AI-Powered vs. AI-Native: What Mortgage Lenders Need to Know

[00:00:00]Ladies and gentlemen, welcome to a new episode of the Fintech Hunting Podcast.

[00:00:04]I have an amazing guest for you today.

[00:00:06]Today I'm joined by Brianna Lynn, co-founder at Copper Lane AI.

[00:00:12]They're building an AI-native mortgage origination platform designed to reduce administrative friction, improve consistency, and give mortgage professionals more time to focus on borrowers and decisions that require human judgment. Brianna, welcome to the show.

[00:00:29]>> Thanks for the introduction, Michael.

[00:00:31]I'm so excited to be here today.

[00:00:33]>> Uh well, it is great to have you.

[00:00:35]Let's start with giving people a little bit of your background because personally, you've been connected to the mortgage industry, but you weren't necessarily in the mortgage industry.

[00:00:46]Talk to me about your background and kind of how you got to where you're at right now.

[00:00:52]>> Yeah, absolutely. So, a bit about me, I'm originally from Maryland, ended up going to school in Pennsylvania. I went to Penn, studied finance and computer science there. I'd say from very early on, I've always been very interested in technology applications to financial verticals. So, even before I was working on Copper Lane, I had explored building technology in other spaces as well, and I've always found applications of AI to be most interesting. In terms of how I actually got into mortgage, I'd say earliest exposure was just through family. So, I had family members who had worked in the mortgage industry, more on the secondary markets actually rather than on the origination side, but I think that through a lot of conversations I had with those members growing up, I realized just how many inefficiencies are still in the mortgage process. Specifically, I remember one of the conversations I had early on helped me realize that in the secondary market, there are a lot of very costly things, for example, buybacks, and a lot of these things trace back to earlier on in the process in origination and a lot of it comes down to human errors, specifically from loan officers.

[00:02:07]Me and Ethan then got the idea of, "Hey, how can we eliminate a lot of these very costly human errors that come from the fact that loan officers need to handle so many different files and so many different things at once?" And that's how Copper Lane came about.

[00:02:23]>> I love it and as we dive in a little bit deeper, as you were exploring that and you're listening to people, what part of the origination process do you believe was most ready to be reimagined?

[00:02:35]And what did you learn once you began working directly with lenders and mortgage teams?

[00:02:41]>> Yeah, so I very clearly remember one of our earlier conversations with a lender was about how their loan officers spend a lot of time doing small tasks every day. And each of these small tasks include things like answering questions that a borrower might have, reviewing their documents, following up on documents that are missing. All these things each only take, you know, a couple minutes to an hour. But if you add all of them together across a single file, these things really compound and add up and it takes away a lot of loan officer time from actually closing more deals.

[00:03:20]We realized that in this entire process, a lot of it is human driven. One of the first things we realized when we entered that industry is that the loan officer or some human will always need to be there to own the relationship with the borrower. So we realized that, "Hey, we should let humans do what they do best and have them focus on owning the relationships, whereas AI can come in and take over a lot of these smaller manual tasks that take away loan officer time from doing what we as humans do best.

[00:03:52]>> Very well said. And you and I have been at a lot of conferences this year. I've seen you at a number of them. It's great seeing you guys promote it. And one of the terms I keep hearing almost everywhere I go now and every solution says they're AI-powered.

[00:04:08]Everywhere.

[00:04:09]You guys at Copper Lane describe your approach as AI-native.

[00:04:15]Describe to me and explain the difference for our listeners so that they understand the difference between everyone just saying, "Oh yeah, it's AI-powered." First, what does it really mean to be AI-native?

[00:04:28]>> Yeah, Michael, I think you bring up a very good distinction there. I think in general, when people think of software or technology, they think of point solutions. For example, like a CRM addresses a very specific need in a lender's workflow, which is the need to store and manage their leads. And same for other pieces of software they've seen. So, when they say AI-powered, typically what that means is there is an aspect of AI in that point software solution, so they incorporate AI into an existing piece of software.

[00:05:02]>> [snorts] >> I think that what we mean when we say AI-native is that AI is at the very core of what we do.

[00:05:09]Giving Copper Lane as an example, our core product is Penny, which is an AI loan officer. Under the hood, she's an agent. There are multiple sub-agents that support her. And at its core, what we do is purely AI. We're not selling a piece of software to these lenders with AI in it, but rather we are selling the AI agent itself directly to the lender. And that's what makes us AI-native.

[00:05:35]>> I I love it. And that's a great explanation as people are wondering, "What's the difference?" And especially as trade show season will ramp up again in the next month or so, you're going to see that probably on everybody's booth.

