Editor in Chief Sarah Wheeler sat down with Christos Bettios, chief information officer at NFM Lending, to talk about the company’s private, proprietary GPT that is making his whole team more efficient.
Bettios has been at NFM Lending for seven years. He previously worked at US Bank Home Mortgage and First American Mortgage Solutions. He started his career as a developer and spent 10 years as a senior product manager at Fannie Mae.
Sarah Wheeler: You’ve been on different sides of the mortgage transaction. How does that experience inform the way you think about technology now?
Christos Bettios: NFM Lending has been successful because we are often able to build solutions — we have both a technical background to build full-stack development efforts, but we also have the business expertise. Part of the role that I was fortunate enough to have was managing vendor relationships, so I’ve been on the technology side of the mortgage business the entire time. NFM is fortunate that we can actually execute and not always have to rely on vendors.
SW: I’m always interested in buy versus build. Do you take things off the shelf and make them yours or really build from scratch?
CB: We build from scratch, and this is a decision we made years ago. I think if you’re going to be a winner in this space, you have to own your own destiny, meaning you can’t always be relying on the road map of some vendor. So, we decided to take on and build where we could. I have enough expertise and we have enough experts in the company that can drive that. And we have 500 loan officers that are on the street every day, and I can always go to them for ideas.
The big effort we have is a GPT. But we also own our own point of sale system. We’ve always owned our own loan application — it’s something we’ve built. We own it, we control it and we customize it for the best usage.
SW: What does your tech team look like?
CB: My tech team is hybrid. I have my expertise in technical analysis on-shore here, and then I use some overseas developers to do the actual coding. But all of the other product design and execution and rollout happens out of our offices.
SW: What differentiates your tech?
CB: Because I only have one client, I’m able to do the most customized thing that helps our business the most. Our team’s success is we’re very close to sales and operations. I’m not removed in some sort of headquarters building somewhere. I get my best ideas from my salespeople and from my ops people. So, the fact that I can collaborate with the users to get their ideas embedded into the solutions for NFM, that’s what makes it successful.
SW: How do you think about security?
CB: We put together a plan that is focused more on business continuity. I mean, obviously, we have all sorts of things set up to defend ourselves, and many tools. But we put a lot of effort into making sure we can get up and running again quickly if something happens.
SW: What keeps you up at night?
CB: I want to be best in class. So, our new point of sale system, we want to be better than anybody else. I’m competing with the companies that are at number one, two, three and four, not with number 25.
SW: How are you leveraging artificial intelligence?
CB: We’ve been talking about AI for a while, but in April of 2023 I formalized a two-year plan that consists of a three-pronged approach. So, the first is certainly the GPT side — LLM (large language model) solutions. The second area was, how do we build a system for the loan originators? So, our loan app has copilot features. And the third prong was, how do you introduce AI in the manufacturing process?
I participated in an MBA event hosted by AWS [Amazon Web Services] and presented on AI for lending. We have our own GPT that links to all our data, and is proprietary, that privately and securely can hit the internet — not because you got an account, but I’m going directly through private, single sign-on from within my enterprise.
But I also have allowed our company to have GPTs with specific company content, so we can have our own non-QM GPT, and a GPT that goes into our HR, and I can ask those GPTs a question. So, I’m giving our users proprietary, private GPT capability inside my enterprise on our SharePoint, to ask questions of our own code.
And we took that idea to the next level, where not only can you safely ask questions on our own content, but we can also query and do data mining at the same time. So, I can ask a question: ‘Show me all the loans that I closed six months ago above 6%,’ and I get my list. And then on that list, I can say ‘Oh, for this person, write an email to this borrower based on this data.’ That’s been in production since November and people are using it. I have, on average, 350 daily hits internally of our users hitting our GPT.
SW: What’s left on your two-year road map for AI?
CB: We are just releasing a new version of our online loan application that has Copilot features for our loan officers, where it’s not making decisions for the loan officer, but it’s assessing the app as it’s coming in in real time, and giving them critical information where they can move the loan approvals faster. And then we’re working on introducing AI-enabled OCR (optical character recognition), so we can real-time compare documents using AI to catch things and for that we are working with an AI vendor.
SW: What part of these new AI capabilities is exciting to you?
CB: The GPT. The ability for me to write a query right and ask it to give me my loans, and then write an email. Or to ask for files and a status update — it pulls my data out of that file and it writes an email to the borrower or to the referring Realtor about the status. I’ve just saved 15 minutes. And our people are doing that right now.
But I think where it’s exciting is for the loan officer. You are coming up with this content securely — you’re not posting anything to the internet. And this is not customer-facing. But if I can make my people 50% or 60% or 70% more productive, that’s the pick-up for me. And I can do all this on my phone — I don’t have to be on my computer
I’m a true believer that the winners of this world are the people who embrace these technologies. And we’re looking to make this available to other lenders, because our friends in the industry are asking for it, so we’re looking at that too.
