AI Agents – From Insights to Action | S5E5
In this episode of the Adventures in CRE Audio Series, Spencer Burton and Michael Belasco explore AI Agents, a revolutionary advancement in technology designed to automate repetitive yet sophisticated tasks. Unlike traditional SaaS solutions, AI agents are highly adaptable, functioning like bodiless robots that seamlessly integrate into workflows without requiring new software or complex interfaces.
Spencer provides real-world examples, including how AI agents can streamline tasks like construction draws and broker opinions of value (BOVs), freeing professionals to focus on higher-value strategic and relationship-driven work. The conversation also touches on the long-term implications of AI agents for both large firms and entrepreneurial ventures, including their potential to level the playing field by increasing efficiency and lowering barriers to entry.
If you’re curious about how AI agents can transform your real estate operations, this episode is a must-listen!
AI Agents – From Insights to Action
Or Listen to this Episode
Resources from this Episode
Episode Transcript
Michael Belasco (00:00):
All right, welcome back to another episode of the A.CRE Audio Series Season 5. Not live, pre-recorded in Miami.
(00:20):
In this episode, this is probably one that I’m most looking forward to because we’re going to get into the topic of AI agents. And this is directly related, I mean, we’re going to talk about a lot of AI topics during this season, but AI agents specifically ties directly into what Spencer’s doing since he recently launched out of Stablewood, which is CRE Agents. So why don’t we kick this off, Spencer? I’m going to kick it over to you. Let’s discuss high level what are AI agents.
Spencer Burton (00:53):
Mm-hmm. Yeah. So these are not real estate agents, although when I say CRE Agents, occasionally I get that. AI agents are a bodiless robots that can reason and, when properly taught and equipped with the right tools, can perform repetitive tasks that free up time for us.
Michael Belasco (01:18):
Okay. So give us some examples. That sounds incredible. We all have these repetitive tasks. God, I can think of just from day-to-day stuff. I mean, there’s a lot of stuff in real estate, but I mean, every month I got to look at all my credit card bills, my mortgage statement, pay my bills. These are repetitive tasks where it sucks the life out of you, right? And it’s just this incredible opportunity. And it’s not a simple solution. It takes a lot of sophistication and know-how, wherewithal. Give us some examples of how this can be implemented.
Spencer Burton (01:52):
Yeah, so I’m going to give an example from RVAT. So Michael, if you’re unaware, haven’t heard the episode from this season or the past, has a RV brand called RV @, and he has a member of his team that processes construction draws, okay? So I am in this early build stage, and so I’m always looking for repetitive tasks that I can work to automate, and each time I do that, it helps me, it helps my tech team as we’re launching this platform to understand the challenges better, and one of the challenges is they have to do these construction draws. Okay? And a construction draw is all this work was done at the site, but I have to prove that the work was done, and the proof comes through an inspection, perhaps photos, but most importantly receipts, receipts that work was actually done for some amount. That is compiled into a report and then sent to your investors, sent to your lender, who then provide you capital so that you can continue to operate, right?
(03:06):
And so a member of Michael’s team reached out to me and said, “I am drowning in receipts.” Now, by the way, what we developed for this individual is not particularly cutting-edge, because there are software tools out there that will take a receipt and will use the appropriate technology to extract data from the receipt to help with this. However, none that I’m aware of, or at least that are out there marketed, are like AI agents. And so I created an AI agent for this individual that takes the receipts, extracts the information that matters for you all, and then imports it into your CSV file, into your Excel file.
(04:00):
And why does that matter? Okay, well, you could go out and license one of the software tools and you could use that software tool in order to speed up that task, but the software tool would export it into its own format and then you either have to adopt their format or you have to transpose it from their format into yours. When we taught this AI agent to do it, it just did it the way you always have done it. It took in the receipts and that it exported into the format that you’re familiar with, that your bank’s familiar with, your investors are familiar with.
