Subscribe
Share
Share
Embed
Current Season: First Draft Live
Between economic whiplash, shifting policies and market volatility that changes by the hour, you need industry insights that cut through the noise. That's exactly why we're launching First Draft Live, a new weekly series that breaks down what's happening, why it matters and what you need to know to do better business.
Join us live on Bisnow.com every Friday at 12:30 PM ET / 9:30AM PT for conversations with the industry's sharpest minds discussing the week's most critical stories, or catch the replay right afterwards — here on your podcast app of choice.
Okay. Welcome to First Draft Live. It's Friday, September 25. I'm Mark Bonner, Bisnow's editor in chief coming to you live from New York. Thank you to everyone who's tuning in live today from wherever you are.
Mark Bonner:We really appreciate it. So let's jump in the show. By now, most big companies in America say they're using AI. In the last few weeks, Apollo's chief economist, Torsten Sloc, went through q two earnings and found 69% of S and P 500 companies now point to a live AI deployment. But ask them to show what AI has actually delivered, and the number falls off a cliff.
Mark Bonner:Only 29% could put a number on a result. Just 2% track a metric over time. And not one of them breaks out AI as its own line on the p and l. And when companies do put a number on it, 70% of the time it's about cost cutting. Commercial real estate looks a lot like that.
Mark Bonner:JLL found that the shares of corporate real estate firms running AI pilots went from 5% to 92% in three years, but only 5% of the people it surveyed say they've hit most of the program's goals. Meanwhile, the bills keep climbing. CBRE spent nearly $1,700,000,000 on computer hardware and software last year, up 300,000,000 from the year before. And AI isn't running cheap to run anymore. Meta, Uber, and Walmart have all told employees to rein in their token spending.
Mark Bonner:So the question for this industry may not be whether AI works, it may be who builds it, who answers for it when it doesn't hold up, and who can get the industry to actually use it. To help unpack all of that, I'm joined today by Reeves Davis, president of Technology Solutions at JLL. And to our audience, send your questions into the chat. We'll get to as many as we can. Reeves, welcome to First Draft Live.
Reeves Davis:Hey. Thanks, Mark. Glad to be here.
Mark Bonner:So, Reeves, let's start with what's changed on the client side. Building your own tools used to take an engineering team and a budget. Now it takes an afternoon. In our June reporting, a senior Cushman and Wefull broker told us people across firms were using Claude on their own before the firm had even sanctioned it. At Walker and Dunlop, executives put two analysts head to head on a prospecting task.
Mark Bonner:One did it manually and it took about a week. The other used Claude and finished it in ten minutes. You've made the point to me that this cuts both ways. It's cheap and easy for clients to build their own tools now, but it's just as cheap for JLL. Why does that shift favor JLL rather than the client?
Reeves Davis:Yeah. I think it's been really interesting watching the, like, the flow and the change that people's expectations in this space, right? I think it's it's it's frankly mirrors plenty of cycles and other trends that we see in certainly in technology but I'm sure in in culture as well where there's a super high expectations then you're met with a slight decrease in when reality hits and you have to determine, all right, what value is actually getting created by this. And the reality is that generating software is not the same thing as maintaining it for years. And ultimately what I find as there's the excitement over the technology, there's an amazement of what people are capable of accomplishing.
Reeves Davis:And then as expectations sort of mature a little bit, you arrive back at, well, what I really care about is getting an outcome. I'm not so worried about like, is this happening? How this is happening? I actually want the outcome. But my expectations now are that I'm going to be able to get that outcome faster, cheaper, more consistently.
Reeves Davis:And the way we look at that is in order to leverage that you need data. You need data at scale. You need to be able to see across the entire industry of what's going on. Otherwise, every organization would effectively be trying to roll their own intelligence into a platform. And when all they see is their own portfolio, well, it's a far more limited scope.
Reeves Davis:So I think data really is the key from our perspective to being able to unlock those outcomes.
Mark Bonner:And that's the edge that JLL brings.
Reeves Davis:I think so. You know, you when we look at this, you know, everyone goes through a phase when you look at the the technology like this of I think of it as experimentation. Right? It's what you can do that is novel, what you can do that is more efficient, how people who are experts in real estate outcomes can all of a sudden implement technology ideas very, very quickly. But and that, you know, when when experimentation goes to a process and let's say not all of them succeed, right?
Reeves Davis:That's the stats you shared at the beginning, You know, I think that that's something that we shouldn't look at as just a failure, right? Sure, but not everything succeeds by by any means, right? Things you know, you you look at stuff later and you realize, oh, well, you know, we didn't have the right data for that. We we thought that this process was consistently applied, but maybe it wasn't. Right?
