After seeing the success in servicing, we asked: How do we bring that to our operations? And once we get it with our operations, how do we move it to the front of the funnel?
Our Loanbridge system is bringing that same methodology to our operations team on the origination side that we’re using on the servicing side, and then we’ve been working incredibly hard, principally with our retail sales force, both our NFS team and our call center team, to take their workflows and standardize them as well, and then integrate them with our marketing journeys.
SW: What are some of the other workflows where you are combining AI with human effort?
BP: So we introduced, again on the on the servicing side, but we’re now pushing it forward into our originations area, what we call servicing exception. Prior to servicing exception going live, when a loan went forward to our servicing system, we had a human being who would look at the data in the origination system, look at the data in the servicing system, look at the underlying documents and do stare-and-compare.
Now, with the machine learning tools we have, with the way OCR has evolved — with a light touch of AI and its ability to understand documents — we can put that together and really offload the stare-and-compare components to the machine. It focuses the employee on the exceptions and our cost per boarded loan has decreased by almost 50%. The machine never takes a vacation, it never gets tired, it never gets angry. Does it make a mistake? Yes, but it makes the same mistake over and over and over again, so you’re able to see it.
Something that’s was a little bit surprising is that we found that it’s better at reading handwritten documents than our human being was. So, we’ve seen our ACH errors go down by 67% — so we’re getting two thirds better. In the forward space, if the borrower is late with a payment, we charge the borrower a late fee. In the reverse space, if I’m late delivering the borrower’s draw, I pay them a late fee. So now, by reducing these ACH errors, we’re seeing a dramatic reduction in our late fee payment to the consumer. The consumer is happy because they’re getting their money more quickly, without errors.
SW: Do employees who were used to the traditional stare-and-compare processes find it hard to trust the AI processes?
BP: In my career, it always seems like it falls into thirds. A third of the folks adopt the technology right away. Then a third of the folks say, I’ve always done it this way, I’m not sure I can trust the machine, and they are slow to adopt. And then you have that middle third, who are waiting to see which of the other camp rules the day. And it often comes back to leadership.
We experiment with a lot of these things in servicing first and one of the reasons we do that is because in my servicing group, it’s more like two thirds of the people are willing to adopt these technologies. Over the history of Longbridge, the leadership team has been leaning in more and more to be a tech-forward company. Our CEO, Chris Mayer, identified early on that if we’re going to be a scalable industry, it can’t be limited by what one human being is able to do and what they know. We have to systematize this knowledge and make repeatable processes and use these technologies to make us more productive. And that’s been the DNA of the company.
SW: When it comes to tech and automation, are there any limitations that you have with reverse?
BP: There aretwo big things. One is that FHA and Ginnie Mae currently don’t permit eNotes for their loans. That doesn’t mean we can’t do it on the proprietary side, that doesn’t mean we can’t do a hybrid closing, but if we want to deliver the full eClosing digital experience, we’re going to have to work — and we are working —with FHA and Ginnie Mae to move that ball a little bit forward. And if you think about our borrower, they are more likely than the general population to have conditions which make repetitive signing more difficult for them.
The other piece is that we have not yet introduced automated underwriting into the reverse mortgage world. It’s something we’re working on and we’re very excited about and I think it can help our underwriters be much more productive and steer them to places that might need the more personalized intelligence of an underwriter. So that is another place where the technology needs to be applied.
SW: What keeps you up at night?
BP: I sleep really well. But the only thing that I would say keeps me up at night is information security. We have an immense responsibility given the information that our customers provide, to keep it secure and protected. The threat actors, they only have to get lucky once. We have to defend our company every single day, every single time. But we do spend a lot of time there and we’ve invested an incredible amount of resources in protecting our environment, improving it, and we continue to do so as we see new threats emerge.
I always counsel people: be very careful that if you are going to the internet and accessing something that’s just publicly available, everything you put there is publicly available, and they’re going to use what you’ve given them to train their models. Just be very careful and thoughtful about all of that.