Mortgage startup Copperlane targets origination inefficiencies with AI loan officer

As lenders weigh how to adopt artificial intelligence under evolving fair lending and compliance expectations, Copperlane is pitching a full-fledged AI “employee” rather than another point solution.

The startup, founded by 21-year-olds Athan Zhang and Brianna Lin, recently raised a $4.1 million seed round to build Penny, which the company describes as an autonomous AI loan officer that handles borrower intake, answers questions, and feeds underwriters and loan officers a steady stream of context.

In a conversation with HousingWire just days after Copperlane announced the seed funding, Zhang and Lin offered a glimpse into their startup and why Penny differs from today’s AI point solutions.

Editor’s note: This conversation has been edited for length and clarity.

Sarah Wolak: Let’s start by talking about Penny, which was described in your press release as an autonomous AI loan officer. Can you talk about the tasks it’s able to do independently and, conversely, where humans still need to be in the loop?

Athan Zhang: We call it autonomous because Penny has the functions to be autonomous, but it can be a copilot or an autopilot. It really depends on what our customers want. The way we typically pitch it is, think of Penny as an employee, just as you think of a loan officer assistant. You can have your loan officer assistant do more, or you can have them report more to the loan officer.

In terms of active capabilities, Penny can answer borrower questions. She has her own phone number, and borrowers can communicate with her via iMessage, SMS, or even call her if they want. She can provide suggestions and recommendations for program eligibility to help loan officers understand what a borrower might qualify for.

More importantly, she gives loan officers context around the borrower. Penny is one of the things that interacts with the borrower throughout the intake stage, so we can surface the most important pieces of information to the loan officers. That gives them more bandwidth to understand more customers at the same time, while Penny works to do things such as figuring out why parts of an application don’t line up, verifying documents, verifying numbers and putting together a briefing for loan officers.

Brianna Lin: One thing I’d note is that Penny, as this AI employee, follows the borrower throughout the entire journey of their application. Penny engages as soon as the borrower starts an application and she’s involved until the file reaches closing. Penny is engaged with the borrower throughout the experience and she fully understands the context, the history and the borrower’s background, just as a real loan officer would get to understand a person.

Wolak: How did you develop this idea? Was it a class project that evolved or something you both had a personal interest in?

Zhang: It was not a class project. For background, both Brianna and I are from the Washington, D.C.-Maryland-Virginia area, and both of our parents have worked extensively in the mortgage space. So growing up, we were always exposed to these problems.

When we went to college, this wasn’t necessarily what we set out to do. I was a computer science major; Brianna studied computer science and real estate. We had other ambitions, and we were always very entrepreneurial. We ended up joining this program called Y Combinator, which is essentially a startup accelerator that has produced companies such as Airbnb and DoorDash.

It’s a three‑month program that gives people like us a chance to try building what we want to build. We met each other through Y Combinator and actually realized we had both a shared background, but more importantly, a bigger understanding of how AI could be applied to mortgages.

Compared to many people in the mortgage industry, we understood more about the frontiers of AI, which is important for governance and alignment. On the flip side, compared to a lot of our peers, we just knew more about mortgages than the average person because of our proximity and what we grew up with.

Wolak: Can you explain how you saw inefficiencies in the mortgage space without having direct industry experience yourselves?

Zhang: The actual story starts with a conversation with my mom. I generally know what she does — she works in the secondary markets in risk management. She told me a lot about the quality of data that gets to risk by the time it reaches the secondary market.

By that point, many of these loans have been converted into numbers, and there’s a lot of variance in those numbers. Part of that is a data problem. I thought it was interesting how these loans get converted at the top level into just numbers.

I followed that problem downstream and ended up at origination, which is the process where these applications become those numbers. I talked to a lot of people. One of our first customers, before they were a customer, let us spend a week in their office.

That put me on the ground floor with loan officers. I had perspectives from people in the secondary market and from the first people talking to borrowers — the loan officers. We realized that’s where many of the problems and inefficiencies appeared, in how these applications were made.

