Raj Singh, head of JLL's investment arm, sees opportunities in the AI data centre boom and the prospect of rethinking a document-heavy sector.
Real estate — just like most sectors — is being upended by AI, says Raj Singh, managing partner of real estate services firm JLL’s strategic investment arm, JLL Spark Global Ventures.
“We do seem to have been overtaken by the AI trend,” he says. “The real estate industry is a very good candidate for AI because we have so many documents. Generative AI is all about ingesting documents, giving you access and drawing insight from thousands and millions of pages of documents that would take a human a long time to work out.”
Singh says the unit is especially interested in using generative AI to manage customer relationships. If, for instance, JLL is looking to lease office space in a building, it can use a startup’s AI technology to build a detailed model of every existing or potential customer, giving its accounts team the edge when it comes to contacting the right customers and having the information to close the sale. JLL Spark led a $3.5m seed round for a company, Ren Systems, with just such a product last year.
The data centre opportunity
At the same time, there is a big emerging opportunity in data centres. There’s a rush to get data centres built to deal with the gargantuan computing needs of artificial intelligence, but those structures come with a range of logistical challenges that need to be worked out. And that is fertile ground for the kinds of startups JLL Spark is targeting.
“For JLL, it’s a very interesting intersection [within] the world of real estate,” Singh says. “You’ve got to find the site, get permission, get the permits, build the connection to the electrical grid and get water, get the contractors to actually build this thing. And then once you’ve built it, you’ve got to kit it out with air conditioning, water-cooling racks for your GPUs and so on – this is what JLL does.
“So, we’re looking for software tools that can help that process – the construction process and the management of data centres – run more smoothly. Another area we’re looking at is companies that will have that little edge that will allow the data centres we help create for our clients to be a bit better than others.”
An industry in need of an update
Real estate services is not an area where there are many corporate VCs – the only other sizeable unit in the industry is linked to shopping centre operator Simon Property and is more retail tech-focused – and Singh believes JLL is one of the few businesses in the industry with a sense that technology will change the bedrock of the industry, as opposed to just consisting of tools that could help things along.
“I think we’re still missing AI-powered vertical applications to deliver the right level of information,” he says. “The connection between all the people in the field who are doing something and the central organisation is something that we haven’t done as well as we ought to.”
Other areas that need work involve improving the efficiency of construction projects, which are still prone to run over time and budget, while the process of staffing and the cost of running buildings remains unpredictable. A significant aim has to be making those more predictable while increasing profit margins.
“Honestly, for every subsection of what we do there are lots and lots of big issues, because the real estate industry until now hasn’t been a strong adopter of technology. So, there are opportunities everywhere you look,” Singh says.
“I look at every other industry out there and they’re all being radically reshaped by technology, so I see no reason why our industry will be any different.”
Beyond AI to other software/hardware technologies
For all his enthusiasm about AI, Singh takes a fairly sober approach on AI investments in practice. The goal is to back startups with models designed for use in real estate – small language models, as opposed to the large language models used by the likes of ChatGPT. You need a specialised product with the right guardrails, otherwise all you’re doing is spending extra money to make what you’re doing harder and longer.
“When we look at startups, we often bring them in for a trial,” Singh says. “And what we find is that our customers may like the tool, but because the tool doesn’t understand what they’re trying to do, it takes them longer to do it, even though it’s AI.
“That’s why I’m saying software will definitely be there and AI is going to be a big part of it, but I think it’s going to evolve from these generalised platforms to specific, verticalised applications.”
Beyond generative AI and document management, there are plenty of interesting new technologies coming into real estate, Singh says, and JLL is already actively using tech products developed by JLL Spark portfolio companies in its everyday work.
Those products include Qbic, which can scan empty real estate space, then create a 3D model that can quickly incorporate any changes into that model. JLL takes clients around office spaces and uses the software on an iPad to show them what it could look like were they to move in.
OpenSpace has created a helmet webcam-based system that can analyse building sites to identify inconsistencies with the construction plan and flag them up immediately, and several JLL customers use it to keep track of what is happening with their buildings as they’re going up.
But while hardware will always be part of the what JLL looks at, Singh says investment patterns are moving more from hardware towards software. Part of that is down to advances in AI and part is due to plain economics.
“As people build out new buildings, they’re going to put in a lot of the tech before a person or a company has moved in. But the problem is that it takes a long time to build buildings, and so you’re often outdated by the time you get to market,” he says.
“The focus is now on whether I use what I’ve already got. Because the reality in our world is that it’s expensive to put stuff up. You’ve got to find people to do it, you have to pay the capital expenditures, buy the hardware to suit the client. But I wouldn’t want to ever say that we’re not going to need it, because you almost always will.”
