Artificial intelligence (AI) is a buzzword in virtually every industry right now. And, in the mortgage industry, AI will play an instrumental role in helping loan officers to be more efficient, according to Nima Ghamsari, Blend‘s co-founder and CEO.
“Some of these borrowers have hundreds of products that they can choose from. How is an LO supposed to keep that in their head? It’s too much context and this will be a supercharger for them,” Ghamsari said in an interview with HousingWire.
The most difficult part, however, is building the technology. To be effective, the AI tool would need to understand each consumer’s unique financial situation and all of the mortgage products and programs, Ghamsari explained.
Another crucial part of building the tech is having “a lot of people” using the system, which is an advantage for Blend, according to Ghamsari. Blend’s mortgage banking software processed 23.2% of the total market originations in the second half of 2022, up from 14.5% in the second half of 2021.
Read on to learn more about the opportunities and challenges that AI poses to the industry, what the company has to say about the risk of getting delisted from the New York Stock Exchange (NYSE), and insight into Blend’s roadmap for profitability.
This interview has been condensed and lightly edited for clarity.
Kim: Blend has played a critical role in powering about a quarter of the mortgages that originated during the refi boom. It seems like the next big wave is artificial intelligence. How is Blend preparing for the era of AI in the industry?
Ghamsari: I think on the AI piece, it’s about combining an understanding of what the client is trying to accomplish.
Now that a system can understand the essence of the question the consumer is trying to understand, it (AI) can actually do that work for the LO in the background, and then when the LO shows up, that work is already done.
Some of these borrowers have hundreds of products that they can choose from. How is an LO supposed to keep that in their head? It’s too much context, and this will be a supercharger for them.
In order for that to happen, you have to have a lot of people using it, which Blend does. You have to be connected to all these data sources and internal systems to both the customers and etcetera — and we are. You also have to be something that the LO uses on a regular basis.
So we’re in this position where I think we can really help the industry, and particularly LOs, who are trying to make things work for consumers.
Kim: Then I assume AI could also take that extra step in correcting some information for LOs that they provide for borrowers?
Ghamsari: I think there’s a separate piece, which is for efficiency. There’s a lot more opportunity to understand what’s required to be done on the loan file after it’s already gotten through the space.
Understanding those requirements and documentation, and actually understanding the data and saying, ‘we need this additional piece of information,’ or ‘we extracted this information, and now that loan looks like we need to change something about it to correct it for whatever reason.’
So I think that’s a separate opportunity that I think is also potentially pretty compelling. Almost think about it as like a co-pilot for an underwriter. That same exact capability could exist.
Kim: Are there any features that Blend is trying to build as the industry gets more involved with AI?
Ghamsari: Nothing I’m prepared to share today, but we are definitely looking very closely at the space. Blend has some unique things — like how many people use our system is very important, and all the systems we’re connected to are very important. All the history of data we have is very important.
So Blend is that interface between LO and the consumer today for a lot of our customers.
Kim: I’m curious how in what ways AI can help with homeownership and tapping into a potential customer base.
Ghamsari: I think that’s the area of the market that will benefit the most from AI. Most people who are first time homebuyers, or in underserved markets, don’t understand all the products and all the things that a bank could help them do or a lender could help them do.
Imagine you’re a lender or an LO or a bank who is trying to serve the mass market. In order to serve them really well, you have to be able to do that work on every file, and it’s just not scalable to build something that requires LOs to spend 20 hours on every file bill to answer that question.
So that’s why I think the co-pilot model is especially important here, because you still want that borrower to have that LO. But you want that LO to be able to do a lot less work to serve that customer.
Kim: The big issue when it comes to AI is getting rid of that bias in machine learning. How can we tackle that?
Ghamsari: I think this is where having a human in the loop is important. There are programs – whether it’s the government, or banks – in place to allow for these higher LTV or lower-income borrowers to get access to credit.
I think what this (AI) does is — in theory — this unlocks the ability to make every specific situation as personalized as possible, which is what an LO would do if they could spend 20 hours in every file.
Kim: Are there any other challenges you foresee other than the bias factor in AI?
Ghamsari: I think the technology is extremely difficult to build. It’s not just taking some large language model or adding open AI to your platform. Building something that can understand the complexity of a consumer’s financial situation and understand all the products and programs that are out there — and understand the intent of the consumer. All three of those things are actually extremely complicated.
Kim: I want to shift focus to the notice Blend received from the NYSE about not being in compliance with the bylaws. Blend’s stock price has been up since its first quarter earnings call, trending closer to $1 level led by revenue above target and shrinking operating loss. How confident are you that Blend can meet NYSE’s bylaws?
Ghamsari: We have a plan to meet it. I feel good about that plan.
Obviously, I think there’s just general challenges. We are growing market share a lot right now and helping our customers a lot. What I’ve always said is – first and foremost during times like this – it’s not about selling customers new things. It’s about being there for our existing customers.
I want everyone to use it (Blend) so they can benefit. Let’s get a prescriptive roadmap for our customers to help them, and all those other things will take care of itself.
Kim: I’m curious what the board’s response was when Blend received that notice.
Ghamsari: We knew it was coming. It wasn’t a surprise to us and we had a plan. We wrote a letter back to the Stock Exchange saying here’s our plan.
So we were prepared, we knew it was coming, and we had a plan to deal with it.
Kim: We are in a downturn of a cyclical business. I remember you saying that in Q4 of next year, Blend will have positive operating profit numbers. What are some of the crucial external and internal factors for Blend to recover its share price, which once traded above $20?
Ghamsari: We said net operating loss will be less than $20 million in Q4 of this year, and then we’ll be profitable next year. We’re going to hit that plan.
We have different levers in our business. We have a lot of discretionary investment that we’re doing for the sake of our customers. Blend has the balance sheet to do it. We have the customer base that needs it, and will stick with us. So we have to keep investing; that is our job.
If the macro gets materially worse, we’ll pull back on some investment. But we have now scoped it out to where we feel really good about that.
In terms of getting the stock price back to a certain number, all I think about is, how do I keep making our customers get more value from us even for things they don’t pay for? How do I use that to get customers to want to do more with us? Because if we make them more successful, they’re going to want to do more with us.
