Generative AI Is Coming Soon to CRE

ChatGPT 4, which debuted in March, represented a sevenfold increase in large language models’ (LLMs) brainpower, which is measured in digital synapses known as parameters, since the original ChatGPT wowed the world in November.

The human brain has an estimated 60 trillion synapses firing. The earliest GPT versions had about 750M parameters; GPT 4 encompassed 4 trillion.

Little surprise then that generative artificial intelligence is remaking several industries, and despite commercial real estate’s reputation for being slow to adopt leading technology, it is headed for our space as well.

Proptech players are rushing to license generative AI technology from chatbot makers, deals that give them access to generic LLMs and the fundamentals on training the bots for domain-specific tasks.

RealPage, the revenue rental management platform, has licensing agreements with Microsoft, Google and other LLM makers to use their Gen AI technology, which the company is training for a variety of use cases that focus on speeding the leasing process.

Evan Davies, RealPage’s chief technology officer, says that the LLM deals are similar to APIs—RealPage can use the technology to make proprietary chatbots. “We can incorporate (LLMs) into our applications stack and surround it with our applications,” he said. “LLMs are becoming a utility.”

RealPage has a tech team training its bots, a process Davies described as “tuning down” a generic LLM and focusing the prompts—the questions you ask the bot—on the precise domain-specific queries that need to be answered.

“What you release into production is a tuned, trained bot solution,” Davies told us.

According to Yao Morin, JLL’s chief technology officer, fine-tuning prompts is the key to optimizing the efficiencies offered by Gen AI.

“It all comes down to the questions you ask. They can have all the knowledge in the world, but right now GPT is trained on general knowledge,” Morin says.

“Just like humans need lawyers and doctors, they need to be trained on specialized expertise for that field. It takes the experts to ask the right questions to Gen AI to get the right answers,” she said.

According to Davies, the tuning-down process—known in the Gen AI industry as “prompt engineering”—is critical to controlling issues associated with early versions of ChatGPT, including inaccuracies and “hallucinations”—basically where the bot makes stuff up.

“You have to set the context of what you’re trying to do—you’ve got to set the tone, the context—and give it a hint of which questions might produce hallucinations,” Davies said.

ACCELERATING WORKFLOWS FROM WEEKS TO SECONDS

RealPage is finishing the “maturing process” for several bots it will put to work as support services to dramatically speed the leasing process. With Gen AI, “dramatic” can mean delivering information that can take weeks for humans to aggregate in a matter of seconds.

“Our objective is to make the leasing process more efficient, to reduce the friction of the process and to centralize its activities,” Davies said.

“[An LLM] can help address turnover issues, it can help address inaccuracies, it can help address repetitive tasks, moving somebody quickly through the leasing process and getting them into a unit,” he added.

Markerr also is focusing on using Gen AI to accelerate workflows with the recent addition of Markets to its Data Studio dashboard.

“We’ve collected a large amount of data-sets over the last few years that really help us with real estate decision-making. Generative AI can summarize this literally in 30 seconds,” Markerr CEO Brian Lichtenberger says.

“[Our clients] can pull the data they want and create an analytic piece as if they were an investment analyst at a major company. Just push a button and create their analysis of acquisitions, investments and retail as well,” he said.

“There are people whose entire job is to query [platforms] like CoStar and put together reports and in-depth analysis, and we’re allowing clients to do that in literally a minute,” Lichtenberger said, adding that the bots have the potential to infuse analysis with “new kinds of insights that haven’t been caught by a human.”

“There’s a huge opportunity to build more accurate forecasts, which is going to be a more accurate set of inputs and modeling which is going to impact across the company whether you’re acquiring or managing properties,” the Markerr CEO told us.

Markerr is generating its 5-year quantitative rent forecast using Gen AI to accelerate the analysis of critical submarkets and key indicators.

EXPANDING USE CASES AS GEN AI GETS SMARTER

In August, JLL unveiled JLL GPT and gave its global workforce of 103K access to the bot, which it says will “transform standard real estate space utilization and portfolio optimization dashboards into dynamic conversations that lead to more actionable decisions.”

