This Weekly AI Update explores into one of the most critical—and misunderstood—realities in today’s AI race: technology alone is not a competitive advantage. As companies rush to cut costs and automate workflows, many are unknowingly eroding the very thing that drives long-term success—human capital. Drawing from real-world examples, including a powerful Fast Company article and lessons from Toyota’s legendary operating model, this episode breaks down why AI must be paired with human intelligence to create sustainable value. If you’re a leader in mortgage, fintech, or any service-driven business, this conversation will challenge how you’re thinking about AI—and more importantly, how you’re implementing it.
[David] Listeners, we’re back with Pavan Agarwal with another AI update. Pavan. Good to have you back here.
[Pavan] Good to be here. We got a lot of stuff to talk about, about how to supercharge your team.
[David] Yeah, well, I’m really looking forward to this podcast. know, some earlier today, you sent me a couple of videos and I love the videos you send over one of them had to do with what I want to talk about. It’s the value of human capital. We’re going to be talking about the article you published in Fast Company that you wrote. It’s based on a white paper, by the way, listeners, you could go to the website and specifically the links in this interview and you can download the white paper as well as get access to the Fast company article fast company article. The article is titled why human article is titled why human capital is the ultimate moat in AI first financing. We’re going to get into what he means by that. But I want to talk about the two videos you sent me earlier. One of the things that came out in the video you sent me earlier was a story of about someone selling the automatic door openers that automated, letting people into hotels and they thought and they sold it as what’s the cost of your doorman at a hotel and you can eliminate that cost because if you buy our expensive automatic door openers, you’re going to have an substantial cost reduction. What I noted out of that was cost reduction is value reduction and when you have value reduction, it’s customer destruction, customer relationship discussion. Talk about some of this, because this is what you’re talking about. Everyone’s arguing that we need to have more AI. We can eliminate so many jobs with AI. Pavan, that comes at a human capital risk. Really it comes at a customer risk. Let’s get into this discussion.
[Pavan] Yeah, so that that dormant example is a famous one by Rory Sutherland, and he is a world famous marketing analyst consultant, customer service consultant. And the idea there is these McKinsey style consultants came in and says, hey, we can save you all this money by making automated automatic door openers. Okay, and then these consulting companies, they get a percentage of what they save you. So they get all this commission upfront, okay, but then you’re left to deal with the aftermath. All right, and so obviously a doorman does a lot more than open and close the door. helps you with the, it helps you get the taxi, helps you get the, gives you a personalized experience in the process of checking into the hotel and is aware of what your needs are beyond more than what automatic door opener can do. So what happened was, yes, you save money, but you destroyed 10 times as much in value. So then your vacancy rates went up, your hotel room prices went down, and so on and so on. Right. So ultimately, it ended up costing them. They paid these consultants a lot of money and ended up costing them more money. And I see the same. I see the same thing happening with this transition, this race for corporations to adopt AI. It’s like this, you have this people at a C-suite that are managers, C-suite are high level managers that don’t really understand what’s happening on the ground and have not spent, or it’s been either been too long or have not spent enough time to actually study the process, study what’s actually happening. And they’re just looking at the numbers and then they got these high paid consultants whispering in their ears that, you know, hey, we could save all this money and you could be the hero. And it’s so tempting. And then they rush to implement a what I call a half-baked AI solution and service levels and experience collapses.
[David] Yeah, it’s so true. In fact, you sent me also sent me a Gartner report. That’s from February 2nd of this year. And it’s a Gartner predicts half of companies that cut customer service staff due to AI will rehire in 2027. So we’re seeing this, it’s the pendulum, we all get excited about the new new the what we could do we’re all in this especially mortgage lending, we need to look at ways to cost cut. But when it comes at the destruction of the opportunity to be providing real good, accurate customer service. That’s where it’s gone too far. You have, you have set up a system within angel AI and at SunWest was something called brilliant airs, brilliant ears. And you’ve augmented what you’ve done is you’ve taken AI and human and found a way to mesh them together. And it’s based on I believe it’s a Toyota model, which is just, yeah. So let’s talk about this one. First of all, kudos for you being the forward thinker to figure out that AI is not the end all be all. It’s a great tool.