[00:05:47]It's in all of their literature, and I think it's important for people to understand that that differentiation.

[00:05:53]We talk a lot about technology. We're talking about AI, but you and I both know that technology alone rarely fixes an inefficient operation.

[00:06:02]>> Mhm.

[00:06:03]>> What should lenders have in place across leadership or process and data and ownership before introducing AI into their origination workflow?

[00:06:14]>> Yeah, I mean, great question, right? I think AI very much is something that can improve a good or a mediocre workflow.

[00:06:23]But, if the workflow already is very bad, it can only improve it by a certain increment. What I mean by that is, you know, a lender will always need to have these core people in place, the core people who make decisions on what technology to implement, and also for what the workflow should look like. Once they have a good understanding of what the workflow looks like, then I think it's a good idea for them to start looking towards different technology solutions to augment that, to make an existing workflow or process that they have in place even more efficient. But, like you said, it's very important to have a very strong fundamental workflow built out, and then use technology to improve pieces of that.

[00:07:05]>> Excellent. And one of the most important decisions for any lender is determining where automation should end and human judgment should begin. Kind of walk me through that process, and how do you guys help your customers kind of determine kind of where is that line of you know, let automation and AI take care of it versus when does human judgment need to step in?

[00:07:31]>> Yeah, I think a huge part of this is also how much does a lender trust in automation or a piece of AI. For example, Penny does a lot of different things. She does things that are borrower facing, but she also does a lot of more back office internal facing work. I would say that in general for the internal facing work, it's easier to build trust among lenders with these automations because their loan officers or their internal ops people or their sales people are the ones who are using it internally and they can see, you know, it works a few times and then you trust the system. I'd say the biggest barrier for actually trusting the system right now is actually more externally with pieces of software or automations that face the borrower. So, like one example I'd like to give here is like Penny does document follow-ups. So, if she sees that this file is missing XYZ documents, she can go text the borrower and remind them to upload the documents.

[00:08:31]In general, these workflows or automations, since they are borrower facing, lenders will be more concerned about. So, we try to build more human in the loop and human oversight into the process. And this involves having audit trails of exactly what Penny is saying when she's reaching out to borrowers, having loan officers review core decisions, for example, the pre-approval runs that Penny can make before the recommendation is given out to the borrower. So, how we approach this is to typically have full traces for Penny's reasoning as well as incorporate humans into this process so they can trust and understand what the AI or automation is doing.

[00:09:13]>> Great [snorts] example of how to bring in human in the loop so that there still is that checks and balance, but does the loan officer really need to be the one sending out the text or the email to say, "Hey, I asked you for six months of bank statements, you only sent me two."

[00:09:30]>> No.

[00:09:30]>> Right.

[00:09:31]>> AI can read that and can do it. And if Penny can send that out and I and I love how you're humanizing the technology, calling it Penny and everything else.

[00:09:39]That's very powerful as people get more comfortable with using these tools.

[00:09:44]We know in this mortgage market, lenders are under pressure to improve productivity, right? Lower costs, but speed alone doesn't guarantee a better borrower experience.

[00:09:55]What outcome should lenders use to evaluate AI besides just time savings?

[00:10:00]And how should they weather or how should they measure whether that technology is improving both operational performance and borrower confidence?

[00:10:10]>> Oh, yeah, great question. Like you nailed all the great points here. The industry very much faces a cost problem rather than a revenue problem. It's very costly to originate a single loan. And I think that when evaluating AI technologies, there are both quantitative and qualitative factors that should be considered. I think more quantitatively, you can look at metrics like how much time are you saving per loan?

[00:10:37]Or how much faster is this loan closing?

[00:10:39]How many more loans are you processing every month? But I think qualitatively, there's a lot of factors to consider as well. And I think that the distinction here is that for a lot of lenders, there's quite a big difference between one of their mediocre loan officers and one of their senior and more experienced loan officers. And a lot of that difference comes down to experience, context they have on borrowers, and all the different files and situations they've seen before.

[00:11:06]I actually think that AI technologies like Penny, for example, can excel a lot in this qualitative area because they have context across all your loan files, which means they've just seen so many more situations and have so much more context on what can happen across different loan files. What that means is they can catch things that maybe one of your mediocre loan officers can't catch.

[00:11:30]And these more qualitative signals or qualitative like measurements, although it's not as quantifiable, it's actually quite important because these are the things that prevent more costly mistakes later down the line.

[00:11:44]>> Successful adoption often depends as much on the people as it does the technology. What have you learned about change management, employee trust, helping loan officers embrace the AI that can support them rather than simply disrupt their roles?

