Is Artificial Intelligence’s Impact on Proptech for Real?

Everyone in proptech wants to talk about artificial intelligence. But does anyone really know what they’re talking about?

A harsh question perhaps, but the variety of answers from proptech entrepreneurs and investors suggests that any conclusion about the reality of AI in the sector now and in the future is indefinite at best.

“AI is broader than most people think,” Blake Owens, founder and CEO of Agrippa, an AI-powered platform that connects commercial real estate investors with those seeking capital, said in an email. “Two major subsets are machine learning (ML) and natural language processing (NLP), which have further subcategories, such as large language models (LLM) within NLP.”

While AI is not new, it has become more popular recently due to OpenAI’s ChatGPT, itself an LLM, Owens said. “Many new products are just ‘AI wrappers.’ These are simply front-end user interfaces (UI) that rely on another company’s LLM,” Owens said. “The problem is that they often lack their own means of data collection and interpretation, making them easy to replicate or replace. To be differentiated and defensible, you need a lot of proprietary, high-quality data.”

Agrippa is one of the many proptech companies that claims to use AI effectively in its business.

“We’re applying AI primarily through ML, and we plan to introduce an LLM-based feature in the future,” said Owens. “In our ML use case, we collect granular preference data from users via submissions and platform activity observations. This data is then utilized to better understand whom to connect with and why.” Yet, because Agrippa is selective in whom it places on its platform, Owens admits that “a lot of time-intensive manual effort” is still required.

Ultimately, the impact path of AI in commercial real estate will be exponential, Owens maintained, but it will take time in the current early stages due to the amount of low-quality data that must be filtered out. “It will accelerate. Great things take time.”

Of course, there are numerous proptech companies — too many to list here — that claim AI as integral to their business.

“The first use of AI that we really had in our platform was taking all the operational data that we have — whether that’s work order information, lease application information, last year’s budget information — and using that to underwrite customers for better rates on mortgages, cheaper insurance policies, etc.,” said Grant Drzyzga, founder and CEO of Revela, a Detroit-based property management software platform for lending, insurance and banking.

“AI gets into the optimization component of our business,” Drzyzga added. “We have gone down the road with AI in 2024 with chatbots for leasing and using AI to query a database or build a report, but that’s really our exposure to it right now. In 2025, we have some big areas and plans for improvement and expansion.”

One example of Revela’s AI use is in the area of delinquent rent collection, said Drzyzga. Instead of a property manager having to hire office clerks or on-site property managers for such work, their AI product does the entire process automatically.

“This product allows you to generate a state-specific demand for nonpayment of rent letter,” he explained. “Every state has different criteria, so we run that through the natural language model to make sure we’re adhering to each state’s specific laws. The AI will actually call, text, email, and send additional threatening letters in the mail. In that regard, we’ve helped customers avoid having to staff up as they’ve grown over time.

“One instance of that was a student housing operator who was going to hire two new accounts receivables clerks,” Drzyzga said. “They didn’t have to do that, which saved them 140 grand and then any other additional overhead on top of that, while it took its collections percentage from about 80 percent to 96.5 percent.”

Amanda Orson is the founder and CEO of Galleon, which she describes as “a platform for consumer-driven real estate transactions.

“We see a future where residential real estate does depart from its 111-year-old status quo of constantly needing to have an agent on both sides of the transaction,” Orson said. “We think that there should be at least one alternative, and we’re building tools to enable that alternative.”

Founded in August 2023, Manhattan-based Galleon’s peer-to-peer residential real estate platform claims AI is a major part of its design.

“It’s baked into every part,” said Orson. “Our open and direct marketplace takes two minutes to list right now. It’s entirely off-market, non-MLS inventory. We’re developing software called Navigator that will allow you to co-pilot a transaction with software in the same way that TurboTax does for tax filing. It’s entirely self-driven. We’re simply helping buyers at the moment, and soon we’re going to follow up with sellers, probably next month, [help them] navigate their transaction from start to finish.”

AI is already accelerating Galleon’s engineering and development operations, Orson said. That includes tools such as Cursor, an AI-powered code editor, and Replit, a development and deployment platform, for tasks that the company would otherwise hire people to perform. 

“Call it a junior software engineer,” she said. “The work can now be done in a very automated fashion. It’s not visible to the consumer eye, but it is at the very foundation of how we’re actually building. That gives us the ability to hire fewer people and to be able to stay leaner and more efficient than people or companies that are currently hiring a lot of people or who have a lot of people on staff.”

Calling AI an “absolutely common” prerequisite in the real estate industry, Orson sees the actual efficacy of the tech, but also the hype.

