Jason Bressler, UWM’s chief technology officer, said the wholesaler is one of the first companies to leverage Google’s Gemini Flash 1.5 model to enhance underwriting automation. Underwriters at UWM have already increasing their capacity from underwriting six loans a day to 14 (the industry average is closer to 3 files per day).
Bressler said UWM has been working with Google Cloud since it built out its Bolt underwriting platform about four years ago.
“We decided to really take the plunge into everything AI probably about two years ago, and we’ve been working with Google on their different extraction modeling, their platform, their data lakes. We’ve got a pretty large partnership with Google at this point.”
The partnership will also explore the use of Google Cloud infrastructure to improve scalability and security. This will lead to a more seamless and expedited mortgage experience for 50,000 mortgage brokers and their borrowers.
UWM, known for its build-don’t-buy tech model in mortgage, will also collaborate with Google Cloud to enhance data-driven insights, enabling more personalized loan recommendations and more intelligent tools to instantly identify the right mortgage for a borrower.
Oliver Parker, Google Cloud’s vice president of AI, said AI is a great use case for summarization of documents in mortgage. And they’re only just scratching the surface.
“One side is helping on the back office side, the origination side and providing brokers more data quicker,” he said. “The other is how you think about things on the front side. With a lot of the automation, you start to think about voice on the front end. You think about a full customer journey. Jason and I have often talked about how much more we could be doing here…that we can do when you start thinking about experience for a home buyer as well as what you could truly do for brokers and putting them in a position to be the most successful.”
The tech giant is also working with capital markets, insurance companies.
“There is sort of a full scale set of capabilities that we are now beginning to see are incredibly valuable and starting to get what I would call sort of traction at scale,” Parker said. “Twelve months ago, you would see pockets of what I would call experimentation. You see proof of concepts, and they would light up certain things…but we’re now moving more towards these systems starting to scale and actually being applied, especially externally.”
