[David] Listeners, we’re back with Steve Butler and True AI is making a real difference in how lenders operate, reducing costs, causing loans to get flow through their systems and get processed cheaper, faster, better. I love it. Steve, congratulations on the successes you’re having. Love having you back here.
[Steve] Thanks, David. Great to be here.
[David] You’ve had really some pretty amazing successes in what you’ve been doing. for those, by the way, that don’t know you, you are the TRUE CEO. you’re we’re the king leads the company in growth strategy, enterprise expansion across the mortgage industry and financial industry. You’re doing a great job, and you have an extensive background and really having some successes, which we’re going to cover in here. I’m really interested in getting into Fannie Mae’s new AI governance framework. it goes into effect August 6th. So, it’s really timely that we have you on here. We’re covering this. What’s the biggest misconception you’re hearing from leaders? And I’ll share some of the ones I’m hearing.
[Steve] Yeah, great great question, David. well it’s super interesting in that you know, lenders have had for decades, you know, governance in place, right? And so they could support and survive, defend any audits, right, that are done and they regularly get audited of course on their loans. And so the biggest misconception is we got that handle. We’re already there. But what they’re missing is that as vendors and themselves deploy more, you know, AI, you know, generative AI in particular, that confidence needs to sort of be re-looked at because those solutions bring a whole another dimension to what deterministic results. And you know, if you don’t get the same results for every time you run that application, if the results can change, then you’ve got you’ve got a governance issue. And so I think the business misconception is they have it handled. And I also think most of them don’t really understand what it means. I mean they’re having enough time, difficult enough time to sort of understand what AI is all about. Now you’re adding another layer, which is okay, it’s affecting your governance. And so I think those two issues are and that’s why we put out an educational piece, the content, a guide to help them figure this out.
[David] I’m really excited about what you have published on this, but you touched on that you do this, you solve this. Just for those listeners that are listening to you for the first time, could you just real briefly explain how you solve for this?
[Steve] Yeah, yeah, no, which certainly so the issue is that generative AI is gonna generate, you know, probabilistic answers, right? It’s gonna make the best fit answer it can.
[David] I love the word and I want to stress the probabilistic versus deterministic because there’s a big, big difference. We’ve covered that. We’re going to put a link, listeners, to a previous podcast. This is really important that you understand this, listeners. If you’re not getting the difference, you’re missing a big, big piece as you’re forming your AI strategies, as you’re implementing solutions. You must make sure that they are deterministic.
[Steve] Yeah. And and you know, as they get better and better, they’re putting out better accuracies. But the problem is, is if you if you put you know, a year from now, two years from now, you get audited, you go back in with the same prompt, right? and say, give me the answers, whether it be a data extraction or a chatbot response, whatever it might be, the answer is going to be different, possibly. So it’s not deterministic and what AI governance is all about is having deterministic, repeatable results. And so what we build from the ground up is the ability to give you always deterministic answers with our platform. What we’ve done, David, is we’ve combined a deterministic model that we built over the last nine years, heavy machine learning, you know, and a bunch of provable correctness kind of methods that we added. And we blended that with the Frontier LLMs that we used so that we basically have machines validating machines. So now we’re fully traceable. That’s basically what’s happened. Very different. I call it a third generation architecture because the first generations basically are not deterministic. Now we have a deterministic architect. And we’re the only ones that have this kind of a approach to solve AI governance.
[David] Yeah, it’s really, really you know, it may be a good idea for you to explain the Fannie Mae AI governance was about because we’re talking about that and again some people may not be up on it. So explain that if you could, Steve.
[Steve] Well, as it pertains to AI, you’ve always had to make sure that every data field that you’ve got, every document, every decision has is grounded with a deterministic answer. So you have to have full traceability and full provability. So, you know, because what happens is the data is what’s used in decisioning, and we’re starting to see lawsuits, by the way, showing up where borrowers felt like…
[David] That’s a risk factor that wasn’t there before as a result because people are saying, prove how you came to this determination. Give us a roadmap.
[Steve] That’s right. There was a big big case just got settled in Massachusetts for almost three million dollars related to student loans and there was a class of there was a class of of students that basically were biased against and the vendor basically didn’t have the governance in place to basically be able to prove, you know, that the data, you know, was the data itself. And so as you know, the data was repeatable. And so this is gonna come heavy, you know, class action, whatever you wanna do is gonna come heavy at the mortgage industry and but for big numbers. And so we are really advising lenders to get their act together on traceability and provability.