[Pavan] Yeah, so, you know, when we rolled out AI, Angel AI, what, 2018? And we were very upfront about this is augmented intelligence. This isn’t fully AI. This is AI plus human assistance, right? Because you want things in the mortgage business in any service business. You cannot lose the human touch. One of the things that Rory quoted was if you drove, if you have a website, doesn’t have your phone number on it and you drive and you’re forcing people to only web interface, right? And only deal with you through the web. Your customer conversion rate is 5%, but if you have a human interaction with that person, then your conversion rate is 30%. So yes, you have a call center, gotta pay for those people. You gotta have that human. You pay for the human touch, but you’re getting that back in a 25 % increase in business, right? So and the numbers are there that this is this isn’t new science. This is old science. The numbers are there. You gotta have that human touch. So our strategy when we roll out Angel AI was that the human will be involved in the process from beginning to end, will always be there and available as a touch point to the customer from the beginning. And so that solves two problems. Number one, with the human oversight, we can issue a warranty. So we have 100 percent warranty on Angel, so if she gives you a wrong answer and approves your loan when it should have declined, I’m going to fund it anyways. So that’s nobody else in the industry can ever give that kind of warranty, even on their human underwriting process. No, no, no single lender can say, hey, if my underwriter says it’s approved, there’s no absolutely no chance of this being you know if my underwriter issues the pre-approval there’s absolutely no chance of this being declined when we go to closing right so we’re the only ones that can that could actually do this okay so that’s you know Rory’s point and the whole point of the Gartner report is people and you have people at the board, they’re really tech CEOs, they’re not really tech leaders, right? And that doesn’t mean that being a tech leader doesn’t mean that you know how to code and be able to build systems. Being a tech leader. What it means is you have to have a visceral understanding of your whole organization top to bottom. And so that when you apply technology to your organization, you apply it in the right measured way and you have a you have a very good instinct of what technologies will survive and what technologies will die. And the industry is littered with that technology like Nokia, for example.
[David] Oh yeah. That was such a good example. And I mean, what they had over 40% of the market share. They were a dominant leader. And I’ve heard, use this story when I speak a lot, they are actually the ones that invented the touchscreen first and their engineers. And this is one of the things it’s, the feedback. They didn’t listen to their engineers. The culture in that company was don’t bring the truth to the own, to the board or to the executives, because they’re going to penalize you.
[Pavan] Yes.
[David] for doing so. And in that they ended up losing their company. And it was just a tragedy that was avoidable. They could have been the market leader.
[Pavan] Yep. I mean, that’s, you know, I like how Andy Grove of Intel put it, right? Only the paranoid survive. Right. And it’s a shame when we see the trouble in Intel is now that Andy is not running it anymore, that’s, you know, ultimately is what I mean by the definition of a good tech leadership of a tech company is, you shouldn’t have a middle management team that’s afraid to talk to the board. Because if the board knows everything, if the leadership, if C-suite knows everything that’s going on in the organization and has, like I said, a visceral understanding. Then no one can hide anything from them, or at least not for a materially long amount of time. So in the Nokia case, company that had 40%, 50% of the market collapsed literally overnight with Apple making the announcement because the team that built it left Nokia, walked over to Apple, and Apple ran with it.
[David] Yeah, it’s so astounding. talk about what Toyota did because they’re one of the reasons that works so well is where you picked up the billionaire model was from Toyota because they integrated engineers into the process. And so they made corrections almost in real time as things were being manufactured, which is really the purpose of the Brillianeers.
[Pavan] Yes, that’s exactly right. So I first studied Toyota back in like early 2000, 2001, 2002. There was a lot of papers have been written about Toyota’s just, they call it just in time, just in time manufacturing. And it’s built around teams of two or three people, maybe four people, small teams that have complete autonomy and authority. Most important authority, they have real authority to make decisions and get things done, and this is why Toyota, I mean, in my opinion, they make the they build the best cars, period. And each team works on whatever part of the car. And they work on their part of the car and the whole car at the same time. And they take that team takes complete ownership of the of the product, right, from design through factory production. So, if something goes wrong, it’s on them to straighten it out. ownership and accountability means, people, like especially in this industry, which is so commission oriented, right, people equate, in this industry, it equates accountability to commission and pay. That means if you did it right, you get paid. If you don’t do it right, you don’t get paid, right? or we take away a bonus or something. That’s the wrong form of, that’s too negative an accountability. The right accountability means if you do it wrong, then you have to go back and fix it yourself. So you’re incentivized because the assumption is that you will always ultimately do it right. So, you’re always gonna get paid the commission. You’re always gonna get paid the bonus. You’re always gonna make a lot of money because if you made a mistake, you’re go fix your mistake and it’s gonna get done right in the beginning. So the person working on the project is never under this fear that, I’m gonna lose my bonus, I’m gonna lose this, I’m gonna lose that, right? And it keeps all the incentives aligned.
