JLL Ups The Ante In CRE’s Artificial Intelligence ‘Arms Race’ With In-House ChatGPT

JLL is the first major commercial real estate brokerage to launch its own large language model artificial intelligence product, announcing Tuesday the firm deployed “JLL GPT” to its brokers around the world.

Similar in name and function to ChatGPT, the large language model, or LLM, that has taken the world by storm, JLL GPT harvests information from the company’s proprietary databases and offers it up in bespoke responses to questions posed by its users.

“This is just the beginning. I’m very excited for all the things we are going to build. I think it’s going to be a game-changer,” JLL Chief Technology Officer Yao Morin told Bisnow Tuesday. “I think this is really the technology that is going to change CRE, how we work.”

JLL will train its model on internal data, supplemented with external resources, to create a tool its 103,000 employees can use to speed up how analysts work, create portfolio dashboards and generate property flyers.

For example, Morin said, rather than a broker tracking down an analyst to ask for rental rate data in a given market, only for the analyst to consult a property database and compile a spreadsheet, that broker can simply log into a secured interface, pose the question to the JLL GPT chatbot and receive an answer in a fraction of the time.

JLL Technologies, the company’s tech division, developed the tool, Morin said, setting a team of in-house data engineers to the task. 

While JLL is the first in CRE to announce such a tool, it won’t be the last, according to Spencer Burton, president of Stablewood Properties and co-founder of the Adventures in CRE blog.

“It’s an arms race, but it’s not just the brokerage firms,” Burton said. “Any real estate firm that’s thinking about the next 10 years is already doing the same thing JLL is doing to some degree.” 

Stablewood Properties itself is working on ways to make LLMs work for its team, Burton said, albeit on a smaller scale than the giants of the CRE industry. However, getting results from these programs isn’t as easy as it seems to casual users, he said. 

Despite the flurry of attention around ChatGPT when it first launched and the speed with which it has proliferated, LLMs are often wrong, Burton said.

“We call them hallucinations,” he said. “So the challenge is, they’ll train a model on their data, and they’ll run into a lot of situations where the model will hallucinate even based on their data and they’re going to have to train through that and it’s just a tall order, an incredibly difficult task.”

Training the model well enough to avoid these factual errors is a matter of throwing resources at the problem, Burton said. 

JLL did not specify its financial investment or its timeline for creating JLL GPT in an interview with Bisnow. 

But the company has demonstrated a willingness to spend money on AI and its applications in the real estate world. In 2021, it acquired startup Skyline AI Inc. At the time, The Wall Street Journal reported JLL planned to use the company’s technology to help clients predict property values and hone in on investment opportunities.

The company’s venture capital arm, Spark, has invested $380M in 45 proptech startups so far, according to its website. Among those investments is EliseAI, a multifamily and single-family property management service that raised $35M earlier this year.

JLL also backed up its enthusiasm for AI in a research report last month, emphasizing its capacity to “transform the sector.”

There is widespread concern that transformation will come in the shape of tech taking over jobs done by humans, but Morin said that is not JLL’s aim.

“It’s absolutely not a replacement. It’s actually quite the opposite,” she said. “After a few years at JLL, I found that there are a lot of ways we can make our researchers’ and our brokers’ lives a lot easier. I see this as an enhancement of our humans.”

Burton echoed those comments, saying he views AI tools as more of a time-saver for CRE professionals than a substitution. Analysts, for example, can focus on actual analysis rather than data entry or sourcing.

“It elevates people. This is not something that, I don’t think, is going to replace jobs, it just to me elevates people to focus on higher and better work,” Burton said.