Reeves Davis:Or maybe our people weren't ready to adopt that change. So I think it's a matter of looking at effort, setting goals, tracking your success, understanding the ROI, but accepting the fact that a normal cycle of experimentation and then productization and scale does does have some waste. There is some expectation of failure. But if you learn from that failure, then have you really lost a 100% of the investment in that? I I would argue you haven't.
Reeves Davis:And I think that's a large part of the the the process that we're seeing right now. I think it's honestly so acute to people because the the cycle is so fast.
Mark Bonner:Right. Right. You know. Speed of
Reeves Davis:It's faster than anything we've ever seen before. Right? You know, just even look at the introduction of the, you know, personal computer or the Internet or email or mobile. Like, those things are all massive, like, sea changes, but they took anywhere from decades to years. And we're looking at cycles that operate in months or weeks now.
Mark Bonner:Right. I mean, put quite simply, clients are building the tools themselves. JLL and many other companies out there are building tools. Right? They're still coming to JLL, right, for that expertise in that edge.
Mark Bonner:But I think the question now is, is that edge durable? How long do you think it will be before clients have access to data and validity that's just as good as or better than JLL? How are you guys thinking through that?
Reeves Davis:You know, I I think one of the things that we see with this type of, you know, the trends in the industry, you know, they've been running for for a long time outside of this where, you'll see a through line of organizations sort of intentionally focusing on what their core value proposition is. You've seen this in the outsourcing of IT functions or finance functions or whatever else. And I think that the ability for organizations that bring together professionals who know how to achieve those outcomes on a, even a variable basis can create better outcomes for an organization who may be focused on manufacturing or building software for a consumer. Like they might be a university or a hospital system, right? Their core business is not building technology that allows them to make better real estate decisions, right?
Reeves Davis:Or workplace decisions. They are better. They're going to optimize their outcomes by focusing on the thing that they can build that make them more competitive in their space. And I think that they'll continue to lean on organizations that will also invest like people like us and folks in our industry who will invest in technology and leverage the data advantage that we have. That's going to create a better corporate outcome for those clients.
Reeves Davis:So I think it's a very natural phase. Like, look what I can do now. And it's quickly met with, oh, why would I do that? Why would I invest our money as a different organization to build out tools that I'm gonna be the only one that uses? I'd rather rely on somebody.
Reeves Davis:Like, I don't, we don't build our own HR software here. Right? We don't build our own finance software. I would argue that very few people do, and it's because someone else has a has an advantage on that.
Mark Bonner:That's right. And they're looking for someone to validate what they have found because at the end of the rainbow on that, you're making a a major capital decision. Right? And you don't wanna be 98% sure or 99% sure. You wanna be a 100% sure.
Mark Bonner:And if it goes wrong, then it gets into, well, who do I hold accountable? Right? And want to bleed into something that I let off on in the top of the program. Because here's the other side of that Walker and Dunlop test that I brought up. Right?
Mark Bonner:It took it took that executive a week to put it together traditionally. Claude finished it in ten minutes. But guess what? It missed several results. And the risk isn't hypothetical.
Mark Bonner:In an survey, two thirds of companies let employees build or deploy their own AI agents, but only 60% have formal policies governing how. Almost every c suite executive surveyed reported financial losses from inadequate controls averaging roughly $4,500,000. So the tools, as we've discussed, are getting cheaper for everyone, but someone still has to own the outcome. When a client's homegrown tool gets it wrong, Reeves, who's accountable? And what does JLL actually take responsibility for that a client's own tool can?
Reeves Davis:Yeah. I mean, I think it all rolls back to what an organization hires you to accomplish. And in our case, obviously, you introduced me at the top end. I'm a technology solutions person, right? I think technology first in my day to day but the reality is that a firm like JLL, organizations hire us to help them create real estate outcomes.
Reeves Davis:That means, you know, what buildings am I buying, selling, refinancing? How can I lease fully the asset I own or find space as a tenant? Right? How can I make sure that the experience of my employees when they show up to work is as productive and safe as possible? Those are the things that people are really buying, right?
Reeves Davis:The technology that they, that we use, that they use, those that's a means to an end, right? That is what allows us to scale our services effectively to generate consistent data, to make predictive forecasts about what is going to occur in the future. And ultimately, like, the key is someone is hiring us to help create those real estate outcomes and you know, if if someone builds their own technology, that's totally fine. The technology in and of itself doesn't achieve the outcome, right? Someone has to take that insight.