[David] It really it’s just brilliant. And how do and the manufacturing on the mortgage loan is as complicated as a kin because of all the various every human has a different scenario they’re going into it. When you have been oftentimes criticized saying, well, they have all these engineers, they have all these people behind the curtain running AI. That’s my design. I mean…
[Pavan] You have to have engineers, right? And you have to have people, you have to have data scientists and people monitoring the data every day, looking at every single, we have a team that. that we have a huge team of 100 plus people that looks at all the millions of chat stations and they have these tools, they analyze it, they look for patterns, look for trends, seeing where the AI needs to be tuned, where the data needs to be updated, and then they go back and they map it up and put it back in. So this is human assisted feedback.
[David] It’s extraordinary what, when you team the two in this way.
[Pavan] Yeah, the technical term is like reinforcement learning with human feedback.
[David] Say that again.
[Pavan] Reinforcement learning with human feedback. Reinforcement learning with human feedback. So when you’re looking at the advancements in AI, we are watching continued ongoing advancements. Does this principle change at all as we go forward?
[Pavan] Well, I think it’s the application of this principle, right? I think a relevant principle is the Peter Principle. I’m sure you know of the Peter Principle, right? So this guy made this satire back when in the 60s about how people in corporations get promoted until they’re no longer competent to do the job that they’re promoted into, and then they get stuck at that job. And hence the creation of middle managers. And all that he meant as a satire is actually ends up being very much true. And I think now the Peter principle is being applied to AI. We’re taking a technology that’s kind of human-like, feels like human, and it does a great job at certain things. And yes, it could replace human on certain basic things. And then we’ve all suddenly decided, let’s promote AI to this new level that it’s not equipped to do.
[David] So first time we’ve ever seen AI get elevated to the Peter principle where, and that’s where I think this is as much educational video for those that are developing companies and leaning into AI to a big degree. So Pavan, as we wrap this interview up, ‘ve got a couple of things I want to talk to you about, an upcoming event, but why this matters? Why does this matter and for the future of finance mortgage finance specifically?
[Pavan] Yeah, because I think the best models are like 97, maybe 97 and half percent accurate. So that means, and if you just rely on the model, which is great, mean, machine, making human-like decisions or human equivalent decision 97 % of the time, right? I mean, it’s a breakthrough. mean, it’s nothing like no machine like this has ever been built before. But in finance, could you imagine if out of every 100 loans you did, you have to repurchase three? It’s not acceptable. You know, the same thing when if you’re self-driving car, right, if out of 100 hours of driving, you know, and it made an accident in three hours.
[David] And this is where we’ve talked, we talked about this with Jennifer when we’ve had her on the difference between probabilistic, excellence and determine or decisions versus, deterministic decisions, what you’ve done because you’ve combined the brillianeers, these people that work be in concert with the AI technology, you have gotten to the point where you have really gotten it deterministic. can absolutely with 100% accuracy, determine
[Pavan] Yes, between the algorithms that we have in the system. And we touched on that a little bit earlier this year when we talked about MIT’s latest paper called Recursive Language Model. The meat of that paper we had invented and patented like eight years ago. So between that and how man and machine come together inside Angel AI. So it’s transparent to the user and it literally gets trained in real time. So you get the reinforcement, human reinforcement learning is happening in real time. So that more people use it, the faster and better it gets. And like the latest build of Angels, like people are just blown away by how fast it is, it’s blazing fast.
[David] Yeah. I know even since when I started working with you on the speed at which it’s rendering decisions is amazing. We’re to be talking more about this in the Las Vegas event that’s coming up this weekend on Sunday. Talk about the mindset summit.
[Pavan] Yeah, this is going to be really amazing event. It’s a combination of mindset training from one on Sunday from noon to four at Delilah’s, we took over the entire Delilah’s, which, as you know, is normally you got to wait three months just to get a dinner table reservation. So we took over the entire Delilah’s and we’re hosting the mindset summit. We got some great artists that we performing for us. Along with if you go to mindsetsummit.com and look at the speaker list, some of the best and brightest from all over the world are coming in. have people, the top people from SoftBank are flying in to Vegas on Saturday just for this event. And one, yeah, so and they’re,
[David] Wow. For those that don’t know who, anything about Softbank, talk a little bit about that, why that’s so significant.
[Pavan] Well, I mean, it’s like any other major wealth fund in the world. This is one of the biggest wealth funds in the world right now. And we have people from, you know, in the audience and some of the speakers, and you can look at the speakers that’s there, but especially in the audience, and they didn’t want us to list them on the website. Their CEOs of some of the largest street firms and tech businesses in the world will be in the audience.