Michael Belasco (04:35):
Which was total… You didn’t have to learn a new system, you didn’t have to learn a new way of doing something. This leads me to my next question, and you’ve kind of already hit at it, but tell us what is the core difference between agentic AI and a typical SaaS offering, software as a service, which has been the way of the world for decades now.
Spencer Burton (04:56):
Yeah, yeah. So software and early-2000s SaaS, software as a service, and software as a service came around because the protocols in web browsers allowed for you and I to run programs in a browser, but that actual application was running locally but through your browser, right? And I won’t get into the technical aspects of it, but effectively your browser became an engine to run software. And when that happened, software could become a service, and thus was born SaaS. And there’s other things that were part of it, and I’m not getting into the history and I’m not a historian on SaaS, but essentially that’s what it was, and for decades now, we have used software in order to make our lives a little better.
Michael Belasco (05:44):
Yeah. Everything from like Salesforce to QuickBooks or everything like that.
Spencer Burton (05:49):
Yeah. And we all use a ton of software, right? And software’s great.
(05:52):
What’s the problem with software? The first problem is that in order to software, you have to speak a different language than most of us speak. That’s the coding language. But because of that and because software is so valuable, a lot of people went out and learned to code to write this special language. But software is intended to solve a problem for a user. The user doesn’t generally have this specialty or able to speak this special language. And so what you have is you have a user who says, in their own mind first, “Oh, I wish X, Y, and Z,” and then over here you have the software developer who through some means discovers this problem. The problem is they don’t really speak to each other.
Michael Belasco (06:40):
Yeah. You have a person creating a solution for someone, but he doesn’t experience the problems firsthand.
Spencer Burton (06:46):
And so what generally happens is you have a translator in between. Okay? Product manager is one of the titles of this individual, and this product manager lives between the nerdy software developer sitting in a dark room somewhere writing code that most of us don’t understand and the user who actually understands the problem but, first off, doesn’t know how to architect a solution. They may have ideas, but they don’t know how to architect a solution, let alone know how to code. And so the product manager’s job is to grab this problem, and the product manager may be a startup founder who also happens to code in some context, but generally once you get to critical mass and a software company, software, you got the developer, you got the translator, and you got the user, and the user who knows the problem but doesn’t exactly know how to architect a solution, this piece in between helps develop an architecture for the solution, gives it back to the software developer, that software developer writes thousands, or tens of thousands, or hundreds of thousands of lines of code in a different language ultimately to solve this problem.
(07:49):
And then early on, this software developer would go away for a year and come back and hand the solution back to the user and the user would say, “That doesn’t actually solve my problem.” And so agile and some of these other methodologies were introduced, would essentially the software developer would go into a dark room for three days or for two months, come back with a minimum viable product, an MVP, deliver that back over here, the user would say, “Yeah, it gets me about 10% of the way,” give feedback, the feedback would come back slow, expensive. And think about this, that problem, and generally the problem is solved through what we call a feature, or it may be like the core software, but over time, more and more problems pop up and features are developed, and every feature is thousands, tens of thousands, hundreds of thousands of line of code, which, by the way, you want to change the feature, you’ve got to modify the code. Something’s wrong in the code, someone who actually knows how to read this other language has to go into it.
(08:48):
So that’s software, okay? AI agents, once you have the foundation, there is no software developer, there is no other language. The solutions are still architected, but the features are written in natural language, okay? Such that the solutions happen faster, they’re easier to audit, and they’re closer to the user.
Michael Belasco (09:20):
So I’m hearing two sides to this story. One is just the actual development of a solution. There’s a problem and a solution and a way to solve it that has been revolutionized essentially, right? So that’s sort of what you described here. You described this whole story of this problem, takes a year to solve, it doesn’t actually do it, there’s this communication breakdown. And it’s inherently messy, right, this software process? We’re generalizing, right?
Spencer Burton (09:51):
Yeah, yeah, of course.