Reeves Davis:It might make the process faster. It might make it more consistent, but ultimately for many of these things, we still have a significant amount of human in the loop that is taking that insight, taking that workflow and making a strategic decision that allows an organization to achieve the business outcome that they
Mark Bonner:want to achieve. And I and I wonder if we're already on the road where is that gonna affect how a company like JLO prices or contracts its work? Right? Maybe it already has or maybe we're we're coming up we're coming to a place very quickly where we're gonna get there. Because ultimately, I think the question is, are clients still paying for the answer, or are they gonna shift towards paying someone to stand behind it?
Reeves Davis:I think increasingly, is that they're paying for people to stand behind it. But the other interesting thing is it's a big industry in terms of what firms like ours do. There are things that like the business arrangement between us and our clients is contingent upon a transaction occurring, right? In other cases, it's a long term staffing model. In other cases, it's a particular project we're managing.
Reeves Davis:And so what you'll see is an evolution of business model over time, but it won't be monolithic, right? Because we're not starting from the same place today. Each of those different areas of the business has a different way of monetizing services and each will evolve from there. And the interesting thing is I always think is like maybe I show my age on this sometimes, but that's no different than it's always been. Business models always evolve.
Reeves Davis:What they're doing right now is they're evolving faster or people have the ability to conceive of should this evolve more quickly today because we can more clearly see a different way of achieving that outcome. And in our minds when we make investments in technology and data at scale our hope is that it creates better outcomes for our clients which makes us more competitive. It might result in a more efficient and effective delivery for the client. They might save money, but in our hope when you get to an outcome based model is you each share in that upside that's created by a shift in the way that services are delivered.
Mark Bonner:If you're just tuning in, we're here with Reeves Davis, JLL's President of Technology Solutions on the AI question, commercial real estate can't dodge. Who builds the tools? Who's on the hook when they fail? And who gets the industry to actually use them? Reeves, I'm gonna go to a couple of questions from our audience.
Mark Bonner:Reeves, what is your comfort level with open source AI and LLMs? Do you look at the cost equation versus security concerns as applied to specific uses?
Reeves Davis:Yeah, it's an interesting question these days, right? We recently had some of the pacing debate come out of the leaders of the major LLM providers. And it's largely those frontier model closed weight systems are four, five months ahead of open weight models these days and some, you know, prognosticated that pacing was a a measure to to keep the competition at bay. You know, I think ultimately these things have a, like they will seek their level, right? Any organization deploying this technology should really first be looking to bring it into a platform that then allows them to scale as appropriate with the right tool at the right time for the right job.
Reeves Davis:So like, I know, like we have a platform here at JLL we call Falcon. And that's where we ensure that we are able to vet these various technologies and to bring them into a centralized toolkit, if you will, that is ring fenced to our networks where the use of those LLMs is contained and the data that we use to deliver our services is not shared out to train other environments. And that we're able to test from information security and cybersecurity perspective fully. So what I see with that is the open weight models play a huge role for some tasks, right? You don't need to use the cutting edge model even if you've standardized on Anthropic or on OpenAI or what have you.
Reeves Davis:Even just within those, you don't have to use the top end model to achieve tasks, Right? And those open weight models, I think will just become an additional option for tasks that don't require the frontier model capabilities to achieve the type of outcomes organizations want to achieve. So think I we're going to end up with a very hybrid world in that way where the right model is used at the right time for the right thing.
Mark Bonner:You bring up a great point. I don't know if you saw this, but last week at Dreamforce, Anthropic CEO, Dario Amodi said that even if AI development froze today, we'd only be using maybe five to 10% of what the technology can already do. So that means the near term value really isn't in the next model. It's an adoption. JLL, as you know, has done a lot on both fronts.
Mark Bonner:You mentioned Falcon. It has put more than $445,000,000 into nearly 60 startups through JLL Spark since 2017. Internally, about a quarter of JLL employees use its AI platforms daily and estimate they save about two hours a week. Its proprietary model has more than a 100,000 users and has processed more than 35,000,000 prompts. So on that point, Reeves, where is JLO where is JLO actually putting its weight right now?
Mark Bonner:Is it pushing capability further, or are you guys driving adoption of what already exist?
Reeves Davis:I I would say it's both. Right? So we look at our AI, you know, capabilities in through three lenses. One is, like, enterprise AI productivity. Something that you or I, even though we're in different worlds would also need to be able to do more effectively and efficiently.
