The Firms Using AI to Boost Predictive Maintenance in Real Estate

These days, summer in the city — any city, or pretty much anywhere — often means temperatures in the 90s with a heat index over 100. Not a time when you want an unexpected air conditioner failure.

To prevent such costly failures, real estate owners and operators are increasingly turning to proptech companies that use artificial intelligence for predictive maintenance.

That’s especially true in preventing HVAC fritzes before they happen.

Having grown out of the Brooklyn Navy Yard-based venture capital platform Newlab, climate tech startup Thalo Labs now calls Manhattan home. Residing in one of the hottest of hot spots for urban carbon emissions, Thalo Labs aims to bring high-scale, less expensive HVAC technologies to a broader market, focusing on the 85 percent of buildings without advanced management controls, said Brendan Hermalyn, the company’s CEO.

“We like to focus on all this unmonitored, un-sensored equipment that makes up the vast majority of stuff on the planet,” Hermalyn said. “We look at all of the different constituent gases — carbon monoxide, oxygen. We can see methane detection in real time, natural gas leaking. Sixty-five percent of boilers are leaking natural gas. From that data set, we can predict, ‘Hey, is this boiler on the way out or not?’”

Thalo Labs’ hardware and software systems monitor and diagnose issues in real time, using machine learning models based on equipment physics. This can help technicians proactively address problems and reduce reactionary maintenance.

“New York City is one of the best places for developing this,” said Hermalyn. “We think of AI as a San Francisco Bay Area thing. I lived out there for a decade. I moved back here to start this company in New York. This is where we should be bringing proptech and AI to the real world. And that’s what we’re doing.”

The Bronx native holds a Ph.D. from Brown University in planetary sciences as well as a master’s in engineering, mathematics and geological sciences. He is focusing his expertise on leveraging proprietary data for more effective analysis to benefit HVAC service contractors and internal HVAC groups.

Thalo Labs is attempting to take expensive high-end technology for mission-critical facilities like aerospace and defense power plants and bring them down in scale and cost. That way, the technology can be deployed across the built environment, said Hermalyn.

“It’s all fantastic because it does all this preventative maintenance,” Hermalyn said of the high-end technology. “You’re constantly looking at stuff. You constantly have people looking at things. But you would never install a $5,000 heat pump and put a $5 million spectrometer on it to look for leaks, right? It just doesn’t make any sense to do that.”

One result of Thalo Labs’ predictive technology is that during the latest heat wave in New York City, it had a zero failure rate for hundreds of HVAC units it was monitoring, said Hermalyn.

HVAC predictive maintenance isn’t the only use for AI in real estate, though.

T2D2 is a Manhattan-based proptech company that uses AI software for predictive maintenance, including as part of a machine learning technology called Computer Vision, to monitor building exteriors.

“T2D2 is used by property owners, managers, and their engineers and architects to analyze building exterior data for the purpose of finding small issues before they become large, essentially targeting items for preventative maintenance,” said CEO Jonathan Ehrlich.

The company separates the data capture component from data analysis, allowing the former to be done on anything from a cellphone camera to a drone, Ehrlich said. 

“Then, when you upload it to our platform, we automatically process it through our proprietary computer vision,” he said. “And, using AI, we compile reports that provide you with information regarding where these conditions are. In addition, we have our interactive portal.”

Demand for T2D2 is highest in commercial office buildings, including the Empire State Building, and a number of Major League Baseball stadiums, said Ehrlich. He declined to say if CitiField and Yankee Stadium in New York were among them.

Despite the demand for AI predictive maintenance tools, Ehrlich doesn’t see the human aspect going away soon.

“There is certainly an importance to the human touch and feel when it comes to exterior building or structural inspections that is going to play a critical role for many years to come,” he said. “I have seen experimental technologies that look at robotic solutions for resolving these types of issues, but there are a lot of safety concerns around that.”

Getting back to HVAC issues, Visitt, an AI-driven integrated property operation platform,  focuses on predictive maintenance in large commercial buildings by analyzing data from human activities and building management systems (BMS), said Jonathan Kroll, the company’s co-founder and chief product officer.

“The major piece in our platform is everything on the back end, which mainly involves maintenance activities,” said Kroll. “So all the assets in the buildings, whether it’s the critical machinery, HVAC system, elevators — all this important system data is being stored in Visitt. And all their preventive maintenance schedules are being hosted on our platform, and all the data is being documented.”

Visitt began incorporating predictive maintenance onto its platform in 2022 to identify recurring issues, Kroll said. 

“The idea was to use large language models (LLM) to be proactive in using signals to basically escalate certain activities or certain events,” he said. “For example, if there is a certain work order that is being created again and again, it might indicate a reoccurring or greater issue. So we are escalating it using AI.”

The Manhattan-based company aims to bridge the gap between humans and data in sensing and dealing with maintenance issues.