Michael Belasco (09:52):
But then there’s the end user output product that comes from this where like you had to have an interface, you had to buy a program, you had to install something. You know, there’s the software, you have to integrate it, right? Where it’s not your company’s integrating into a software. Now you have kind of an employee who is integrating into you and not only just learning your skills but then taking it a step further. I mean, it’s beyond like, “Oh, you have a problem? Let me try to solve it,” and maybe that problem is obsolete by the time I come… This is like real-time practically human, beyond human. Some of these capabilities, I remember I was watching something where, I mean, it’s known like these models are solving things that humans haven’t thought of yet. So that is what you’re saying here basically? Or did I take it wrong?
Spencer Burton (10:44):
Essentially. Well, what I’m saying is producing a solution to the problem is highly complex, very expensive in a SaaS context. On top of that, the user, in order for it to solve its problem, must learn a new way, whether that’s change their workflow or learn the tool. I mean, all of us have done it, right? We get a new app, it’s going to save so much time, and a lot of times it does, but a lot of times we spend so much time learning that new app only to discover that it actually doesn’t entirely solve our problem and it creates other problems. AI agents think of it this way. Number one, the solution is much closer to the user, but number two, the user doesn’t have to learn a new application. What it does is it’s almost as if the user is outsourcing the problem to someone else. That someone else happens to be a bodiless robot.
Michael Belasco (11:44):
Wow. And not that every software solution is a one-size-fits-all, but like you have your own customized problem-solver with these, as long as it has all the data and all the things that are needed to be plugged in. Is that-
Spencer Burton (11:57):
Okay, so let’s bring it to the real world. Okay, so the user’s problem is that they have to produce a broker opinion of value. Every single day, it takes eight hours to do, and because it takes eight hours to produce that broker opinion of value, that broker is spending too much time creating BOVs than actually bringing in new customers, going to lunch with people, going to dinner with people, talking to people on the phone. And so what do they do? They go and they hire an analyst to come in and do the BOVs for them, which is expensive, and that analyst doesn’t really want to do that for the rest of their career. So that means every two years they’re having to teach a new analyst how to create a broker opinion of value for them. That’s their problem.
(12:39):
So what’s been a possible solution is SaaS, so a software tool that helps create broker opinion of value. And so the broker either uses that tool themselves, and so an eight-hour process becomes a two-hour process, or that the analyst that works for them can now produce four BOVs in the same time that the analyst would’ve produced one, okay?
(13:05):
The problem still exists, which is it still takes two hours to do it, and your poor analyst is not enjoying doing that work, okay? But imagine if that broker, instead of outsourcing to a human analyst to do that very tedious mind-numbing task, could outsource it to a bodiless robot who was diligently doing it all the time, 24 hours a day, 365 days a year, happy to do it, and the broker and the analyst could both lift up and, instead of focusing any of their time on BOVs, could focus their time on things they’re great at or the things they actually enjoy doing. And instead of bringing in a software solution, they just simply hand it off to this bodiless robot who produces it for them.
Michael Belasco (13:50):
Yeah. And that’s the incredible, I guess, opportunity here. You had mentioned to me, you mentioned this a lot, it’s vertically integrated, right? Vertically integrated. And what’s the difference when you talk about vertical integration versus horizontal? Why don’t you give us-
Spencer Burton (14:06):
Yeah, so you may have heard of the concept of vertical AI, there’s also vertical SaaS, and therefore the alternative to that is horizontal AI, okay? Vertically-integrated anything means you go really deep but narrow. Horizontal is your thin and wide. So vertical AI refers to and vertical AI agents refer to agents that can do very technical nuanced things. They can go very deep. But they don’t go wide. So we’re building CRE Agents, which is a vertical AI agentic platform for commercial real estate. You will not train these vertical AI agents to plan your annual trip to Maui. You, I guess, probably could, but there will be better solutions, horizontal AI that will do those more generalist tasks for you. They will not be your email organizer, they won’t be your schedule. That’s not what vertical AI is.