JLL’s plan is to supplement JLL’s proprietary in-house data with “external CRE sources” and to begin offering “made-to-order solutions” to clients later this year.

”Connecting buyers, sellers and lenders at the right time, with the right data in hand—within seconds—is going to determine success in this new generative AI era,” said Richard Bloxam, JLL CEO, Capital Markets.

Markerr is planning to offer customer-specific models in H1 2024. “Our phase 2 approach will allow clients to engage in an open-ended way with the data—not just our data, also their own data,” Lichtenberger told us.

“The next step for us is the ability to engage at a more granular level and ask a series of questions that might be unique to that particular client,” he said. “We want our clients to see the next three levels of insight they can derive from Markerr data, as well as their own data, to see those trends in more depth.”

Northspyre, which is tailoring its automated platform to achieve “more predictable outcomes on even the most complex CRE projects,” is developing LLMs that are going to be “much more powerful than a prompt-chatbot role,” according to CEO William Sankey.

“We’re going to surprise people with how we can put the knowledge of the market in their hands,” Sankey told us. “We know thousands of vendors, we know what these vendors charge, we know their contracts. We’ve indexed our data for years.”

“Now we’ve got the ability to mine contracts and change orders. There’s a revolution to be had on how you plan projects, how you procure vendors, how you decide who you’re going to work with and how you make sure agreements and proposals you’re putting together are airtight in terms of scope,” Sankay said.

AN “EVERYTHING TEAMMATE” FOR MULTIFAMILY OPERATIONS

Travtus is deploying advanced analytics and “knowledge mining” using machine learning and generative AI to “radically” change multifamily property management, offering resident-centered design, rapid prototyping and “customer journey mapping.”

According to CEO Tripty Arya, the company has been working with large language models for four years. The result of this work is Adam, “the digital teammate for all multifamily operations”—or, in bot-speak, your “everything teammate.”

Adam is a multi-tasking multifamily bot trained for operational and technical use cases, Arya says. “The first use case is curation of knowledge. The second use case is using that knowledge for better decision-making and benchmarking,” she said.

“We then do a layer of generative AI and statistical analysis for ratings against performance of the community and try to find the needle in the haystack of which of your customers need your attention and why,” Arya said. “Finally, the third use case is automation.”

We met Adam on the bot’s web page (hiadam.ai), where our new everything teammate explained to us how we could quickly become best buddies. Adam said there were three steps we needed to take:

1: Connect Adam to your community inbox, reviews and forms to gather “field intelligence.” 2: Recruit Adam to learn and respond to customer emails, messages and chats with knowledge. 3: Let Adam automate operations and workflows from lead capture, maintenance, to payments “and beyond.”

“The first step was mining conversational data from our clients. We connect to voice as well as chat. We connect to client in-boxes,” Arya explained. “We want unbiased interaction with customers.”

“We mine that to understand what your clients are asking you about. Then we mine your company’s knowledge to build out your knowledge base to understand what it is that your employees know,” she said.

“When you have the knowledge and understand what is needed to be done, you’re able to leverage the productivity of the generative AI model,” Arya added.

ARE BOTS CO-PILOTS OR BODY SNATCHERS?

While Microsoft is promoting LLMs as “co-pilots” that will free up humans for more important tasks, the early emphasis by adopters on productivity may be a prelude to bots replacing large numbers of workers.

“In the first phase, Gen AI will be a co-pilot and [it] will be a huge productivity booster for a lot of the things we are doing today,” Morin told us.

In a white paper JLL released this summer on the implications of generative AI for CRE, Morin said JLL views generative AI as “a valuable human enhancement, not a replacement.”

“The co-pilot model allows people to understand the potential of this without the threat of immense change,” Arya told us.

“It can only be disruptive if it is widely adopted. For adoption to be full scale, you need to minimize change management as much as possible,” she said. “Not rocking the boat is part of the disruption. To be transformational, it has to be something that can be adopted with ease.”

Adam helpfully adds: “A digital teammate is an automated team member that is trained to carry out a business process just like any employee, only faster and without mistakes.”