Ben Miller of Fundrise: 5 Questions

Having co-founded investment portal Fundrise in 2012, Ben Miller is no newbie to proptech. As a CEO, Miller has seen many changes in real estate technology and nomenclature over the past 12-plus years, and such trends are only accelerating with the introduction of artificial intelligence.

For Washington, D.C.-based Fundrise, along with the rest of the industry, the rise of AI is changing everything. No wonder then that in late January 2026, Fundrise unveiled its newest product, RealAI. Developed over the last three years, Real AI uses a trillion data points to automate real estate analysis for large-scale multifamily owner-operators, said Miller.

Having amassed $6 billion in customer investments, Fundrise claims to be one of the 50 largest real estate private equity investors in the world by total annual outlays — deploying more than $1 billion of capital annually in 2021 and 2022. Its portfolio is largely composed of more than 20,000 residential units and e-commerce-centric industrial assets.

Commercial Observer spoke with Miller last week about the evolution of proptech and real estate investment finance, as well as the impact AI is having on every aspect of the industry, including Fundrise’s expensive efforts to develop and successfully market RealAI.

The following conversation has been edited for length and clarity.

Commercial Observer: You have made big claims about Fundrise having created “a new real estate world” for individual investors. How has that happened and how do you define it?

Ben Miller: To invest in big real estate buildings like the World Trade Center, you previously had to be an insider or even an institution to invest in that kind of real estate. In 2012, we changed that. 

We were the first to ever make it so that anybody could invest in institutional real estate. The market for democratizing access to private markets is now normalized. Our focus has been mostly multifamily, but we’ve done other asset classes, including build-to-rent, industrial, and event technology venture capital.

It’s been well over 15 years that I’ve been doing this, but the name of what we do keeps changing. When I first started, I just called it an online syndication. But there’s a long financial history of access to markets. We spent 15 years at the Securities and Exchange Commission creating regulatory wrappers or frameworks that allow anyone to invest into private technology companies or private real estate. And then we also built technology to make the cost very low.

Fundrise has now brought RealAI to the market. What is it and what do you expect your clients will achieve using it?

AI is going to eat half the white-collar jobs in the world. It’s obviously eating journalism and customer service. It’s going to eat the real estate profession as well. It can do the job of a real estate analyst. 

And what we created with RealAI is an artificial intelligent real estate analyst. It does acquisitions, analysis, and asset management analysis. It can produce pro forma and investment memos. It does the work of a junior analyst. I would think of it as a real estate AI co-pilot for the real estate financial professional.

Everybody makes grandiose claims about their AI. What is your AI based on? Is it proprietary?

What makes it unique is that we built a trillion data points, including every property and person in America, with lots of macro-economic data. It’s a really smart AI model like Anthropic or ChatGPT, which are smart generalists but don’t know anything about real estate. RealAI is trained up to give it a master’s in real estate, so that it can now do what a real estate professional needs, and it knows as much or more than almost any real estate analyst. You’d have to be 10 to 20 years in real estate to know more and be better than it.

How was RealAI trained on real estate data that’s notoriously fractured, inaccurate and hard to parse? How did you clean and vet the data?

We have a team of research and data engineers, data scientists and machine learning experts, software engineers and product managers. Fundrise has more than 200 people, and we’ve been working on it for a long time. It cost us millions and millions to develop.

Vetting the data was the hard part. That’s why it took three years. Data engineering is the long-lead item. So we had to build essentially a virtual model of the real world. And not just the physical world, but also the mental models of how real estate people think about investing and operating real estate. And then how to structure and organize it.

Who will use RealAI, and what do you expect will be their results over the next year and beyond?

No. 1, it’s built for large-scale multifamily owner-operators. That’s the primary. Obviously, though, if you invest in multifamily and partner with an owner-operator, it’s useful to you. The person who will most value it is the owner of five to 50,000 units of multifamily. The cost is very affordable to start. It’s only $30 to $100 a month to try for an individual. You get basically 50 analyses a month for 30 bucks. And, then, for the enterprise customer, the cost and the functionality increases. For a large real estate company, there’s an enterprise product that has more functionality and lets you do way more. 

This is the kind of data that real estate people haven’t really ever had, because not only is it all this real estate data, we have people’s data, including income and net worth. It’s never been done before. The Internet is a giant advertising machine, and so the advertisers on Facebook, Instagram, Google, J.P. Morgan Chase know a lot about each person when they serve them an ad. There’s tons of data about every person in America. That people data has never been available to real estate professionals.

With RealAI, you get to be an investment alpha, because you have more and better data, and more insights than ever before. And you have automation of the work — building pro forma, doing the analysis, building investment memos, etc. So, for the analyst, instead of taking eight hours to do an analysis and an investment memo, it may only take one hour, and they’ll have more data than they currently have.

The problem, though, is the people, not the technology. A person who doesn’t know how to use it is intimidated, making them skeptical of it. It’s a human adoption problem at the moment. It’s not a technology problem anymore, which is crazy. That’s wild to me. The internet had its claims, and ultimately it delivered, just slower than people thought it would. AI is moving much faster than the internet, but it will also deliver, with the good and bad of that for sure.

Philip Russo can be reached at prusso@commercialobserver.com.