[David] I’ll never forget one time my wife and I were on vacation and we were going through a hotel. We just happened be staying at a really nice hotel. And we walked by this suite and that or there’s room that like they have any conference, you walk by there, and there I noticed as I sat, I said plaintiff bar, and they’re how to sue the mortgage industry. Now, this was a number of years ago on other issues. But there’s a whole it’s important to our listeners understand, there is a whole legal enterprise out there, lawyers. That are looking for ways, and that’s how they prosper. That’s how that’s our business model, looking at how to sue lenders. And you’ve got to have this covered. And it’s so important. Now, one of the things I’m really impressed about your guidelines, it really addresses some of the biggest issues, which is AI accuracy and its traceability. Can you explain the difference between why that is becoming so important? You’ve touched on it already, the class action lawsuits or is there more to that we should go into?
[Steve] No, other than as you point out, litigious lawyers see AI as in a huge enabler of them to sort of cause a lot of this problem, the legal community there’s sections of lawyers that basically they’re looking for how do we how do we find something that’s ripe and fertile, right, for lawsuits that can be settled for, you know, six, seven figure kind of lawsuits. And then and AI is creating that because of the probabilistic answers which lead to decisioning and et cetera, et cetera. And so it’s something that lenders have to get set to defend themselves against.
[David] One of the ideas that you introduced is something called trust architecture. I love that term. You trade you TM’d it. And what does that mean? And why do you believe lenders should be start evaluating our platforms through the lens, through the lens of a trust architecture? Love that expression.
[Steve] Well, I’m glad you like it because it really to me this is all about can I trust the data will be my friend a year from now, two years from now, five years from now when I might get sued, right? or I might get audited. Can I trust that that data is my friend? And it’s my friend when it’s the same data every time, no matter what when I run it, right? Same inputs should give me the same outputs. And so we built an architecture so that enables that you could change you can change the date. But if you give it the same inputs, you’re gonna get the same outputs, whether it be data or whatever other AI application you’re running. And so that’s but we did that by integrating a deterministic model with these frontier LLM models and they validate each other. And so you reduce the human in the loop requirement and you basically make sure you’ve got full traceability and deterministic results. And so that’s the key to what a trust architecture is all about. We build it from the ground up. I mean and it leverages the model that we built for you know nine years by running tens of millions of loan through it. It’s very deterministic. So you take that and blend that with the Frontier LLMs, Claude etc. and you basically have a very, very powerful but very deterministic result.
[David] So Steve, if you’re sitting with the CIO or COO seated at the mortgage lender today, what questions would you be asking every AI vendor before August sixth?
[Steve] The most important question and by the way, I gotta tell you, we looked out and we see so little conversation going on about AI governance among vendors. They’re so focused on building the great next big tech stack. You know, let’s get the best AI out there. They’re forgetting that enterprise AI has to have all these other features like traceability and deterministic results. Otherwise, to me it’s just gonna be get them in trouble down the road. So the first thing I want lenders to ask their vendors is what are you doing about AI governance? What have you got in your system that means that if I gave the same result same answer, same prompt a year from now, two years from now, I’m gonna get the same. What have you done to ensure that? Do you have something like a trust architecture in place to do that? and what does it look like? Those are the key things. Otherwise, that lender is super exposed. They may get great results, but you know, and it will they probably will. There’s a lot of good vendors out there. They’ll get good results. But the fact of the matter is, is it could be a time bomb kind of waiting for them down the road if they don’t if they can’t ensure that there wasn’t bias in the decisioning, for example. Right, and borrowers that should have got loans didn’t get loans, etcetera. And that’s only gonna come by being able to defend an audit, right? And defend a lawsuit. And so that’s why we think this governance is so important.
[David] Have you gotten any insights of what these audits are gonna start looking like, Steve? Have you heard of anyone any insights of what the lenders can participate?
[Steve] Yep. Well, we you know, we poured through whatever literature has been put out by Fannie and Freddie and they put out some pretty good stuff and what we do know is and you know, they do their standard audit but they add the what are you doing on supporting governance, you know, in terms of the framework and the guardrails that you’ve put around your AI to make sure it gives you the same answers every time and so, you know, we expect an audit which, you know, I haven’t heard yet what what actually has happened. I’m sure they’ve started to well, August sixth. Yeah. If
[David] Well yeah, they’re planning it. I mean they’re gonna be doing this. They don’t put out so it’s something. Yeah.
