SW: AI can mean different things to different people. How do you think about it?

JB: AI, at its core, is nothing necessarily new. All AI is predictive analytics, even in generative AI like chatGPT and chat bots and things like that. All that it’s doing is taking data and predicting outcomes, or in generative AI, what you expect to hear out of an answer. Machine learning is probably where AI is utilized the most for people that really want to gain efficiencies within their business platforms, their workflow, things like that.

SW: Some parts of automation and AI have been there for years, especially for a company like UWM where you have all this data. And if you don’t have the right data and know that that data is good, you’re actually starting from less than zero.

JB: You’re right, you are starting from less than zero. And the good news is that all of us are starting from less than zero. There is not an organization anywhere, whether it’s in mortgages and fintech or any industry that does a great job curating its data year after year.

If we take the mortgage industry —  every time we get busy, nobody really cares for their data or had a good long-term strategy. We were all just kind of trying to stay afloat. No matter how you slice it, the fact is: garbage in will give you garbage back out, it’ll just be prettier and packaged a little bit nicer. And so the key behind all of this is making sure that you have a solid data structure and a solid data plan before you really start to move forward,

SW: How you think about the balance between humans and AI?

JB: In the 36, 37 years that we’ve been in business, we have never laid anybody off. And that will absolutely continue. No matter how much we utilize AI, we are so invested in our team members and their growth with UWM and their career paths that we would not [replace them with AI] — absolutely not.

People should be thinking: How do I keep the people that I have? And how do I train them properly? And how do I utilize AI in my existing workflow, in the existing products that I have, so that I don’t need to hire more people. I need to be able to be very good with the team that I have, make them feel challenged, make them feel fulfilled. But as more business starts to come in, I shouldn’t have to hire more people, I should be able to allow the technology to delegate a ton of the tasks that I was paying people for.

SW: How do you think about the rate of change with AI? Are you taking a measured approach or do you want to be first and fast?

JB: I’m notoriously famous for saying that I don’t use vendors — I keep everything in-house and rely on my family of 1,600 IT members to be able to manage that. But the truth is, we do have to partner sometimes with other companies. But I never want to be beholden to anybody who is not part of UWM so what we’ve done in all of the modeling that we’ve built, including the generative AI, is we’ve built our back end to be able to do plug and play.

We’ve built it so that we can pull out one vendor and one partner and replace them with another if we needed to. We’ve made it a somewhat headless experience, to be able to say, hey, we’ve got a product, we built this product, we own this product, we control this product, and it doesn’t need to be partnered with somebody. And if a partner is lagging behind in technology, then we can shift into somebody else very quickly and allow them the opportunity to provide the service that we need to provide for the team members of our brokers to be able to continue to level the playing field and then exceed the playing field for everybody else.

The question that I get asked the most is: Will AI replace my job? And the answer is very simple and very definitive. No, AI will not replace your job. People who use AI will replace your job.

SW: You have a huge team and you have open office hours where people can come in and like pitch stuff. When it comes to AI, have you seen people just be really lit up about some of the possibilities here?

JB: Oh my God, they’re so excited. It is a constant stream of people with different ideas, different solutions. But what’s even better is everybody wants to understand the strategy of where we’re going and why we’re choosing to go this way. Our team members stay with the company and with us longer and longer. They understand why we built things the way that we did and they can offer solutions because they’ve been enmeshed in building the technology or using the technology.

So they’re able to come in and say, hey, now that I have a basic understanding of the AI and the direction and the strategy that we’re going, I think we should totally shift everything and do it this way. And that’s where the best ideas come from. And I can empower them to actually go run with that.

SW: You have such a large number of employees, that saving 10 minutes, maybe even five minutes, times thousands of employees — that really is where ROI can add up pretty fast.

JB: The real ROI is: Can we give the same experience that brokers and borrowers are expecting from UWM?  We all have secret sauces that make us successful as mortgage companies. And it’s embracing whatever that secret sauce is and then figuring out how to never lose it, no matter what the business is. And that’s really where technology and solutioning and things like that should really come into play — not just with AI, but in general as it relates to ROI.

SW: What part of what you’re using is gen AI?

JB: From gen AI, we want to serve up information to our brokers and our internal team members in a conversational manner that happens incredibly quickly and is giving them answers in a different way than they would traditionally get them by going to the library and trying to look up a book. And you can take it so much further with those generative pieces to start having automated conversations.

And then where AI really gets into the next level is serving up questions based off of their behavior inside of our systems. As they’re moving through a screen, we can actually serve up generative conversations as they’re working. To be able to say, hey, you know what, you might want to do this right now or you’re choosing this product, but by the way, there’s four more products and if you add one more document, we can analyze that this second and tell you that this product may actually be better for your borrower right now.

That’s really where what we’re calling the traditional AI merges in with generative AI to create an experience that truly enhances the work that somebody’s doing and makes them smarter and faster and more efficient at what they’re doing.

SW: What parts of using AI worries you?

JB: In the financial space, what it boils down to are the ethics around making financial decisions based off of computer modeling. That really is where I think the biggest issue falls. You can worry about fraud but… what’s more important, I think, is not denying potential borrowers just based off of computer models. And that’s what the next step and the next evolution in AI in financial services is going to start to come from. The technology is moving so fast and it can learn so fast that it’s hard to figure out where these guardrails are going to be because nobody’s really utilizing AI in its fullest capacity just yet

SW: To that point, I think one of the things that we see the most hesitation around is decisioning. Lenders are using AI to do all sorts of things to make it a faster transaction, etc. What nobody wants to touch with a 10-foot pole yet is to have AI do any sort of decisioning for a loan. What’s your feeling on that?

JB: I would say that there is definitely somebody out there that looks at having AI do decisioning and would be happy to touch that for sure! So I think it’s just a matter of what your appetite is for technology and how quickly you can move and what your understanding of the mortgage industry is and how that technology could fit into that process to be able to do that.

SW: We’ve heard regulators say that when it comes to AI, all the same rules still apply. If your AI decides something, we’re still holding you accountable for whatever that technology does, because you built the technology. And one of the lawyers I talked to said, if a human made a mistake, I can explain that to a regulator, but if a machine that you built made a mistake, it’s less understandable.

JS: I’d almost argue that that’s not the case. The biggest piece if you look at decisioning — it’s all a level of confidence in what the data is, and what you’re trying to solve and how you went about solving it. So if you took the time to make sure that you have fully verifiable confidence in what you’re doing, what that modeling looks like, and you’ve tested it to death and you have that comfort level — it’s almost like talking to a human because you can show the logic behind exactly what that modeling did and exactly why it did it.

And then you can actually point to where it found all the information, how it extracted it, how it calculated — all of it. Everything creates a paper trail: people and machines create paper trails. It’s just making sure that when it comes to technology you’ve created the proper paper trail to be able to show why the decisions were made to either adjust it or justify that decision.