that the loan applicationroval is perfect and to make sure they get those same-day approvals. If that's their whole value proposition, they're in trouble. And and a lot of them define that because I understand that because it's so hard and it takes years to get that right. But now that mechanical part of the job is automated. So the loan officers that are doing really well are the ones that are focused on relationship, focused on building the market and building their referral partner network and their referral channels, and keeping those relationships alive and not focused on manufacturing. My message here is not very different than uh many of the other big top traditional LO coaching programs, which have always said, hey, you as a loan officer, you need two assistants, three assistants, and you need the expert processor and all that, so that you get out there, get relationships, and let your team handle your deals. I'm saying let's continue that, except the team is now the AI and the team of AI agents that we give you. Robbie ChrismanI think ultimately the appealing part of all this to me is making smaller loans or more complex loans much more economically viable for companies. I'm hoping you can expand on it because that seems like where there's a lot of the hang-up. If I get a conventional conforming right down the fairway, vanilla loan, sure that's easy. Robbie ChrismanBut if the borrower has a special type of income or there's some complexity to the file, that's where it gets bogged back down and and I'm really not able to make money. Can you can you speak to I guess at what point you see those complex loans becoming something that AI can can totally handle, and then just also the economic opportunity of of being able to be an early adopter and get it? Pavan AgarwalI I'm so glad you you brought up that question. And and these days, you know, with gig economy, complex loans are now the norm, they're they're not the exception anymore, right? I mean, the gone are the days where everyone had a W-2 job and money in the bank, you know, all that. Now it's multiple jobs, it's side hustles, it's uh maybe a small business, you got or partnership tax returns and all of that, you know, usually, you know, if you're buying a $3 million house and you got a couple of partnership tax returns and complex income, that's worth it for a lender to go and and do that heavy underwriting. But if you're buying a $100,000 house and you got the same level of documentation, no lender wants to deal with it. So I'll give you an example of a deal that we did. It was a traveling, she was a traveling nurse. So that means like she had a different job assignment every every 30, 45 days. So over two years, she had like 20 W-2s and 20 last pay stubs. So she had talked to three loan officers before she found AngelAi. They wouldn't even take the loan application because they're like, and I don't, you know, I don't blame them, like get through all that paperwork. They don't want to promise or even imply they could do it or get the hopes up and then spend all this work on a small loan. It was a small loan because you know she's a hardworking lady and and making a living, and to put all that effort, so they just didn't even take the loan on. And then when she found AngelAi, she she basically put all of those documents, you know, all those W-2s into the system, and she had an approval in like two hours. And that's the difference. When we break down the MBA number, their top line numbers like $12,000 alone. If you look through it and you peel it back and you take out commissions and you take out, you know, corporate overheads like legal and and occupancy and things like that. And you say what portion of that is actually manufacturing cost, it comes out to around $5,000. And it also depends on how you look, what class lender you look at. I've seen it as high as you know, six, seven thousand dollars and as low as you know, you know, $2,500, $3,000. Okay, so the median is about $5,000. That's what what our analytics have come out with. So let's just say it's $5,000 and you're doing a $100,000 loan, that's five points. And we we all know there isn't five points in this business, at least not in the agency world. That's why those loans get get pushed aside. And as a result, and our findings from when we look at the HMDA data for the last seven years, and I'm not talking uh, and I'm not just just us, but it's also our findings have been correlated by uh the Urban Institute and also by a researcher at Lehigh University, which basically the findings are that basically out of every two black borrowers that are declined, one should have been approved. And out of every five Hispanic borrowers that are declined, three should have been approved. Those are those are like startling statistics. What we found is that because we run it through the AI, it doesn't care how much work, how many pay stubs you have, how many bank statements and exceptions and whatever. Because it doesn't care about any of that, it's just going to grind through it. We found that within basis points of each other, we've been able to achieve in terms of approval parity. Robbie ChrismanWell, I wish you the best of luck. The website is angelai.com. Pavon, hopefully we'll speak again soon. Thank you very much. Thank you, Robbie. Robbie ChrismanThis week is quiet on the domestic data front with Tuesday's housing starts, the main report. Starts likely declined further under pressure from elevated mortgage rates. Highlights from the rest of the week include September pending home sales tomorrow, a 20-year bond auction on Thursday or Wednesday, excuse me, the September leading index on Thursday before the week closes with Flash S&P Global U.S. Manufacturing PMI and Services PMI, as well as final October University of Michigan Consumer Sentiment. Today has the October NAHB Housing Market Index, and we began Monday with agency MBS prices slightly improved from Friday's close, the two-year yielding 4.71, and the 10-year yielding 5.07 after closing last week at 5.10%, down 14 basis points over the course of last week. Let's wrap up with a joke and some housekeeping. How do you feel about Roman numerals and potentially bringing them back? Who thinks we should revive them? I for one. Experian Verify’s integration with Workday, gives mortgage lenders broader access to reliable, up-to-date income and employment data, streamlining verification, enhancing efficiency, and delivering a smoother borrower experience from application to closing.