[00:12:02]>> No, great question. I think that through all the conversations we've had at conferences, at trade shows, what I've realized is that a lot of people in this industry have been scarred by technology. A lot of vendors in the past have made big promises on what software or AI could promise them in terms of certain business outcomes. And then when the lender actually goes to implement it, one, it takes a lot of time and cost and energy to implement. And two, it really doesn't that deliver that 10x that the vendor initially promised. So people are already generally cautious about new technologies. So a lot of it is about being building trust and also rolling it out slowly with these lenders. So I think a few good ways to do this is like for one, what we do with Penny is we first roll her out to a small group of senior loan officers who can really well evaluate how she performs relative to an experienced loan officer. From there, once we've built trust with the senior loan officer team, we roll it out to the branch or even larger across the organization. So part of it is just building trust incrementally by doing slower rollouts.

[00:13:17]I would say also for us at least, we handle all of the implementation for lenders. In general, people don't want to have to set up the the and have to go through all the effort of doing it themselves. So, we handle that all for them. And then we also have certain structures in place that give them flexibility on how they want to use Penny or how much of Penny's features they want to use. So, it's very much about building trust and then also really listening to what they want and tailoring what we offer to those kind of um desires.

[00:13:48]>> Well, and I think that building trust and buy-in is what's going to help with the adoption throughout the rest of the company. So, I think that's an excellent point. For lenders listening who are interested in AI but unsure still where to begin, what's the most practical first use case they should evaluate and more importantly, what questions should they ask vendors before selecting the technology partner?

[00:14:13]>> Yeah, absolutely. So, I would say to your first question there, where I see the most value for AI initially is more internally. So, helping your loan officers improve their internal workflows. Right now, coming back to the point I said at the very beginning, your loan officers probably spend a lot of time handling those borrower questions, reviewing those documents, you know, setting up follow-ups. All these things are very easily automatable and AI can already handle many of these things very well. So, I see a lot of value initially in using AI to speed up some of these internal or back office tasks.

[00:14:52]In terms of more of your second question about the questions that lenders should be asking when evaluating AI tools, I think one is definitely security and compliance. So, often times these AI tools will be touching your borrowers' PII, their sensitive information, and it's very important for lenders to understand what the AI is doing with that information and how that data is being securely stored. So, security is always going to be one of the most important questions.

[00:15:26]I'd say another important question for lenders to ask AI providers is also just what value can this AI deliver? I think there are a lot of AI solutions out there that sound very flashy or can make very big promises, but in practice, AI makes a lot more mistakes, especially on more complex files. So, whether or not this AI is advanced enough in order to handle those edge cases, the more messy files, is another question that lenders should be evaluating out of the vendors they work with.

[00:15:58]>> As [snorts] Copper Lake continues to grow, what do you hope lenders and mortgage professionals and borrowers will ultimately experience differently because of the work that you guys are doing?

[00:16:10]>> Yeah, I mean our goal is always for lenders and their loan officers to be able to focus on what matters to them most, which is really understanding the people and understanding the borrower. I can say for very like with very high confidence right now that a lot of loan officers, they don't like spending a lot of time, you know, chasing down borrowers. It's very tedious and it's also just not necessary anymore.

[00:16:35]After like with AI and the current capabilities of what agents can do, a lot of these workflows that have existed for decades in this industry can now finally be automated.

[00:16:47]I think there is a lot of potential for lenders and I think that most lenders understand that it's not a matter of if AI is coming, but rather of when it is coming. I would say my recommendation would be start thinking about how you can use AI internally in your workflows first, test it out, see how it works with more of those internal workflows and I can promise you that often times you'll see it provides a lot of value to automate a lot of these manual tasks.

[00:17:15]And then from there, you can see more value of how your loan officers can focus more on doing sales, doing top of funnel, bringing in new loans for your business, focusing on really what humans do best.

[00:17:28]>> Briana, you're a wealth of knowledge and expertise, and I love talking AI with you. You are welcome back anytime.

[00:17:35]If people want to find out more about Copper Lane and all of the incredible things that that you're doing, what's the best way they can get a hold of you?

[00:17:43]>> You can directly reach out to me. My LinkedIn probably will be attached with this podcast. My email is [email protected].

[00:17:52]You can visit our website copperlane.ai or LinkedIn page to learn more. We're really active with posting blogs, posting on our LinkedIn, so I'm sure you'll be seeing what we're up to very soon.

[00:18:04]>> Briana, thank you for being a guest on this episode of the Fintech Hunting Podcast.

[00:18:10]>> Thank you for having me, Michael. You always ask the best questions. Really enjoyed this conversation.

[00:18:15]>> Thank you.