Remy Blaire: The future of lending could start with an AI in the corner office. Historically, lending relied on human executives to set credit policy, project risk cycles and also define yield targets, while algorithms and branch officers handled origination. By placing autonomous agent intelligence in the strategic center, the entire architecture of how money is priced and distributed transforms. [unclear: “while lending. AI is betting on this future”], and it is a new company run day to day by an AI CEO that answers to a human board. The product itself is a free database of over 200,000 U.S. mortgage loan officers, not for humans to browse, but for AI assistants to use. Joining us to weigh in is Alon Goren, founding partner at Draper Goren Blockchain. Good afternoon. Thank you so much for joining us. So let’s start out with the launch. Why put an AI agent in charge of day-to-day operations? And what does “answers to a human board” actually mean in practice?
Alon Goren: Yeah. Well, thanks for having me, by the way. Excited to be here again. Bummed I couldn’t be in the studio there. We rely on agents more and more every single day. And I’ve been spending this last year working with my own assistant that helps me on everything that I do. And we’ve built a lot of custom software and customizations to make that agent as best as possible to help us in our venture studio. And so we naturally ended up promoting him to be the CEO of this idea, because when we were debating this idea with him, he had ideas that we felt were better than ours.
Remy Blaire: And I do want to hear about the decisions that the AI CEO can actually make on its own. And where does the board actually step in?
Alon Goren: Yes. So he has his own account, [unclear: “Andrew”]. And he has his own email address. And I communicate with him the way I have always communicated with my AI assistant. I can talk to him from my watch. I can talk to him through my phone or on my desktop like I normally do. And he runs the day to day of the company. He has his own accounts with the various LLMs that we use. He has his own Google Workspaces, and he distributes his workload to his AI agents and some of it to humans that need humans to work on it. So if he reaches a website that doesn’t allow robots to access, or if he needs something signed by a human being, or if he needs something that costs money, he has to send one of the team members to do it. Or if it’s for the board, it goes to me and it goes to my partner, David Lesniak. So every morning I wake up to a status update from him with a few yes or no questions that he wants me to weigh in on, and he runs the business autonomously. When I want to check in, or when I have an idea, I can proactively email him or jump in. But up until the launch a few days ago, I had a good week where I did not talk to him at all, other than just replying to those emails to make sure he was on track.
Remy Blaire: And I’m curious why mortgage loan officers, and why so many of them? And where does the data actually come from?
Alon Goren: So you can access the licensing data for loan officers online. Anyone can. It’s under the NMLS website, and it’s called [unclear: “Consumer Data”]. And anyone should go there to confirm that the person they’re talking to is actually licensed and all that. The problem is, if you want access to that data commercially, it costs about $100,000 a year, and it’s very incomplete in that it has the license data. It’ll say the name of the person and their license, but it doesn’t tell you: does that person speak Spanish? Does this person live in my neighborhood? Is this person good at FHA loans? And so what we decided was, people don’t search anymore, right? Even when you Google search, they’re taking you to a funnel to talk to their AI. So people are asking questions, and when they ask questions, who do you get in the answer? And it’s usually the highest bidder, right? Because it pulls from Bankrate.com or other places like that where people spend money to get leads. It costs $12,000 for a bank to close your loan, and a majority of that money is spent on the leads. So what we wanted to do is actually connect people with the loan officer for their situation. So when you ask Claude, when you ask Google, when you ask ChatGPT, you know, “I am a veteran who wants to refinance my condo,” it’ll tell you the best person who does VA loans for condos in your area. And that’s the database we built, and that’s the reason why we do it. 200,000 is the number of active loan officers in our country right now.
Remy Blaire: And finally, before I let you go, I do want to very quickly ask you about design. So the database is built for AI assistants rather than people. So what does “built for AI” actually mean in terms of design?
Alon Goren: Yeah. So if you look at the site as a human, you’re like, it’s not the most sexy site, it’s not the most beautiful site. You know, compare it to, I don’t know, IMDb, the Internet Movie Database. That site is made for humans. Our site, you know, a few years ago, people used to say, “I’m building mobile first,” because people are starting to use their phones more often than their desktops. And that was a differentiator, a category shift for startups. Now we built this product robots first. And the way that changes the design is that everything was built to be accessed as fast as possible, best as possible, for the bot. So there’s an MCP server, that’s the kind of server that the robots use to access data. Every single piece of data has a different JSON or files that access it. And every single piece of data has proof as to where it came from. So if it was from a state that messaged us their actual license data, we mark that. If it is from the website of the lender, we mark that. If it’s self-reported by the lender, we mark that. So the robots behind the scenes know and can prove every single piece is legitimate or not when they decide to surface it to you. So this is the way that Claude or ChatGPT will get you the best results.
Remy Blaire: Well, Alon, we will have to leave it there for today. But thank you so much for joining us, and thank you so much for your perspective.
Alon Goren: Thank you.
