How underwriting's thorniest issues could be solved with AI

September 5, 2024 5:00 AM

While lending volumes may have fallen from the heights of a few years ago, a mortgage fintech is rolling in, looking to tap into the opportunity presented by the surge of generative artificial intelligence.

Formed in 2022, Tidalwave plans to employ AI that provides real-time answers for loan underwriting that will lead to a more equitable origination process. Tidalwave was co-founded by longtime technology executive Diane Yu, who also heads the firm as CEO.

Tidalwave rolled out its first product this year, an AI-powered underwriting engine called Solo. With integrations to Fannie Mae and Freddie Mac automated underwriting systems within the pre-approval stage announced this summer, Solo has already attracted two beta clients. The company hopes to build Solo, which Yu described as a copilot to serve both lenders and borrowers, into a tool that will eventually serve all parties through closing.

Reggie Ige

With a background in digital advertising technology, Yu was an engineer at Doubleclick, which was eventually purchased by Google. Several years later, she co-founded and served as chief technology officer at advertising startup Freewheel and continued there following its acquisition by Comcast in 2014.

Pivoting to mortgage banking. Yu took on the chief technology officer role at Better in 2021 before she set out to create Tidalwave.

In a recent conversation with National Mortgage News, Yu talked about the path that brought her to the mortgage industry and her decision to start a new business in the current market environment, while also discussing the vast undertaking of introducing a new product and expectations for the future of AI in loan underwriting.

Portions of the interview, edited for length and clarity, follow.

Can you describe your tech background? How did you first get into the field?

I started my career at this point, 27 years ago, as an engineer. I have always been an engineering tech executive, starting as an entry-level engineer. I spent 10 years at Doubleclick. I only left Doubleclick after the Google acquisition. I was running their engineering team. That's when I decided, basically I'm going to leave and build my own company, Freewheel. So I built Freewheel with two of my other partners.

You had the tech background, but what attracted you to mortgage exactly?

I am based in New York. There are two industries in New York. One is advertising, the other is financial, so I always had my heart in the financial industry.

After many years in advertising — when I think that I wanted to do something brand new and the next industry that could be interesting for me — fintech would be my number one choice. That's when the founder of Better, Vishal Garg, reached out to me.

Being at Better, I got to learn everything about the mortgage industry. I realized how outdated the tools have been in the mortgage industry and also realized how painful it has been and continues to be for borrowers to get a mortgage.

So I want to use my knowledge and my capability to change all of that for the entire industry, not just for a single lender. That's my mission — basically creating a tool to change the mortgage industry.

The other one is also very personal for me. I came from a minority background, and I'm a female. I remember 20 years ago when I first got my first house as a female homeowner with a minority background, the loan officer asked me multiple times whether I wanted to put my husband's name on the mortgage application.

It is personal for me to have the capability regardless of an individual's unconscious bias. As a minority, as a female, my mortgage application journey can always be treated fairly just based on my financial information — not based on the color or my status, or or lack of status. Having the capability of using the machine to evaluate that, it's the way to provide the most fair treatment to borrowers.

The industry has changed a lot in just two years. What drove you to create the company at that particular time? Was your plan to form the company independent of what was going on in the market?

I left Better in April 2022. After I left, I took a break. In November of 2022, all of sudden ChatGPT came in. I have been following ChatGPT very closely. That's when I realized, with the ChatGPT capability, there is a possibility to combine the technical platform, the automation, along with the ChatGPT-like conversation to create this tool to be able to benefit the industry. That's when I found the raw ingredients were ready.

November 2022, that's really when we started to form that idea and see whether it's possible. Then in March 2023, that's when I started taking money. I took money from the investor who basically invested in my first company. That's when we started.

The reason I started to build a mortgage company when nobody was investing in mortgage technology at that time was because I see the pain the consumer has to go through. I feel with the latest technology available now, an ultimate tool can be created to solve that problem.

The mortgage industry overall has been slow to adopt digital tools. How challenging is it to start a fintech with that kind of history?

It's not easy. Mortgage underwriting is super complex even with our capability.

You know, I read through Fannie's underwriting selling guide multiple times. Fannie's selling guide is 1,200 pages. I read through every single structure. And Freddie is even thicker. It has about 2,500 pages.

I read through both of them because I want to make sure what we build has the capability to automate, basically creating our ultimate solution to make all this pain go away. It is pretty complex, and I think it's the complexity stopping the industry from evolving quickly.

Tidalwave starts at the right time because the technology breakthroughs make it ready so that we can take advantage of all the latest technology has to offer. On top of that, using our experience to build a scalable tech platform, we can quickly evolve and build that capability.

Apart from reading thousands of pages of selling guides, how did you ensure Solo could meet regulatory and compliance rules?

That's why our initial product is a copilot. We test around everything, but the loan officer and underwriter can still evaluate everything we've evaluated. We're not relying on Solo to make this decision on its own.

At the same time, Solo has the capability to collect everything in real time. Any type of compliance audit can come back and see everything Solo has made a decision or recommendation based on the interaction the loan officer has with the borrower. They can track everything and ensure compliance.

Right now you're using Solo in the preapproval stage. But what else can it do? Do you have a plan for how you want it to grow?

If everything goes well the mortgage application coming out of Solo should be error-free. Then a seasoned underwriter, when they look at that, they're saying, "You know what? All the documents have been verified, all the questions asked and answered and everything looks good. Now let's just draw the doc."

My hope is that it will save a lot of effort for the lender. They're really just doing the last-minute check without the back and forth between the processors and the borrowers asking for additional information, because we already asked for all the additional information an underwriter would have to ask when they look at the financial information about this individual.

We are not focused on the process improvement. We're actually focused on automating everything and creating the copilot. The copilot idea existed after ChatGPT became a possibility. You can see that our system is actually a purposely built AI engine, powered by basically our mortgage engine, plus the AI capability.

What do you think about the future potential for AI in the mortgage industry?

I think AI is really just starting. I understand a lot of people are scared about AI capability because it feels like it's a black box. What I would say is it doesn't have to be a black box. In order for the AI technology to be applicable in mortgage underwriting, it has to be transparent.

That's why we purposely built Solo to be the transparent white box, not like a OpenAI ChatGPT solution. Our Solo is equipped with underwriting guidelines, which we sourced from Fannie and Freddie.

AI as a technology is a starting point. Everybody feels like, in order to get AI-like capability, you have to use OpenAI, which is a gigantic black box. You don't know what will come out. It's not predictable.

I can see that several years down the road, mortgage application utilizing AI would create a capability where the information source is well controlled in providing 100% transparency about that AI engine's decision making and making sure the compliance officer, the loan officers and the underwriters can write through the inherent decision making and then be comfortable of that decision making. That has to be the first step.