“There is a lot of noise, and what AI does really well is to triage this noise in detection,” said Kroll. “So we use this first triage, using the AI to help us reduce more than 90 percent of the alerts or events being collected, mainly from the building management system. Then we use pattern recognition and past data that was gathered to understand what’s the next step, whether it’s escalating to a human that will proactively go to inspect something, or it’s something that should not get through to a person, such as an engineer. Or, it can just go to the reporting side to be logged, but not something that requires immediate attention.”

David Blumenfeld, co-founder of NextRivet, a San Francisco-based proptech consulting firm that helps real estate clients implement various maintenance technologies, highlighted AI’s importance and particular effectiveness in energy optimization and predictive maintenance. Both, he says, can produce significant cost savings.

“CFOs especially are looking at what they should be spending money on,” said Blumenfeld. “When you talk about being able to save money, either it’s energy optimization — whether that’s HVAC or lighting or other things — or whether that’s the predictive kind of cost savings, essentially to drive a tighter bottom line, it’s something that vendors who are installing these things and coming up with these things, to the real estate companies implementing them, it’s a real thing and real savings.”

Companies like BrainBox AI, Vertigris Technologies and Conservation Labs are leveraging AI to monitor HVAC systems, optimize energy usage, and detect water leaks early, said Blumenfeld, noting that demand for AI in proptech is high across sectors, with multifamily and office leading the way. He also maintained that venture capital interest in AI proptech is robust, despite the potential pitfalls of the rapid evolution and potential obsolescence of technologies.

AI is also providing greater efficiencies and cost reductions in work orders, a ubiquitous real estate document challenge. 

Property service company Lessen has invested more than $200 million in AI tools. That has helped reduce human intervention to 25 percent in residential work orders, while handling over 2 million residential and 1.5 million commercial work orders last year, said Sean Miller, chief revenue officer at the Scottsdale, Ariz.-based company.

“The predictive side is starting to emerge more now,” said Miller. “We have clients that have deployed all these different Internet of Things devices. They can see things like HVAC data. They can see water leak data. But what do they do about it? We’re starting to take this and compare it to the actual work we’ve done, which helps us identify when we see this signal that it could mean something.”

AI for predictive maintenance is a truly national trend at this point, with international inroads too. 

Overland Park, Kan.-based Lula, is an AI-driven proptech platform that helps property managers and tenants prioritize and dispatch a nationwide network of professionals to respond to maintenance tasks. It focuses on single-family rentals, and some 95 percent of its work is with middle-tier property management companies, which the company defines as having 300 doors or more.

“We do work with some institutional operators that have tens of thousands of doors, but we don’t go downmarket to DIY landlords,” said Bo Lais, co-founder and CEO at Lula.

“I think AI in general is moving extremely fast,” Lais said of the industry. “I think it’s moving a little too fast, in my opinion, in some cases. We’re getting into voice AI now ourselves, but we have a product that basically is a web app that’s conversational, so it feels like you’re just chatting with a real human. … Let’s give you an example that you know your garbage disposal is not working right. We have no problem asking the resident to reset it. We’re not going to ask them to stick something down in it to try to unjam it, or maybe someone puts their hand down it. If you use open AI models, they might ask you to do that. We have our algorithms to make sure we’re not asking residents to do something that might put them in harm’s way.”

Two U.K.-originated proptech startups using AI for predictive maintenance are Demand Logic and Verv Energy, the latter of which has just entered the U.S. market, working out of Los Angeles.

London-based Demand Logic collects data from various systems like building automation and energy management networks, contextualizes it, and provides automated or advisory services to customers. This customer base varies from large blue-chip companies to smaller firms.

“The technological challenge is about the volume of data that we’re seeing in a building,” said Mike Darby, CEO and co-founder at Demand Logic. “If you take a Canary Wharf building in London, there’s a million square feet. You look in a building like that and we will see maybe 100,000 to 200,000 data sources from a single connection. We have a hardware device that can understand the languages of most building automation systems. One device has sufficient capacity to stream data from that kind of volume we’re talking about.”

Demand Logic, which is expanding globally and operating in cities like San Francisco, claims it can handle around 150 billion data points, using AI to reduce the time-consuming process of contextualizing data from hours to minutes. Its primary customers are asset managers, but it is also engaging with facilities management companies. The company is exploring AI for real-time human-machine interaction to enhance customer engagement.

Similarly, London-headquartered Verv uses AI for predictive maintenance in commercial air-conditioning systems. The company interconnects equipment to the Amazon Web Services cloud for data analysis using machine learning to detect issues like coolant leaks and filter blockage.

“We have a lot more machine learning models, probably, than any of our competitors,” said Peter Davies, founder and CEO at Verv. “We’re pretty accurate as to how that operates, and it’s probably the cheapest solution out on the market.”

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