For commercial mortgage lenders, artificial intelligence (AI) can help automate the underwriting process, source deals and eliminate human error from traditionally manual processes.
NREI spoke with Kim Nguyen, credit risk manager at life insurance lender Pacific Life, and Marc Rutzen, CEO and senior vice president of information technology for Enodo, a tech startup purchased in 2018 by real estate financing services provider Walker & Dunlop.
This Q&A has been edited for style and clarity
NREI: In what areas of commercial real estate lending and underwriting, do you see the most opportunity for the use of AI products?
Marc Rutzen: We see a ton of opportunity in using AI to automate the underwriting process. I think the decisions that are made when you are underwriting are very trainable. People learn them over time doing many many deals, and we’re able to look for patterns and AI is very good at mimicking those patterns and helping people make those decisions. So, what our product does is it puts all that collective experience that you get from years and years in the industry.
NREI: Which are the most useful AI products/programs currently being used? Why?
Marc Rutzen: Real estate is very fragmented. If you think about ownership compared to browsers, Google Chrome is about 90 percent of the browser market, but in real estate, no one owns [such a high percent] of the market. So, you have a lot of fragmentation, different documents in different formats, and AI programs are able to take in different documents in different formats and convert them to a unified template and look at deals on an apples to apples basis. I think those are going to be the ones that push the market forward the fastest.
NREI: How do firms/companies determine whether an AI product is worth investing in? How much are you willing to pay for these?
Marc Rutzen: I’d say looking at workflow or your way of doing things if you’re an organization, and specifically what it’s going to add. I think a lot of people look at AI as some sort of cover-all that’s just going to add value. [You have to] articulate specifically what it’s going to do to your workflow and specifically how it’s either going to remove steps from your workflow, accelerate the process, or help find deals that you otherwise couldn't have. You have to quantify how much, which is very much a case-by- case basis, you have to quantify how much it’s going to reduce the time to get a deal done, how many additional deals and how much money you’re going to make on those deals [that] will result from the introduction of the software. Then base what you’re willing to pay off that.
Kim Nguyen: First of all, we look at what a product can offer, will it meet our needs, and then we look at the costs. That’s how we determine whether we should move forward with the product. So, for SpaceQuant [an AI program for value and monitor the performance of real estate properties that the company uses], for example, what we were looking for is efficiency and automation, and that was provided to us on both fronts with the product. Then, the next step was pricing and if it makes sense for our team, and of course if the benefits outweigh the monthly fees. So, that’s how we justify moving forward with the product.
NREI: What are some of the challenges in integrating AI products/programs into lending and underwriting practices?
Marc Rutzen: Adoption is tough. Real estate people tend to be very conservative by nature. They tend to not want to adopt new technologies because those people have made a lot of money in real estate doing things very much the same way they’ve always done things. That is becoming less and less possible as time goes on and there is more competition. So, I’d say the biggest [challenge] still is adoption and openness to new technologies. The other is trusting an algorithm on a deal that can be tens of millions of dollars. It’s not the same as AI in terms of lead generation software. So, you think about selling a software subscription and you use AI to look at who might be the best lead, that’s very low impact if you can’t close that deal. But if you’re asking AI to value a property and it comes back with an answer that is $50 million versus $60 million, that’s a substantial difference in the ability to get a deal done, and people are a little reluctant to trust AI because of the huge impact it could have on valuations and deal outcomes.
Kim Nguyen: For us, of course, there is always a learning curve. So, I think that was the biggest challenge for us. We were so comfortable using the old product and some of the manual work. So, to move away from that and to try something new and to test the product, I think that was the most challenging for us. Once that’s implemented and we were comfortable with the process, it really helped us with, not just servicing, but all the lending and underwriting.
NREI: What are some of the products currently on the market that you are bullish on?
Marc Rutzen: We’re developing algorithms to source deals and acquisitions, as well as good refinance opportunities using AI right now. So, we’re very bullish on that. I think, in terms of adoption in the near future outside of what we’re doing in our specific lender and broker use case, I think any sort of software that points you in the right direction in terms of what types of deals to pursue for your organization, to the extent the product can use AI to tell you this is a good or not so good deal to pursue for you and what your competency is, it’s going to help people make those decisions better as to where to focus their resources. I think AI is very good at that and you’ll see many more products adopted that allow you to focus your resources more effectively going forward.
NREI: Any other trends in the use of AI and proptech in commercial real estate lending that you are seeing that are significant?
Marc Rutzen: As time goes on, companies tend to be putting more money into proptech, not specifically AI, but proptech in general. We’ve seen a huge consolidation of startups, especially with the COVID-19 crisis going on right now, a lot of startups have not fared so well, and a lot of others have been acquired or teamed up and merged their companies. A lot of them are being sucked up by big companies such as CoStar and RealPage. I think that consolidation is going to be the biggest impact because it’s taking all the data from all these different sources and aggregating it to really large data players. With that data, the challenge of implementing AI comes from data aggregation and putting it into an organized format that can be analyzed. With all that aggregation, it’s going to become a lot more firepower when it comes to building AI models.
Kim Nguyen: I think with the [move to] digital and [to] efficiency comes eliminating human error and this is a big component to that. So, we do not just save time, but we also eliminate human error. Before, we looked at the financial statements and literally keyed in visually what we saw from the financial statements and then categorized them. Now, we’re able to have all that done via SpaceQuant.