Raj Singh of JLL Spark Global Ventures: 5 Questions

While Raj Singh, managing partner at JLL Spark Global Ventures, is, of course, a technology enthusiast, he also has the cool analytical detachment of an experienced proptech investor.

As such, Singh has observed and shares the industry’s excitement about robotics, though he remains a skeptic about how fast the innovation will become a common feature of the built world.

Last week, Commercial Observer spoke with Singh about robotics in real estate, China’s industrial use of robots, and why JLL Spark has not yet plunged into robotic company investment.

This interview has been edited for length and clarity.

Commercial Observer: How is robotics impacting real estate right now?

Raj Singh: If we ignore specific use cases around things like distribution centers and warehouses, the answer is almost not at all. Robotics is still relatively early and in a pilot phase, trying to work out what we can do, with the possible sub-exception of cleaning. 

You do see a number of cleaning robots in buildings that have a large amount of floor space — hospitals, airports or something of that nature — where it’s relatively easy for the robots to get around and clean. That’s probably the only use case that has any significant deployment at this time.

Then what does robotics have to offer in the near future? What’s changing?

The context is important here. The macro context is that the real estate industry is not able to find the people that it needs to do the things that we want to do in construction. We know there’s close to half a million jobs in the United States that are not being filled on construction sites. Then, inside buildings, think facilities management. It is a relatively low-paid job. There’s high turnover and absenteeism in those jobs. 

So, we have an issue, which is we don’t have the labor force that we need in place. Robotics is seen as a potential solution for that. That’s why we’re all interested in robotics.

There’s also interest in robotics because potentially using a robot in those environments may be more cost-effective than using a person, a person that we can’t find right now. I don’t know when robots will become mainstream, but the feeling is that they will. The issue is that, up until recently, robots were a one-trick pony: You had a specific robot; it could do a specific thing; and that thing was not a job. A human does a job, and within a job there are many different tasks. Robots tend to be able to do one task, like plaster a wall, or do layout, or clean a floor. There aren’t really any robots that can do a multiple of different tasks that a human might do.

What changed is the advent of physical AI, which essentially is taking robots and putting in some form of an AI brain, so that the robot can do multiple tasks. That now makes it attractive. 

Let’s take the setting of facilities management. I might want a person or a job to include restocking inventory, cleaning the kitchen, or resetting meeting rooms to be ready for the next meeting. Up until now, I wouldn’t have been able to do that with a robot. With physical AI, I could potentially do that. 

But the reason why we’re still at the pilot phase is that for the robots we’re talking about, for them to be enabled, a couple of things need to happen. The equivalent of the large language model (LLM) that the robot would use has to be created. This model is different from an LLM, because LLMs are trained largely on the written word, and the internet is a huge source of the written word. There is no equivalent database of physical actions, the things that robots would actually have to do, and so we need to create that. 

Sometimes these are called world models. There’s various different methodologies like vision, learning action and visual-language-action models that would allow a robot to see how tasks are performed and then be able to riff on that — do that task and also extend that logic to other tasks. All these things are kind of happening now.

The one other thing that still has to happen — and when it happens is not entirely clear — is dexterity. We will then be in a position to use robots for the things that we want them to be able to do. Change is happening very, very quickly. But it’s a fool’s game to say exactly when it will be in place.

We hear a great deal about how advanced robotics is in China and how they’re using it throughout their manufacturing chain. Are their robots still doing just one task at a time, or are they doing something that we’re not catching on to?

Industrial robotics is different from what I think we’re talking about here. It is true that China far outpaces everybody in terms of their deployment of industrial robots. But that’s not quite what we’re talking about here. I think it’s also true that China is advanced in the sense of the types of robots that we would be thinking about deploying for these use cases.

However, to my knowledge, they have not solved the issues that we just discussed around the world model and dexterity. But I would say that the most advanced robotics companies are largely Chinese companies. In part because it’s become a national priority there, and so it’s funded as such, and in part because there are a lot of local Chinese ecosystems that allow companies to move very quickly because all of the relevant technologies and parts are readily available and concentrated in a small area. 

And, so, we don’t quite have that level of sophistication, but we definitely are trying to catch up with that. I think it’s an exaggeration to say that they are out there doing these sorts of tasks. They’re not. But are they leading in terms of the development of solutions? They might well be.

Taking the industrial broad view, is that applicable to things like the construction industry — which is gigantic — or is that sector still unable to have robotic supply match demand?

There’s a disconnect. The industrial robots can’t be taken off a factory floor and put onto a construction site. You need to build robots that are appropriate for the construction site. The factory floor is a relatively predictable place and has relatively few humans. Construction sites tend to be outside in variable weather conditions with not necessarily smooth floors, and there are people.

And, so, for construction, we still are looking for robots that can do multiple things at the same time. A lot of the companies I speak to are trialing robots, but even the Chinese robots that they’re trialing are not quite ready for prime time yet. So there is an opportunity for the U.S. or European robotics companies to deliver against that result without having to face overwhelming competition from China.

Then, if we want to get even more detailed, the business model that robotics companies are thinking about using is potentially also problematic because, from my perspective, I’m looking to capture greater margin by deploying robots. If the robotics companies are looking to capture all the margin for themselves, then me using a robot doesn’t necessarily help

Given all those challenges, are there any robotics companies that JLL Spark has been looking at or in which it has invested?

We have not invested in any robotics companies to date. We are looking at robotics companies, but I’m not sure that we are expert enough in robotics to choose a winner. I’m interested to see when the ecosystem is ready so that we can use those companies. 

My interest in terms of investing is investing in the software and services that go around the robots, because another reason robotics isn’t ready yet is that the tech stack inside a building doesn’t know what a robot is. So when your building management system spits out a work order, a human can easily take that work order, understand what needs to be done and do it. A robot does not understand what that work order does. 

We need some way of translating the tech stack and its outputs that we have today such that a robot or multiple robots of different types can actually go and solve the issue at hand.

We also need software to be able to coordinate or orchestrate the use of multiple different types of robots in the context of people also being there doing jobs. I have no illusion that we’re going to be replacing a lot of the workforce anytime soon using robots. It will be a mixed environment where there are people doing some things, robots doing some things, and situations where people and robots need to work together. Making that happen is where we need to work quite intensely to get a solution.

Philip Russo can be reached at prusso@commercialobserver.com.