“Especially for a small organization, it’s one of the only ways in which you can have an accelerant or a leverageable tool to work with incumbents that have a 6,000-person operational expense. It’s not a way for you to replicate what 6,000 people could do, but it does give you much more bang for your invested dollar in terms of product output.

“But right now, there’s still a lot of people that are kind of forcing AI down consumers’ throats, making AI things that no one actually asked for, or that don’t actually have any utility.” That means a learning curve ahead, Orson added.

A challenge for the use and growth of AI in real estate is the often siloed and closed technology hardware the industry uses, said Edi Demaj, co-founder of Detroit-based Kode Labs,  a smart building operating system.

“We have a multitude of AI applications within our platform today,” said Demaj, who estimates that some 25 percent of his firm’s software uses AI. “Everything from energy forecasting and predictions using AI models to generative AI that actually helps you operate more efficiently. We’re positioned perfectly to be able to take advantage of these different gen AI algorithms that are being pushed, as well as to work on new ones ourselves.”

In that, Kode might be ahead of its time.

“Frankly, the built world is just not ready for it because of the foundational issues that exist within the space’” Demaj said. “The big four or five hardware manufacturers have basically taken all these territories around the world. Locally, it’s like one or two of them own the territory. So they have no incentive or desire to build hardware that has open APIs, which allows for the ability to take advantage of new applications that are coming out now specifically related to AI.

“The problem is that if your hardware isn’t open and is proprietary, and I have 10 buildings, with Johnson Controls there, and Siemens there, etc. I can’t take advantage of any one of the new support applications that are coming out, because they don’t even speak to each other.”

From the investor use point of view, AI still needs to prove itself, said Sarah Liu, partner on the real estate investing team at proptech venture capital firm Fifth Wall.

“AI had impacted investing even prior to the current wave of hype,” said Liu. “There’s been tools like Harmonic, for example, which fits well and picks up on certain signals, like a founder’s background — have they founded a previous company, spikes in employee count from business, and things like that. All those existed long before the generative AI pipe really emerged.

“Right now, what we’re seeing more on the investing side is probably less value from AI’s ability to source incrementally from what we were already using, but more value in terms of being able to just summarize information quickly. Like, basically, helping with some of the drafting of a memo and being able to write summaries of our findings from a data room more quickly, and things like that. That tends to be where we’re seeing the use cases much more, compared to the core functions of actually being able to talk to founders and their teams, ask some questions, or being able to think more critically about the business and develop a thesis.”

While investors and startups are not yet demanding AI for their work, it is becoming increasingly expected as an everyday tool, Liu said. Still, its impact is nowhere near what some fear, she said. 

“I think it’s still relatively limited, to be honest,” she said. “I think it’s expected that everyone stays on top of what’s going on. But it’s not like AI will suddenly replace people or anything close to that.”

It’s more right now along the lines of AI saving companies 30 minutes’ work here and there, Liu said. 

“That’s the level of adoption we’re seeing,” she said. “Much more so than, ‘We have an AI associate.’ I know there’s been a couple of those companies that have raised capital trying to be AI associates for equity, for example, but that’s not something we’ve really seen. Given that investing is so numbers driven, we need to be able to do a lot of math in our day-to-day. And AI has that as its Achilles’ heel at this point — it’s not great at math.”

Josh Panknin, director of real estate AI research and innovation at Columbia University’s engineering school, spent 15 years in the industry before entering academia. He’s blunt about AI’s current role in real estate.

“It’s a mess of a space to be, in my opinion,” said Panknin. “A lot of people out there are trying to do things and don’t really know what they’re doing.”

Real estate’s heterogeneous complexity makes it extremely difficult to build AI that can be used efficiently and replicated at increasingly lower cost across the industry, he explained. Plus, traditional real estate companies have little tolerance for investing in new technology that doesn’t produce quick profits.

“We spent a lot of time trying to develop a structure for what most people talk about today as AI, which is actually machine or deep learning,” Panknin said of his academic work at Columbia. “And one of the ways we separate things is between operational and strategic technology. Operational is more about efficiency, day-to-day automation. How do I just make things a little bit faster and remove the manual component of that?”

Strategic technology, on the other hand, is more about using tools for a competitive advantage, Panknin said. “But what I don’t see is companies using it for more strategic things like, ‘How do I automate the analysis of markets so I can see things such as which markets are going to have higher growth rates relative to other markets over the next five years?’

“That’s really where I see the gap: Automation is relatively easy compared to new analytics, largely because when you’re automating things, the data is already there. Whereas when you’re doing something new, typically you need new, different data. That data has to be worked with a little bit differently, and that’s a harder thing to do. It’s more complex and much more expensive to pursue those kinds of things than basic automation.”

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