(15:13):
Vertical AI is, let’s think of a very technical task, perform loan valuations for your entire book every quarter for regulatory purposes in a very defined way. Or you are an ODCE fund, an open-end diversified core fund, and you have to do reporting in a very specific way. You’ll use a vertical AI who does that. Why? Because they’re going to have very specialized instructions and they will be given access to very specialized data and very specialized tools, and those things are necessary in order to perform these very specialized tasks. You won’t expect to OpenAI to develop integrations to highly specialized tools or to train models on highly specialized data. They’re not going to have access to that data. Let alone expect OpenAI to write instructions for you for highly technical, very specialized tasks. And so that’s where you get vertical AI.
(16:24):
Horizontal would be… OpenAI is going to release an AI agent tool at some point. Google already has one. I think it’s called Vertex. Microsoft, obviously, has theirs. That’s the Copilot. Those are horizontal AI. They’re intended for users to develop generalist AI agents. Those are absolutely necessary. You’re going to see a ton of those hitting. They’re already there or they’re already here and more and more are coming. But the vertical AI, it’s a much bigger lift, takes longer to build, but they also perform expensive, important tasks that in industries like ours allow analysts, associates, all of us to lift up and focus on the things that we do best.
Michael Belasco (17:09):
Yeah.
(17:10):
Let’s shift to somewhat of an elephant in the room here, which is, and we talk about this a lot and I think we alluded to it earlier, so you have AI agents doing very… They’re repetitive, but somewhat sophisticated. They’re sophisticated, but repetitive tasks, right?
Spencer Burton (17:26):
We don’t think of them as sophisticated because we’ve done them for years and years in our career, and we just think of them as mind-numbing, bang-our-head-against-the-wall sort of tasks. If you talk to anyone outside our industry, they would be highly technical, nuanced.
Michael Belasco (17:42):
Now, traditionally you’ve needed a handful of people to do this stuff, right? And this skill set… Or it’s not even a skill set. It’s like this task then shifts, right? And you say this, and I believe this too, it’s like people will shift to be able to do more high-value things, right? But there is this reality out there where like now these tests are no longer needed, right? And this is coming, by the way, in every industry. I mean, it’s just a matter of just where the world is going. And so how do you see from the ethical perspective and like how do we safeguard people from… Or what are your general thoughts? And I have my own thoughts too, but what are your general thoughts on how we go into this? It’s coming, right? This is happening.
Spencer Burton (18:35):
Yeah, yeah.
Michael Belasco (18:35):
And so maybe we save this, maybe this is another episode we have, but…
Spencer Burton (18:38):
Let’s just touch on it really quick. So there’s like in my mind this graph with two lines that look like this, okay? To each individual, one line is the value of a dollar to us, and it’s time-series’d, so the left end of the X-axis is when we’re born and the right end is when we die, and one line is the value of a dollar and the other line is a value of an hour of our time. And paraphrasing a little bit of what Warren Buffett said, which is the older you get, the more valuable an hour of time is.
Michael Belasco (19:23):
Yeah.
Spencer Burton (19:24):
The older you get, the less valuable a dollar is. I saw it as a young land broker where I would go sit down with a 92-year-old man and we’d talk about buying his land. You could offer him a billion dollars and it would mean very little to him, because he’s not going to ever see that, right? He cares about those last few days or months or years of his life. And they would often say to me, “Spencer, I don’t want to spend the last year or two of my life negotiating the value of my land.” Okay?
(20:05):
And so where I’m going with this is, you may be young in your career and you’re thinking about the value of a dollar. I’m at a stage where I’m thinking a lot about the value of time, for me, for the people I care about, and for the broader CRE industry, and I go, “If I could create something that gave time back to people, that’s something worthwhile in my life.” Now, the elephant in the room is this technology is going to cause disruption. Roles that used to be performed by people will be performed by digital coworkers, by AI agents.
