UWM’s Jason Bressler says in-house AI agents are changing underwriting, servicing work

Artificial intelligence is reshaping nearly every corner of the mortgage business at United Wholesale Mortgage (UWM), where chief technology officer Jason Bressler said the company is betting on AI agents to automate underwriting support and servicing operations at a massive scale.

Speaking with HousingWire at the UWM Live! event in Detroit, Bressler said the company’s strategy centers on using proprietary technology to increase loan production efficiency, strengthen broker relationships and help independent mortgage originators compete against larger retail lenders as consolidation accelerates across the industry.

Editor’s note: This interview has been edited for length and clarity.

Sarah Wolak: As CTO of UWM, where it’s all about developing in-house tech, can you talk about what managing that looks like from your perspective? How are you organizing not only your priorities but others’ priorities in terms of building and managing it?

Jason Bressler: We’ve got about 2,000 IT team members here, all in the building. I’ve been here for 10 years, and I brought in this “build versus buy” mentality. I believe very heavily in building everything myself. I was the CIO of Guaranteed Rate for 15 years. We only used vendors there, and anytime they would go down or we wanted changes, there was nothing I could do.

When Mat (Ishbia) and I sat down here, we talked about how we were going to position ourselves to be the No. 1 overall lender. I said we need to build everything ourselves so that we can move very quickly and respond to broker feedback. We should build technology to give away, basically for free, to all of the broker partners, to level the playing field against the Guaranteed Rates, the Chases and the Rockets.

We want to tell them they don’t even have to use UWM; if we’re really supporting the broker channel, let’s just level the playing field with technology. Hopefully, that loyalty will create loyalty to send the business to us.

Wolak: With 2,000 team members, how do you manage the workflow and the sheer volume of projects?

Bressler: We probably have over 500 projects actively in flight right now. That’s pretty normal for us. I’ve taken the approach that I’m the CEO of IT. We are an IT team inside of a mortgage company, but we’re also a separate IT organization. I have senior vice presidents for application development, security and enterprise technology who all report to me. We meet and collaborate on a daily basis.

I set the direction and strategy, but I allow the SVPs to work with themselves and really manage everything else. I do something called “Feed the Bear” to be able to manage it. I require that they feed me and all of my direct reports the same information every single day so that we can all work on the same transparent playing field. They know where I’m coming from and why we may need to shift priorities to get a product out faster or support a strategic initiative.

Wolak: Do you find it difficult to recruit for tech positions inside of mortgage rather than a big tech firm? How are you finding and building that talent?

Bressler: I took a very different approach that everybody in tech said I was a fool for. We’re in Pontiac, Michigan, which is a niche area and hard to recruit to. Instead of playing the round-robin game of overpaying for developers from Ford or GM, we created what we call “X programs.” I started training people from the ground up to be developers within my code base and tech stack. They don’t have any bad habits; I teach them exactly what to do.

I expanded that to networking, business analytics and quality assurance. Instead of just finding people from boot camps, I went into the business and took underwriters and operational people. They already know my systems and the mortgage industry. That is half the battle in technology.

Now, more than half my staff have never been in IT before; they came from the business. I don’t really recruit anymore because UWM IT has such a great reputation that people from Google and Microsoft want to come here. But they often don’t add the same level of value as someone who understands the mortgage industry.

Wolak: How do you measure whether AI is augmenting jobs versus reducing headcount needs within the business?

Bressler: Technically, it’s doing both. One of the secrets to our explosive growth is that because we built our systems in-house, we can move very quickly. Take our underwriters: We bring them into an X program, and because our technology has so many guardrails, they don’t need to know as much as a retail underwriter.

We were at the forefront of AI and signed a partnership with Google four years ago to build large language models and data extraction. We can now measure not just what an underwriter can do in a day but their accuracy. A typical mortgage underwriter at a place like Guaranteed Rate can do four loans a day. Our underwriters currently handle a minimum of 16 underwrites a day, and by the end of the year, they’ll be well over 30.

I haven’t eliminated the need for people, but I’ve made it so the easiest parts of underwriting are handled by AI, so our underwriters only handle the most complex cases.

Wolak: How do you balance creating transformative functionality versus “AI washing” or just creating a basic chatbot?

Bressler: We don’t talk about what we do from a technology standpoint for marketing; we just say we have invested heavily in AI. We keep it in-house and just release it. A lot of times, people don’t even know it’s AI; they just recognize they have less work to do.

For brokers, we basically force them to use it because we give the technology away for free. If they want the tool, they use the AI. Take Mia, for instance. She is automatically making calls on behalf of the brokers.

At first, there was thrash — brokers didn’t want someone they didn’t know calling their clients. But then Mat would get on stage and show testimonies of people getting seven deals in a week because Mia called 25 people for them. Because it’s free and brings them business, they adopt it.

Wolak: Can you give an update on the different AI agents like LEO and Mia? Do you have any updates regarding servicing tech?

Bressler: LEO is an AI agent that analyzes loan estimates (LEs). A broker can upload an LE, and within seconds, LEO reads it, shows everything wrong with it, grabs market data and tells the broker exactly how they can win the deal based on their profile.

Mia is different because of the scale. She can make 100,000 concurrent calls at any point, including inbound calls. Every broker partner has their own inbound number so Mia can answer for them. It’s all generative, so her conversations change based on who she’s talking to.

She can talk about Fannie Mae and Freddie Mac guidelines or UWM overlays. She can even see where a loan officer is licensed and offer to transfer the call.

We have a new agent like Mia, called Nora, who will answer all of our servicing calls. She is a full IVR (Interactive Voice Response). You can call Nora and she can take payments, go over escrow balances or handle payoffs with the ability to transfer to a live human being if needed.

Wolak: You spoke at HousingWire’s AI Summit last year. During your presentation, you mentioned that brokers should use AI to safeguard their independence and skills amid rapid M&A activity and change. Does that sentiment still stand?

Bressler: It made my blood boil sitting at a conference with CEOs from Rocket, Redfin and Mr. Cooper right after their big merger. They were creating this utopian scenario where they weren’t actually going to go after all the brokers, but I was telling the brokers, “You worked so hard for your business. Why would you give it away to an entity whose whole goal is to keep everything within their own ecosphere?”

Independent mortgage brokers need to use these tools to take control of their destiny. Our approach is to get these AI tools out to our broker partners so they can keep that borrower for life. That’s also our servicing play with Bilt.

We partnered with them so the loan officer stays top of mind on servicing statements for two or three years, with the ability to send the borrower dinner or a gift. We want to keep our loan officers top of mind.