JLL GPT: the good, the bad and the future

There’s a lot we like about JLL’s decision to launch JLL-GPT, a generative AI chatbot able to converse and mimic human intelligence akin to Open AI’s Chat-GPT.

That might surprise you. Start-ups don’t often welcome competition from companies with annual revenues of more than $20 billion, but the move provides some insights into how generative AI is being adopted by the real estate industry and raises questions as what the most effective products powered by generative AI might look like.

The Good

Let’s start with what we like. Investors are still debating whether generative AI will be most successfully deployed via horizontal, cross-industry products like chatGPT, or vertical, specialist products like us, Fifth Dimension AI.

The truth is, the real estate industry produces unique data and workflows that are best catered to by specialist models. Most of the information that will underpin effective generative AI in real estate has never been, and will never be posted online, so models like Open AI’s GPT3 haven’t been trained effectively to use it. That lack of specialist information prevents it from scaling in the way Fifth Dimension AI will (more on that in a minute).

We also like the urgency sparked by the announcement of JLL-GPT. We’ve spoken with competitors and we’re aware of growing fears about being left behind. The recent fall in traffic to Chat-GPT’s website attracted questions as to the true staying power of the latest generation of generative AI but it’s now clear that real estate decision makers view AI as a crucial tool for the future.

The Bad

That said, a fragmented industry served by JLL-GPT, CBRE-GPT, Savills-GPT or the dozens of other potential agent-GPTs ignores what’s possible. Each would only have a slice of industry data and workflows relevant to the controlling firm, which creates a permanent limitation. Biases will inevitably creep into the models. Not seeing the whole market will leave agent-GPTs capped out as OK, but not transformative.

And while this is partly about limiting the potential of generative AI in real estate, it’s also about trust. Can any investor, landowner or competing agent rely on investment decisions and insights created by a model controlled by another market player – potentially with competing interests?

The Future

Businesses remain unsure of the potential of generative AI. Many people I speak to still think of the technology just as improved versions of customer service chatbots or conversational Google search engines.

We need to do a better job of dispelling those misconceptions. The power of generative AI lies in its ability to learn the reasoning behind human decisions. The more you interact with Fifth Dimension AI’s assistant, for example, the better it will get at utilising your investment criteria to provide shortlists of deals that are worth a closer look. It will turn historic investment memos, reports and research into structured data that will enable you to make better decisions, quickly. It will write advisory, valuation or research reports in minutes.

The more businesses use the product, the more it scales in a manner that would be impossible for an in-house GPT that relies on data from a single firm. Specialist, cross-industry models are the way forward.

Companies that opt to take this route will see better results than any in-house GPT, with no delivery risk and at a fraction of the cost. So while we like the launch of JLL GPT, we like our approach more, and we think you will too.