Michael Belasco (20:46):
For the long-term net benefit, but for the immediate, there’s some disruption, right?
Spencer Burton (20:49):
Yeah, the immediate, there’s real disruption. I’m not blind to that at all. In fact, when this idea began to became real, I called you, and the very first thing I said was, “This is real. This is happening. I’m thinking of taking it on. But my concern is people, this will create disruption.” And we’ll have a whole episode and we can kind of talk through it. The point is, it’s going to happen. To me, though, the net benefit is it gives back people time, and time to me is far more valuable than dollars. Although if you can’t eat, then what’s the point of the time? What’s time worth to you? But-
Michael Belasco (21:36):
Let me add, because what has happened to date is a lot of people have to learn skills that are mind-numbing. That’s a great word to use.
Spencer Burton (21:43):
Yeah.
Michael Belasco (21:44):
There’s going to be a shift in where people need to retain higher-value skills because these, they’re repetitive. It’s an opportunity, and that’s how I look at this too when I think about this. And what I love about just our platform in general is that we have an opportunity to prepare and make sure… You know, what are the new skills? There are going to be new skills. It’s not like jobs are going away. It’s not like when the Industrial Revolution happened, jobs disappeared. Things shifted. I mean, it’s a thing we talk about. When I was at Heinz, the long wave of office, right? When shift went to white-collar work, right?
Spencer Burton (22:25):
Yeah.
Michael Belasco (22:26):
Anybody that built an office building anywhere became successful. But it’s an opportunity to grow and become more sophisticated. And yeah, like you said, these were, to the outside industry, seems like technical stuff, but it’s really just like there’s no deep thinking into any of this, and so the opportunity here in my mind is like now people can start to have deep thoughts, shift into higher-value things they can learn, and now that level of thinking becomes accessible earlier on in your career, is sort of where I think. It has to be done very responsibly, I think. I think there’s no better person in the industry, I think, or even with our platform, you know? But like for CRA Agents, so-
Spencer Burton (23:08):
No, and we should talk about this more in depth on a shorter episode because it’s a important topic. I think the broader point though is time matters, time’s important. We should spend our time on things that add higher value and that bring more fulfillment. And so when I think about it in the context of real estate, that’s relationship management. I’m not going to trust an AI agent to manage my investor relations, certainly not my face-to-face, certainly not my phone calls and most of my email conversations. So relationships, then there’s strategic thinking. I’m not going to trust my AI agent anytime soon to do-
Michael Belasco (23:49):
As a companion-
Spencer Burton (23:52):
… to think strategically. As a companion, it’ll speed that up, and certainly the AI agent will produce outputs that get me to decisions. That’s the third piece. Decision-making to me is very much a role that we continue to play and will continue to play for the foreseeable future in real estate.
Michael Belasco (24:05):
Can I ask you one other question?
Spencer Burton (24:06):
Yeah.
Michael Belasco (24:06):
I want to get your thoughts on how do you foresee the playing field leveling out when AI agents are introduced between like larger institutions and the more entrepreneurial shops. Do you think the entrepreneurial shops, it levels the playing field and gives them, maybe there’s some focus there, or does it actually… What are your thoughts?
Spencer Burton (24:27):
So again, think of AI agents as inexpensive work power.
Michael Belasco (24:38):
Yeah.
Spencer Burton (24:39):
I don’t know.
Michael Belasco (24:40):
It’s what it is
Spencer Burton (24:41):
Yeah, I don’t know how-
Michael Belasco (24:42):
That’s fine. Yeah.
Spencer Burton (24:44):
And so if you are a small firm, you now have access or you may have access to a agent that can perform a lot of the tasks that you did before and, therefore, that frees you up to do more and whatever, and if you’re a big firm, you’re going to have the same thing. But look, the edge for everyone is data, and AI agents don’t come equipped with some special data. AI agents basically perform tasks.