Reeves Davis:These are the things we use to communicate with our clients, to market properties, to respond to RFPs. Like those are relatively general skills and what you need those for has a varying level of need from like a model capability perspective. I would say to a significant degree in that space, we need to diffuse the technology, right? That's the real name of the game is making sure you get adoption and understanding and that you're able to turn that from 25% of those people to a 100% of those people leveraging the technology, right? That's huge value creation for us to scale it.
Reeves Davis:Now there's also second is data, right? What do we know about the industry and the real estate world that other people don't see and how do we make that available to our use? And then the third is things that we think of as like commercial real estate specific AI. Right? And this is both an adoption and a capability thing.
Reeves Davis:Right? This is where you're dealing with everything from, lease abstraction tools, predictive attendance, predictive maintenance, How you tune your supplier contracts to expected attend Like those are pure real estate use cases. And I think in that space that is where I would say we're first focused on capability definition and then diffusion, right? And diffusion is what I mean by scaling and adoption of that technology. So when I look at it, I think enterprise AI, we're largely scaling and adding capability.
Reeves Davis:On the real estate specific use cases, we're focused on capability and then scale.
Mark Bonner:So Reeves, give me a little bit of rope here because we're talking about a few different things and I want to try to pull a couple of these threads together. There are really three questions. Who should build the tech? Who's accountable for the outcomes it enables? And who can actually scale adoption?
Mark Bonner:What I know about JLL is that its answer to the first one has been partly bottom up. Beta testers inside their firm build and try their own tools and agents before they roll out more widely. Rick Forino calls it a citizen driven approach. And one AI founder Bisnow talked to put it bluntly, a top down mandate alone will not work. Can the same company be the builder, the accountable party, and the one scaling adoption, or does that have to eventually split apart?
Reeves Davis:You know, I think it's a reflection of the fact that the people who the problems that people are trying to solve and doing that from a citizen AI perspective, it comes from real deep understanding of what it means to deliver on the outcomes that they're accountable for. And when you go top down with that, it gets very difficult to do two things to really bring in the right number of SMEs, like subject matter experts to define what that should be. And you're sort of starting with the assumption that like whatever we're going to do is going to scale. And the second thing is you have a prioritization problem. When we've noticed this and there's been the basis for some adjustments that we've made over the past several years in terms of how we plan is if you spend all your time evaluating potential use cases for AI, trying to determine the highest value creation ones, you're never gonna stop.
Reeves Davis:You're never actually gonna take action and start seeing value out of the stuff you attack. So we have a model that assumes that all those things are valid, right? That individual productivity and development that I might do to make myself more effective and productive, perfectly fine, right? Something that me and my team or someone else in their team might do to make our work group more productive and effective, that is fine. Then you start stepping into things that actually require true like the infrastructure that will scale, right?
Reeves Davis:What is going to allow us to make sure that the things that are getting developed at that scale are safe and secure and scalable. And so we actually have like a numeric scheme of like, this is zone one, two, two and a half, three. And that creates a real framework for us to talk about what type of opportunity are we looking at and who should be involved to making sure that we develop that. Now, I don't know the answer fully. Your question about like, is the same group who's going to build it, should they be the ones who use it?
Reeves Davis:I don't I don't know that it I guess the way I think of it is I'm not sure it matters. As long as what is there can deliver on the outcomes that are required, then that would be our priority. We look at this this way. There's things that we have engineering teams that build. And then there are partnerships that we have to lean into other major technology platforms to bring their best in class tools and talent together.
Reeves Davis:And who does it? Doesn't matter to me. I just want it to be done. And I think that my opinion is just a reflection of what my clients tell me.
Mark Bonner:Right. And I would have bring up your clients. I mean, I bet you the answer varies depending on what their needs are or the scale of their own business. Mean, how do you determine what what the answer might be per client per situation? What's the tell?
Reeves Davis:I mean, there's always the interesting dynamic of supporting we work with thousands and thousands of companies, but those also do include some of the massive technology providers that are the platforms of choice for this technology. What you need to be able to do certainly for them is, well, they're often going to want to use their own tools to have teams deliver outcomes. That's not a huge cohort to be honest. Right? But then you walk into organizations who also have made commitments to some of those providers.