[Steve] If they go by what they’re saying, they’re basically gonna be asking lenders to support, you know, the traceability improvability for all their data fields, right? And and
[David] Yeah. Yeah. And they’re gonna be putting up files. They’re gonna be a random selection of files. They’re gonna give you a file, then you’re gonna say, okay, here’s the thing. And they’re looking for the ones that are the more complex of the decisions where they’re where there is risk. And listen, Fannie Mae is a counterparty. That means they share in the risk in the decisions that are being made and so they’re they have a very keen interest in what is this this regulation didn’t come out, or this new guideline did not come out because they’re not at risk as well. They’re feeling the pressure and they see the trends on the plaintiff bar that the legal industry, that’s suing letter. So it’s really, really important that they focus on that. When someone’s working with you, Steve, I’m assuming that you are. We asked the question if you were sitting down with a letter. How much do you spend actually advising some of them as every good vendor partner should? How are you doing that?
[Steve] First of all, in our platform itself we have what we call the data panel which provides the traceability. So we show them that for every field that’s in their you know in their system of record, you know, we can defend it with a document, the bounding box on that document, everything. We can give you full traceability back. So we educate them on the importance of that and making sure that they can do that. I don’t want the vendor to just kind of pick the best LLM solution and you know, run amok, you gotta have enterprise AI which includes the support for AI governance, compliance. You know, compliance is gonna become a very, very big issue in the whole AI world as you know, David. And and this is this is the big part of it.
[David] Yeah, the best part is, what do they say? the best offense is a really strong defense or best defense is a bigger strong offense. In other words, it’s going into it every one of those audits with the confidence that you know your system does it. And what I’m as you guys are implementing these or working with your lenders that you’re working with, you’re clearly giving them a point. Come on, bring it on. It’s nothing like a good fight that you know you can win if you have to go into one. So I could as well. But one of the things, Steve, I really love about you and what you’re doing, you are very much a forward thinker. If you’re looking beyond August 6th, do you see this as a one-time compliance event? or does this represent a bigger shift in how lenders should be thinking about AI?
[Steve] If you look at other industries, right, ultimately the chief compliance officer, right, they they and the you know, the head of security, you know, they the CISO, they end up driving so much of decisions because that’s where the huge risk is, right? And so I think we’re gonna see over time those folks taking greater control of what’s happening, you know, inside of lenders. I don’t care if you’re an IMB or a bank, you know, you’re gonna see stronger and stronger compliance as a top priority item and it should be or at the board level, at the ownership level, it should be. because, you know, you could lose big if you’re not you know careful. Now, David, the alternative is, and here’s another thing I don’t want to predict, where AI is supposed to be a cost saving, right? A cost saving and it should be, you know, when you know something where you take a lot of cost out or you build significant efficiencies and productivity enhancements. Imagine a day where because of the governance issue, you actually have to review everything by humans. Humans the only way around not having full traceability by machine is to basically get full traceability by human and so now you’ve got all of a sudden I invested in LLMs, but I have to have my human team validate what’s actually on the document so that I do have traceability then. And so the alternative is adding cost if you don’t have a machine validating machine kind of solution.
[David] And when you haven’t taken the time to really get to know by someone by a leader in the industry such as yourselves. Steve, I’m sitting and looking at this and I’m stuck on the trust architecture. That is brilliant. I understand for your staff, you came up with this. There’s just there’s just so many things that the way your brain works and how you’re designing this thing, give us a little bit more of a background how it is that you function because there is a difference. When you look at technology, it’s built by people. And the nature of those people, the brilliance that those people have makes all the difference in the world.
[Steve] Well, I appreciate it. I mean I just started thinking to myself, what we really are selling is trust because underwriters and whether you’re underwriting or selling a loan off, you’ve trust you’re trusting the loan file. is what it is, right? And so I really started thinking, you know, we’ve got to figure out, you know, what we’re basically selling trust. And that led to this idea of the trust architecture which we’d built. But I wanna just say that maybe because I’m an old guy and I’ve been doing this a lot of years, I had a really good sat part of my career in which I sold compliance solutions, compliance platforms into the enterprise and that became a real big success. And it ended up being a company that is probably doing somewhere, you know, close to maybe a 500 million to a billion dollars in compliance solutions and we pioneered that whole area. We pioneered that here area. It’s a company called Manage Soft. And so I’m taking those learnings and bringing it here and it turned out that and it was banks and financial services primarily, it turned out that the compliance issue ruled the day and that was the big thing. And I think we’re gonna see that in AI. I really believe that. You know, AI’s gonna get commoditized. You David, it’s gonna be a commodity, right? This gonna be a commodity, but won’t what won’t be a commodity is the fact that you’ve gotta basically have compliance and governance in place and that’s the thinking here. Yeah.