[David] Wow. I’m excited to be there. I’m going to be excited to be recording some podcasts. Thank you so much for Pavan coming on again, giving us another AI update, the importance of having humans partnered with the AI and you’ve done that through the brillianeers. love it.
[Pavan] Okay, this one. Yeah, there we go. So that’s actually last Friday in Puerto Rico and in San Juan. That’s that’s me and me, obviously me and Gabriel. Gabriel is one of our billionaires. And we you know, we you know, most of most of the people in my company work from wherever they want to work. And so we met we met last Friday, spent a couple of hours brainstorming through problems right, and brainstorming through ideas and literally, you know, at this beautiful setting, trying to think about and solve problems. We’re problem solving. And we’re in territory, we’re in areas that has never engineering problems that have never been addressed before anywhere by no one, by anyone. Okay, and out of that maybe three hour session, we produce four new patents.
[David] You’re kidding.
[Pavan] Not kidding. And then we got our patent attorneys on the phone, went over with them. They got all the information and they’re off to the races. And this is how we do it. And this is what it means. This is a real world example what it means for just-in-time manufacturing. What it means to be a billionaire. What it means to have ownership and accountability beginning to end. So right from design conception, innovation. And then Gabriel’s gonna see those patents all the way through to implementation and the feedback from customers and how do customers react to the new innovation that we invented today or that day. He’s gonna follow that all the way through along with the team that worked on it with him. So for those four innovations, there will be four small teams of say one or two engineers in each team along with Gabriel. And that’s how things get done. And why does it get done? why do tech companies, how do tech companies achieve this? The fundamental difference between tech company and legacy companies is that there’s nobody in the middle. Because at the C-suite, in my chair, I have a complete visceral understanding of everything that’s going on from training issue to a architectural design issue to how a loan is processed or how closing docs are printed. So having that kind of top to bottom understanding and being able to plug in and work directly with the team that’s on the ground and have intelligent conversations. So there’s no that Nokia event will never happen here because there’s nobody in the middle to filter.
[David] Yeah, filter out important intelligence that you need to know.
[Pavan] There’s nobody in the in the middle who have been who have been promoted to their Peter limit for Peter under the Peter principle, right? So they have been that nobody in the middle have been promoted to the level where they’re no longer competent. And now they’re more afraid of how do they keep their job as opposed to how do they expand their job.
[David] That’s so good.
[Pavan] I encourage everyone to read the white paper we wrote on this. I mean, it’s a roadmap. I’m literally sharing with the world, with other mortgage lenders, my formula, how it works. And it isn’t my formula. I will not deny that we learned it from Toyota and took those ideas and implemented in our company and it works. Toyota is just in time formula methodology works in the mortgage industry. This is what I’m literally giving this to you free information, right? And I made it public so you can use it, right?
[David] Well, I appreciate you sharing it on this podcast and our listeners are and we have a growing number of people listening to these AI updates. It’s really good. yeah, go ahead.
[Pavan] One last thing, one last thing is the second link I just dropped in here is, you know, the link is Angel AI STLM and how it compares with legacy lenders as they, you know, we’re seeing legacy lenders and I know one particularly and actually two, maybe there’s three that they’ve invested hundreds of millions of dollars, one or two of them, even a billion plus in building AI platforms. And this this this link actually gives you a side by side comparison. And you can see literally like if you don’t have leadership that actually understands what the technology is, what its limits are and how to sit down leadership that could sit down with his engineers, with its process, people with his customer service reps, with the sales team, listen to them, listen to the customer needs, understand the limits of what the technology can and can’t do, right? The key to being a good tech CEO is not knowing what technology can do, it’s knowing what it can’t do and then pushing your team and pushing yourself to say, do I get past what it can’t do and get it done, right? everyone right now, you know, everyone’s watching social media and reading up on AI and super excited about the hype and everyone’s all like, okay, I got to learn what AI can do. Right? That’s the wrong approach. Start with what AI can’t do and then think about solutions. If you want to separate yourself from everyone, everyone else running to AI and trying to implement AI in their companies, think about what AI can’t do and find a way to implement it so that you work around those limitations. Either build a better AI or build a better process to manage those limitations. That’s how you’re going to separate yourself and that’s how you’re going to win in this race. And that’s why the Gartner report that came out two months ago basically said that 50% of the companies that implement AI today in 2026 will roll it back in 2027 because they’re doing it wrong.
[David] Wow. Pretty amazing. Thanks so much for sharing all this with our audience. Pavan, appreciate you being here again and look forward to seeing you in Las Vegas on Sunday.
[Pavan] Thank you, David. Cheers.
[David] You bet.