Reeves Davis:They have an enterprise agreement with OpenAI or Anthropic or Google or Salesforce or Amazon, And they want what they're doing because what our teams are doing, they create real estate outcomes, but they also need to cross over into the corporate productivity that an organization is trying to achieve. So we will meet a client where they are, and we try to bring what we know makes our teams efficient. And then the good news on this stuff is it is a reasonably open ecosystem of how to federate how work is done into smaller and smaller packets. So if you think about like an agent actually orchestrating a series of other agents, You know, a few years now folks have been talking about a multi cloud strategy or a hybrid cloud strategy to mitigate risk. Well, have the same thing with AI tools as well, right?
Reeves Davis:And we need to be able to have our agents that might be built on some piece of technology interact with the clients' agents that are built on another piece of technology. And that's a perfectly valid thing. It does mean that you need people who know how to show up and speak to a client about what it means to implement enterprise technology at scale. And like that's why that's one of reasons I have a job here. Right?
Reeves Davis:My job is as technology solutions. We create real estate outcomes, but my real job is making sure that we can operate at scale professionally and introducing technology alongside a client's technology strategy.
Mark Bonner:Right. And I think that's the thing that you just said is what gets lost in a lot of these conversations. People, humans, talking to other people and other humans based off of the data and the expertise that they have achieved. Why is that getting lost? Why does everybody think that AI and technology is gonna take that person human level out of the equation?
Mark Bonner:Theories?
Reeves Davis:I think that a part of it is the novelty of what can be accomplished and and also like a a very real people are impressed by what they can achieve, right? And honestly, think sometimes me included, you overly ascribe value to that tool and that it creates something. When in reality, it's a probabilistic tool that's pulling together information from the whole of human history and oftentimes the local training that we give it. And it's a reflection of the intelligence we've given it. And I think it's a novel, natural thing to take that impression and assume, well, the human won't be needed.
Reeves Davis:But the reality is the human is absolutely needed to take action in the real world and to achieve consensus, to make a decision, to take some action, right? Those are the things that we still need to do. I just think we're gonna be more effective at doing it with better tools to support us.
Mark Bonner:So let's end on this, Reeves. Back in June, your colleague, Rick Farino, told Bisnow the industry was at an inflection point, confident the ROI was real but unable to prove it at that time. And he asked the question everyone's been asking. Is the juice worth the squeeze? Ultimately, the answer is yes.
Mark Bonner:That's what he said. Three months later, is the answer still yes? And what does yes actually look like in practice from your perspective?
Reeves Davis:I would say it's absolutely worth the squeeze. You've in practice, I think what that looks like is an increasingly mature posture that doesn't look to token max, starts to look at ROI to define use cases, to train people how to work more effectively and cost effectively. And it allows you to scale the best in class knowledge that you have in a consistent way. So to me, where the real juice being worth the squeeze is that I believe it's far easier today to bring new people into the organization and have them be as good as our best performer in a certain area of the business far more quickly. It's a huge boost to their learning curve and their productivity.
Mark Bonner:Why I mean, I I hear you. Why is it so hard? I mean, just going back to the Torsten Slack earnings counts from q two, why is it so hard to measure it? Why does it still feel like it's a feeling instead of something that's showing up on a P and L?
Reeves Davis:This is a very interesting question. And what we talk a lot about, you know, being able to define ROI that appears in such small components in some cases. We set a goal for one area of our business this year and our project and development services to save ninety hours per project manager per year. That's spread across thousands of people, right? That's a very specific ROI target.
Reeves Davis:Now that ninety hours is going to appear in twenty minutes here and forty minutes there and two hours here. And you have to have the tools that you can scale across those thousands of people. And you have to have the ability to measure the actual productivity of those people. And to compare that against the cost increase for token use or whatever else that's associated with that productivity. And that's just a maturity you need to build into an organization that gets real about activity based costing and being able it's back to the old Drucker statement.
Reeves Davis:What gets measured gets managed. And when you're hyper slicing stuff at that level, you really have to also build the tools to be able to identify those productivity gains and to prove that like, it's real. Now I will tell you this. When I when someone says like, well, is the ROI real? And you say, how about I take that away from you?
Reeves Davis:And people are like, no way you're taking this away from me. It's usually a good indication that there's something there and you just need a better job, do a better job of identifying and capturing and showing that value.
Mark Bonner:Okay. Unfortunately, we're running out of time, Reeves. I thank you so much for joining us.
Reeves Davis:Yeah. Absolutely. It was fun.
Mark Bonner:And thank you to our audience for tuning in today. We appreciate it. We'll be back soon with another episode of First Draft Live. You can also find today's episode and all of our past conversations on our website and on your favorite podcast app. I'm Mark Bonner.
Mark Bonner:This is First Draft Live.