Learn How Rocket Is Using AI With CIO Brian Woodring

Rocket is embracing artificial intelligence and leaders there say it will be a game changer in the lending industry.

Varun Krishna, CEO of Rocket Companies, emphasized their focus on AI in a statement accompanying the third quarter’s financial results.

“In today’s climate, innovation is essential. With the incredible amount of data in our ecosystem, paired with the transformative power of AI, Rocket is uniquely positioned to disrupt the industry. We’re just getting started and can’t wait to show you what we are building to revolutionize homeownership,” Krishna said.

Last month, continued advancements to the company’s Pathfinder tool were announced. Pathfinder is an AI and machine learning-powered search engine used by more than 40,000 housing professionals, according to a press release.

Rocket is enhancing Pathfinder using large language models and is adding more than 3,300 new loan scenarios. All told, Pathfinder collects 75,000 data points each day that can be used to train the tool to be more effective.

Brian Woodring, chief information officer of Rocket Mortgage, said in a statement that this technology, and others that leverage advancements in AI and machine learning, are changing how they facilitate homeownership.

Woodring recently sat down with Editor Kimberley Haas to talk more about how Rocket is harnessing the power of artificial intelligence and machine learning.

Haas: As you know, you can’t escape a news cycle without hearing something about AI. So I’m hoping that you can tell us about the artificial intelligence and machine learning that goes into Pathfinder. And also if you could tell us a little bit overall about how Rocket Mortgage is using AI and machine learning to solve pain points for clients.

Woodring: Absolutely. You know, if you don’t mind, I’ll probably reverse those and maybe start with a more general case, because then I think you’ll see that what we’re doing with Pathfinder really is maybe an example of an overall narrative that we’re going to be rolling out over the next few months. This is one of many announcements we’re going to be making about some of the amazing work that we’re doing.

So in general, our view is this, that technologies change and evolve and advance and that’s very exciting. But the durable mission of our company does not. And that it is our job is to make the process of getting a mortgage for a client as amazing as possible. We want it to be fast, we want it to be simple. We want it to be low-stress and confidence-inspired. And so we’ve been using technology for years.

We’re the world’s first internet-powered lender. We are on the journey right now of becoming the world’s first AI-powered homeownership company. Really, bigger even than just mortgage. But mortgage is obviously a huge piece of that. And the way we see it is that the role that AI can play is in simplifying, streamlining, automating, and making the entire process much more confidence-inspiring for our team members, for our partners, and for our clients.

We want to build technology that makes this entire process so much simpler. Pathfinder, I think is a really cool example.

Pathfinder is our knowledge base that has all of the data that we know about the right way to underwrite, process, and fulfill a mortgage. It has regulations, it has investor guidelines, it has our own best practices.

So if you imagine our best underwriter here, our best processor, our best loan officer – here we call them bankers – that really, deeply understand how this process works. We took that knowledge and broke it down into structured data and we put together this knowledge base that we actively maintain.

An example I’ll give sometimes is your client’s on the phone with you and they’re asking, “I’ve got a rental property, can I count that as qualified income so that I know what I can afford?”

We want to put the right answer at the loan officer’s or the underwriter’s fingertips at the moment. They need to make that decision or they need to give that piece of advice. So we built this tool.

What we found is that AI has really allowed a paradigm shift in how we make information accessible to people. Because one of the biggest challenges you’ve got at the moment is you might not even know what the right question to ask is. You might just sort of know what the client’s telling you, right?

New technology allows an interface with this information that is not just language-based, but conversational, where you’re literally able to converse with an AI that can give you contextual advice and can answer a series of questions on a related topic, clarifying questions that allow you to drill down to exactly the answer very quickly.

We’re testing this and we’re refining it now with some of our own internal team members. What we’re seeing is that we’re able to get the answers about 69% faster, which is absolutely huge. Not just from an efficiency standpoint. Obviously, we want all of our team members to be as efficient as possible, and they obviously want to get their work done faster. But because so many times these are conversations happening with the client on the phone or with the client in person.

If you’re a broker, you’re sitting across the table from a client who’s asking you a question and you don’t have ten minutes to go look it up, right? You don’t want to get back to them later. You want to be able to answer them in that moment and give them an accurate context-sensitive answer instantly. And that’s really what this technology is.

Haas: How are you doing this, like how is this database built? Is this something that’s all internal? And what information are you pulling from to be able to power these tools?

Woodring: It’s a great question because what we find with AI is that really it’s built on top of the actual data you have and the higher quality data you have.

You’ll hear people refer to things like tagging, as in is the information actually organized and tagged correctly so the AI can interpret it. That’s absolutely crucial.

