Category: Press

JLL Spark VC talks about the challenges PropTech startups face: ‘Ride through this rough period’

Raj Singh, managing partner of JLL Spark, is one of the 100 leading corporate venturing professionals in our Powerlist this year

In recent weeks it has been difficult to log on to any news service without coming across at least a few headlines relating to artificial intelligence (AI) generally and ChatGPT specifically. The rise of the latter, currently the most prominent example of generative AI, seems to have finally brought AI into mainstream conversation.

It’s a divisive subject. For some, the benefits are practically limitless: AI doctors, for instance, have the potential to make medical diagnostics far more accurate, consistent and quick. 

For others, however, the the dangers are terrifying. There is, after all, a cogent argument that AI bots could ultimately decide that “hairless apes”, to quote the broadcaster Kirsty Wark last week, have served their purpose and that mankind has outlived its usefulness.

Those are the extremes. In real estate, generative AI has a huge range of potential uses. React News has already taken a tongue-in-cheek look at ChatGPT’s potential to come up with names for property portfolios, but there is much more that AI can do, from putting together marketing brochures to modeling development opportunities.

JLL pilots

Some companies are already getting started. One of the more advanced property companies in terms of thinking about AI is JLL, which has both an internal technology division, JLL Technologies, and a venture capital arm, known as JLL Spark. According to Raj Singh, managing partner of JLL Spark, the former already has a number of pilot programmes up and running. 

“Those are around applying generative AI to some parts of our business,” he says. “So, an example would be leasing. We’re getting a generative AI large language model to ingest leases and then allow us to essentially interrogate that in a way that you wouldn’t be able to do otherwise.”

Singh admits that it is early days, but says that even nascent projects are starting to yield results. “I actually counted five or six different projects that are going on right now,” he says. “And what’s interesting is that even those of the projects have only been going a couple of months are already showing results because of the speed and power [of generative AI].”

Then there is the work JLL Sparks is undertaking. That involves investing in multiple technology start-ups, many of which are already using generative AI. A good example is Elise AI, which is effectively a chatbot that manages prospective tenants and residents, answering emails, texts, chat message and phone calls within minutes. The point is to automate leasing, service requests and so on. 

“In my previous life, we spent years trying to find chatbots that would take the load off calls into agents and never found one that really worked well,” says Singh. “Now, generative AI does it and it works well. [Elise AI] allows you to have a property that you’re trying to lease out and then use the chatbot to arrange meetings and answer questions, but at a level of sophistication that we’ve never had before.”

Another application JLL is currently working on regards different languages and jurisdictions, which at present requires multiple local professionals with the relevant linguistic and legal skills. Generative AI can already eliminate a lot of that work. “When you’re dealing with international clients, you’ve got lots of different languages and lots of different laws,” says Singh. 

“One of the pilots that we have going right now is ingesting all the different laws and regulations in all the different languages and that then allows you to query all of that using English. So, you don’t have to learn Mandarin or French or Spanish. You can query in English and you can go and find the relevant information in all those different languages. That’s extremely powerful.”

Planning and development uses

AI is also already being used in the field of planning, according to Frances Wheat, a consultant at planning consultancy HGH Consulting, who adds that there is the potential for generative AI to go much further. “The huge amount of data across the built environment needs to be unlocked, with software made available without significant cost or technical barriers,” she says. 

“We can already see the efficiencies emerging across the planning processes: GIS layering of constraints and opportunities; site selection for local plans; engagement in planning; visualisation of proposed development; automating validation and assessment of straightforward applications; and chatbots to answer questions. We can imagine many more benefits.”

The same is true in construction. Ashley Pappin, senior associate at law firm Winckworth Sherwood, says that the results are already being seen in terms of health and safety. “One of the areas of early adoption of AI is health and safety on construction sites,” he says. 

“Visualisation of data reflecting project activity and safety performance has steadily become normalised in development projects, resulting in better practices and the ability to pinpoint risks. The next step is to develop how AI can reliably interpret construction site data to identify and predict risks and actively propose areas to be urgently acted upon.” 

Future gains

As for the future, Singh believes that AI will ultimately prove transformational. “This is going to be a game changer in our world, as it will be in most other industries, and the reason is that the world of real estate runs on data,” he says. 

“The problem is that that data is fragmented and unstructured and so all of our efforts until recently have been about how do we ingest that data and structure it in some way so that we can actually use it for another purpose. Generative AI essentially eliminates that requirement.” 

That means no more tedious file conversion or scraping PDFs for the relevant data, which until recently was a time-consuming and therefore potentially expensive process. “You can just throw the stuff — the documents, the text, the video, the images, whatever it is — into a large language model, actually you’d have to have more than one, and that will allow you to avoid the entire manual and laborious process of structuring,” says Singh. 

Pappin adds that generative AI also has the potential for greater application when it comes to project planning and construction contracts. “There is a clear opportunity for the use of AI to mitigate the risk of disputes that can arise, for instance in relation to payments and project scheduling,” he says. 

“To an extent, software is already used to analyse the critical path of a development and has been for some time. A logical next step is further development of machine learning platforms that actively and reliably spot risks and opportunities, and identify ways to harness and reduce their impact.”

Potential barriers

However, before we get too excited, there are hurdles to be overcome, some of which relate to legal issues and others to matters that are still properly the reserve of human judgement. The first relates to the ownership and reliability of the data that is made available to AI applications. 

“It’s fair to say that you should not use the open public systems because they are using data from somewhere and you don’t know what their data is and you don’t know who generated it and who owns it,” says Singh. “That’s why in the case of us creating something using AI, we would have to know everything was owned and sourced by us.”

That doesn’t mean that every bit of code has to be independently generated, just that the data that is fed into a system is proprietorial. “There are lots of open source, large language models, so we can get one of those and we can put in our data and can train it on our data alone,” says Singh. “Any results that come out of it are based on our data, our proprietary information, and so the responsibility lies with us, as it would have done previously.”

Wheat makes a similar point, albeit more hesitantly. “Will AI be able to discern reliable data and misinformation?” she asks. “Will it be aware of errors and risks and tell us? Unlike government, AI is not subject to data regulation. Indeed the scope to manipulate AI visualisations of proposals and consultation responses already exists and is becoming increasingly sophisticated. Can we trust AI?”

So, it is evident that AI — and to a lesser extent, generative AI — is already being used across real estate. Moreover, it is also clear that leading proponents have a lot of interesting ideas about how it can be used further in the future. Constraints remain, however, and although inconvenient require careful consideration.

June 6, 2023 | Adam Branson, React News

Source: React News

Though overall investment in the space is dipping, the desire for data analytics and efficient use of space grows more intense.

Data from buildings could eventually yield more revenue than the buildings themselves, according to JLL Spark Global Ventures’ Raj Singh