Lin: We first found the idea through our families’ experience in the industry. Overall, our approach to learning about mortgages has been to be very aggressive and very boots‑on‑the‑ground. As Athan mentioned, we would go on-site with customers and sit directly with their loan officers, watching them work to learn the workflow and see the inefficiencies firsthand.

We did a lot of in‑person meetings with potential customers, went to conferences, and visited local banks and credit unions in San Francisco. We were very aggressive about learning upfront, and that gave us the confidence to work on the problem.

Wolak: The mortgage industry traditionally lags on technology, but there are companies with internal AI or AI assistants. For instance, United Wholesale Mortgage has Mia, which can answer calls and contact borrowers for follow‑ups if the system flags them as eligible for a refinance or something else. What makes Penny different from the AI that’s already in the market?

Zhang: This is the classic build‑or‑buy question. In any industry, whether you should build or buy depends on the technical hurdles of what you’re trying to build.

AI has lagged in the mortgage industry partly because there isn’t much regulation around it yet. We’re only starting to see early pieces, such as the Freddie Mac 1302 that came out in March. It’s not even a concrete piece of legislation on AI.

There’s a lot of cold feet around AI because we don’t really know what safety and governance will look like. That’s part of why Brianna and I got into this. We come from AI research backgrounds and environments where you’re exposed to alignment work. Alignment as a process really started gaining traction only recently, so we know it’s still early, which some industry leaders might not fully appreciate.

When we build Penny, we’re positioning ourselves with the expectation that regulation will eventually come down. We have the foresight to anticipate what it might look like and to build Penny in a safe way. That’s important, and it’s something many smaller or even midsized shops don’t have the expertise to do. That’s our right to win, and we intend to exercise it.

Lin: A lot of our competitive edge comes from the fact that, if you look at most tech players in this space, they’re mortgage people who spent years in the industry and then pivoted into building tech. Athan and I come from AI and tech backgrounds. We’ve studied computer science and AI intensively, and now we’re building for mortgage and learning the space very quickly.

We have a better understanding of the models and how to deliver them in a safe and compliant manner to the industry. I’d also note that while many competitors claim they have internal AI loan officers or AI employees, from what I’ve seen, these are still point solutions that address very specific pains, such as document parsing or Q&A. The systems are not fully agentic. They’re not truly functioning as a real person would, and they don’t think in the same way that anyone on a team would.

Wolak: To achieve what you’re describing — thinking and acting like a real person — how long did it take to implement everything? And with compliance, some people might argue that mortgage veterans building AI know how to best address fair lending concerns. How are you building those considerations into Penny?

Zhang: It’s not a set duration. This is always an ongoing process. In terms of legislation, it’s not as if we’re unaware of fair lending, ECOA, TRID and all these rules. If we didn’t know about them, we wouldn’t be running a very good company. Our core competency is AI, but we’re working in a vertical with its own domain requirements.

That’s why we’ve been very aggressive about filling what is most likely our weak spot. We’re very aggressive about being on the ground and learning the problem. We know our competitors have more experience in mortgage. In my view, it’s easier to learn the domain by being there and understanding that this is what you need to learn, compared with someone who has to spend more time learning AI and the technical side.

Wolak: With the seed money you’ve secured, what are the next steps and milestones? What should investors and industry observers expect over the next 12 to 18 months as you deploy that capital?

Zhang: One of our biggest bottlenecks has been engineering. We have to think about integrations with different software platforms, while also running a lot of research and building systems to make sure Penny is safe and aligned.

It’s been very intensive. We feel the pressure because we have a decent number of customers and that number is growing. We’re using capital to reinforce our ability to deliver more feature requests and improve the product for existing customers, while also supporting what we expect as we add more customers.

The other part of the capital is reinforcing the growth motion. A lot of what we do now is simply being present at conferences, but right now that’s just Brianna and me. We want to invest more in those relationships with organizers and lenders and be more present in the space.

Lin: Growth is the major theme. That includes hiring more people and doing what we’re already doing, but more aggressively — being more involved at conferences and seeing more customers in person. Even as we grow the team, Athan and I will continue to be boots on the ground for at least the next year or two.