We have a number of team members whose job it is to constantly curate and edit that information. We have editors who are constantly reviewing the guidelines and working with our business leaders because it isn’t just about what’s the compliant thing to do, it’s about what’s the optimal thing.

We worked with Google and their search technology to expose this through kind of a traditional search engine. And then with the rise of large language models, once again, we’re continuing to partner with high-tech companies.

What it’s allowed us to do was make available what I would call beyond even a natural language interface, a conversational interface. In the new experience that we’re piloting right now, you’re not just typing in a question you’ve got, you’re having a conversation with AI and you’re explaining the context.

You’re not starting from scratch every time you’re continuing the conversation with this AI and it’s continuing to learn more and more context about what you need to know and helping you get to the right answer.

Haas: Say, for example, I’m working with a client. I’m in my office, and they’re sitting there and we realize, “Oh, I don’t know whether or not this particular rental income will count,” do I type the question into the computer? Is it an actual conversation through voice? How is that working?

Woodring: So today what we’re piloting is I’m a loan officer. I’m talking to a client, the client asks me a question, but I don’t know the answer to it. Nobody can know everything. I can type it in, ask for an answer, and get an answer in seconds.

We’re piloting technology also that is a banker talking to a client and the client’s telling them this information, the way people talk. We’re seeing some amazing results with being able to interpret the context of that conversation and turn it into that structure of information.

Haas: As you know, people have some natural concerns about artificial intelligence based on the fact that they’re not really necessarily familiar with how it all works. I think that’s really the bottom line for it. What are some of the things that you have in place that may alleviate concern for some people who are just, quite frankly, a little bit scared of artificial intelligence and the whole concept?

Woodring: That makes a tremendous amount of sense. We have for years now had documented AI guidelines that we put in place that talk about what is safe to do and what is not safe to do. Even with the earlier more primitive versions of AI, it’s still important to have this.

Several months ago with the rise of Generative AI and new sorts of capabilities we refreshed those guidelines and reinforced them.

I was very pleased to see the executive order that came out from the White House around AI, very pleased to see the direction they’re going. Although there were a lot of specifics in the executive order, you could read the tea leaves, you could sort of see where they were going with that. I was very pleased to see what they’re contemplating.

I’ve had a number of conversations with our regulators and our investors kind of, once again, getting a sense of where this is going.

We keep the human in the loop. Our humans are ultimately the ones making the call, but AI empowers them by acting as that advisor.

So imagine you are the underwriter making that decision, or you are the loan officer deciding which products to select. What if you had someone in the room with you who had knowledge of every mortgage regulation and guideline and every product and pricing permutation in their head that no human could have? And you had them in the room as sort of your advisor that you could go to to ask questions, to get clarification in the moment.

So the way I see it is this is really supercharging the people we have because one thing that AI isn’t going to replace is the need for human connection and trust. People trust people, they connect with people.

The AI, it’s not there to make the call or forge the connection or give the advice. It’s there to support the person who does with the access to perfect information and access to all the calculations in the moment so that they can answer the question on the spot and not have to say, “I’m sorry, I’m going to have to research this and get back to you tomorrow or later this afternoon.” And they can say, “Yes, that rental income, it absolutely qualifies. Yes, you can afford this home.”

That’s what we want to get to, is giving the tools to the client, the partner or our team members to make the right call in the right moment.

Haas: It must be very exciting to be the chief information officer at Rocket Mortgage right now. And watching all of this unfold before your eyes. Tell me a little bit about that aspect of it. I mean, just as a person, to watch all of this come into play and to be able to provide all these tools for thousands upon thousands of employees must be very exciting.

Woodring: I joined here six years ago. I wanted to be here because what I could see even from the outside is that this was a company that is obsessed with two things. One, it’s obsessed with technology and innovation, but it’s also obsessed with using those things to help our clients and our team members.

I’ve never worked anywhere else that was as fanatical as this company is at constantly making things better for our clients. We invest enormous amounts of time and money refining this and constantly making it better. Our new CEO, Varun Krishna, just joined a few months ago, he’s like-minded.

It’s beyond exciting to be able to roll these things out and see people using them. Pathfinder, for example, is something not only we use internally, but we work with thousands of mortgage brokers and many times these are small businesses. They don’t have a technology team of their own, or if they do, it’s very small. They’re not able to build things like this. And we’ve made it accessible to them as well.

It’s so powerful for them, to give them something they wouldn’t normally be able to have. They’re out there on the front lines and then we see, we hear reports, we can watch some of the sessions happening. It feels amazing.