[David] And then how do we do that? Maintain the relationship. That’s one of the things I love about how you have blended this, your solution together with the ability to do that. as we wrap this up, we’ve been talking and focusing on Fannie Mae AI governance, and that is a very important topic. But one of the things I underscore that I want to underscore is you help lenders get loans done faster through the system. So the speed by which tasks get done is dramatically reduced. But it’s also we’re scrambling in our industry to reduce cost. Again, those that know you, they understand this, but could you explain how you do that again?
[Steve] Yeah, yeah, no totally. And by the way, this trust architecture comes completely in play again cause what happens where’s all the cost? All the cost is in the labor that’s being used in every stage, often related to data, you know, and in and getting the data to the point where you can trust it so you can use it down in underwriting, right? And downstream processes. And so much of this back and forth. I mean the underwriter opens the loan file four or five times, not because they like the file, but because it’s incomplete. The data’s not ready. Right? so the idea is that you can reduce a ton of cost if you can get your trusted data faster and with less human. So again, machines doing a lot of the validation and it all happening almost in real time is really where you can start to save some real cost. Now underwriting opens the file maybe twice instead of four or five times, right? And so huge cost savings are there. and then the whole process compresses. I mean it I’ve had so many lenders tell me, Steve, nothing starts until we can trust the data. Right? And so with True MOS, you have a trust you have a trusted data platform, right? That’s the idea. And so so that’s where the cost savings really comes from.
[David] Yeah, and you’re having such great success doing that for those that are talk your customers that are using yours. again, we we vet our vendors that are we have on our advertisers and we want to have those with great reparation. Steve, you are a really highly valued vendor for so many of the lenders that have wisely chosen your solution. Kudos to you and the success you’re having.
[Steve] I got one customer now that’s north of two thousand users on our MOS platform. Tremendous reduction of costs and obviously the AI governance.
[David] Yeah. Which as we’re finding out today, it’s growing real important. August sixth is a big date, Steve, and you’re helping people go past till past that with confidence and an assurance that what you’re doing can be trusted and it’s reducing their costs. Great job. So good to catch up with you again. Congratulations on your success. I love your team.
[Steve] Thanks, Dave.
[David] The people I work with within your team are just delightful and encourage people. How can people reach out and get to learn more about
[Steve] I think go yeah, our website is very informative, but also we’ve got a number of we’re doing a lot on social with a lot of posts about traceability. You can access the guides that we’ve put up. So I’d say between our social outreach and our website, you can read all about this and get a hold of us.
[David] Yeah, I encourage people to do so. interview. on. The only good thing I say is we don’t have you on enough. Please come back soon so we can share some more of your brilliance with our audience. Thank you so much.
[Steve] David, I’d love to and I appreciate you carrying this ball as well about AI governance. I think it’s gonna really help the industry in general over a big issue.
[David] Yeah, it is, it is, and I’m honored to do so. Gotta keep this healthy, this industry healthy and strong. We’re doing one of the most important jobs in America. That’s putting people in homes and helping them stay in homes. And doing so,lowering risk while we do it. Steve, thanks so much for coming on and being with me again, my friend.
[Steve] Thanks again, David. Talk to you soon.
[David] Looking forward to it.
[Steve] Okay. Bye-bye.
Important Links:
As CEO of TRUE, Steve leads the company’s growth strategy and enterprise expansion across the mortgage and financial services industry. With a deep background in scaling B2B SaaS and fintech platforms, Steve brings a relentless focus on execution, operational excellence, and customer success. From sales, to marketing, partnerships, business development, and customer success, Steve’s tech-seller approach and customer centric vision are critical to TRUE’s continued expansion into the lending industry.
A true believer in AI, Steve has decades of experience running revenue organizations and building valuable tech companies. Prior to joining TRUE, Steve was CEO of GoDocs, a leader in commercial lending automation, founder of AI Foundry, and led revenue organizations at a variety of successful technology firms. Steve has a BSEE from the University of Rhode Island.