Michael Belasco (25:12):
They allow you to ingest data. I’m experiencing it in my own world. I think it’s created a competitive advantage. I’m not using AI agents, I’m using the generic stuff that’s out there, but it’s given me my productivity, where you don’t have the financial resources of a large shop. It actually has leveled the playing field out in that regard. Now, the AI agents is not exactly what I’m utilizing today. I just-
Spencer Burton (25:40):
You’re using a precursor to AI agents. But still, if you really listen to what you said, you said it’s your strategic thinking that’s adding the value, because what you said is, “I’m going to make a concerted effort to mine data that doesn’t exist in any other shop.” That’s what you’re doing, and you’re using this technology to speed that up or make that possible. And it’s still the edge comes from you thinking about how you can develop an edge and then leveraging this technology to produce that edge, okay? But the question is, who has the advantage with this technology between big firms and small firms? I’m not sure that this levels the playing field, although it makes doing more with less possible for the little guy. The edge, though, ultimately comes through data, how are you aggregating data, are you using the appropriate tools to do so, is that data giving you some edge where you know something that the market doesn’t know. Like, all those things don’t change in this new world.
Michael Belasco (26:43):
Yes. That’s true.
Spencer Burton (26:45):
So what this does for the big guys is it makes them more efficient. If they’re more efficient, they can lower their cost of capital. If they lower their cost of capital, it makes it harder for you to compete.
Michael Belasco (26:56):
Yeah.
Spencer Burton (26:57):
So on the one side, they have that advantage. For you, though, it gives you tools that you didn’t have before. Instead of having to hire 10 people that you can’t afford to hire, you could bring in an AI agent to do a lot for you. You could hire five. It sounds like people are going to lose their jobs, but-
Michael Belasco (27:11):
And it’s a massive industry. Yeah, but it’s like where jobs may have not have been created in the first place.
Spencer Burton (27:15):
Well, then that’s the flip side of it, right? So what this does is it increases human work in other areas. That’s the elevating concept.
Michael Belasco (27:24):
It could also create more entrepreneurs where somebody else might’ve gone into a shop, they go out and… I mean-
Spencer Burton (27:29):
That’s true.
Michael Belasco (27:29):
… like you and I, we’re both die-hard, like I love just thriving in a competitive environment.
Spencer Burton (27:35):
Yeah. I mean, maybe the new world is, instead of having a concentration of specialists in big shops… Because we’re all entrepreneurs anyway. I mean, you look at developers at a place like Hines, I mean, they’re all entrepreneurs even though they’re working at Hines. I’m not suggesting that all these Hines developers are going to leave and go start their own shop, but it does lower the barrier to starting your own shop.
Michael Belasco (27:55):
Yeah, it absolutely does.
(27:58):
So this is awesome. Anything else you want to touch on on AI agents? I feel like we hit a good… Is there anything that I missed asking you that you want to hit on before we wrap?
Spencer Burton (28:08):
I think everyone’s like, “Okay, when do I get to work with one?” So right now you can use a horizontal AI platform to build an AI agent to help you with generalist tasks, in the same way that many of you are using ChatGPT and others to make you better with your generalist tasks. That’s possible right now. Vertical AI is coming. 2025, we’ll get into CRE Agents on another episode, but we’ll be working with firms, developing AI agents for those firms, and very soon we’ll have AI agents that companies across the country can use to perform more nuanced tasks.
Michael Belasco (28:43):
Amazing. Awesome.
Spencer Burton (28:45):
Yeah.
Michael Belasco (28:45):
All right, well, this wraps up our episode on AI agents, and we will see you on the next one.
Announcer (28:53):
Thanks for tuning into this episode of the Adventures in CRE Audio Series. For show notes and additional resources, head over to www.adventuresincre.com/audioseries.
