In this episode of the A.CRE Audio Series, the team explores the impact of artificial intelligence and commercial real estate. While the last episode talked about AI in general, this episode discusses how it applies to CRE. This episode explores how AI is reshaping careers within the industry, and how roles may evolve in the future. From AI-driven screening interviews to the prospect of customized, in-house tools, the episode touches on various use cases that could revolutionize CRE.
Tune in below to explore the possibilities and challenges AI is bringing to commercial real estate.
Artificial Intelligence and Commercial Real Estate
Or Listen to this Episode
Resources from this Episode
Sam Carlson (00:00):
We finished the last episode talking about AI and how it’s … How we’re using it on the site. General discussions on AI. Now I think we want to talk and just jump into commercial real estate. Commercial real estate, not plural. Commercial real estate and how it’s going to affect careers, positions that are existing and have a certain status quo today, and what they’re going to look like a year from today, five years from today. I don’t know where we start. This is interesting. The beautiful thing about real estate is it is tangible but there’s evolution, and speed, and efficiency, and opportunity. Opportunity galore coming if you do it right. So let’s jump into wherever we’re going to jump into with this one.
Spencer Burton (00:47):
Oh, yeah. I mean, there’s so many examples. Let me start with one that is at the base level and that’s when you get your job. Let’s just start there. There’s this process where a hiring manager drafts a job description. You think about it, right? Drafts a job description, and there’s a tedious process that goes into that. And then they post a job on a job board and then resumes start to come in. And there’ve been tools, thus far, AI sort of tools but nothing that compares to what’s coming now.
You get 100 resumes in and you go okay, there’s 15 that look like they could be a fit but I don’t want to interview all 15, or at least I don’t want a subject matter expert, one of my practitioners, interviewing all 15. The hiring manager, generally a part of the HR team, will call all 15 and do a screening interview. I think those screening interviews are done by AI, or they should be, or most of them will be. That’s my prediction. In fact, I would assume many groups are already thinking that. Because then you can quickly go from 15 to five and then you can do now the human interviews. Anyway, that’s the first use case. We could literally come up with hundreds.
Michael Belasco (02:15):
This is a use your own imagination. Where we are in the world today right now this is just-
Spencer Burton (02:19):
Some are imagination some are real but yes.
Michael Belasco (02:28):
There’s a lot of third-party technology that people pay a lot for subscriptions for. I think with ChatGPT … Sorry, ChatGPT. ChatGPT is on my mind. I think with all the AI that’s coming out and all the access you get to advanced coding and advanced technology, there’s going to be a plethora of just in-house, exactly customizable to what I’m doing software that comes out that people are going to use. Whether it’s for property management, asset management, portfolio management, even people’s in-house acquisitions. I don’t know that Excel goes away. Excel probably does go away for acquisitions but not in terms of training, which is actually interesting because it’s so transparent which is why we use Excel. But I think all of those things just proliferate. It speeds up much faster and people get their own customizable in-house.
Spencer Burton (03:19):
You’re describing bespoke tools, right?
Michael Belasco (03:21):
Spencer Burton (03:21):
Bespoke tools. It’s interesting. Many know here that I work at Stablewood and our CTO, he and I were having this conversation about almost this very thing. His prediction is that software as a service largely goes away.
Michael Belasco (03:37):
Spencer Burton (03:38):
And the reason why is because with these tools you don’t need to use a solution that someone else built you could build your own solution.
Michael Belasco (03:46):
Spencer Burton (03:46):
And you pair no code, which has already been building momentum, with now natural language models. And so basically you prompt a tool that then builds what you need. Why would you go through the headache of setting up a Salesforce CRM, I mean, that’s a process in and of itself, if you could actually just build your own CRM in the same amount of time?
Michael Belasco (04:10):
Spencer Burton (04:10):
Right? And so that’s what you’re describing. I mean, coming back to the recruiter, that’s in essence what they’re doing. They’re building a tool, a bespoke tool, that allows them to screen candidates quickly. Let me ask this, right? Let’s say you’re a job seeker-
Michael Belasco (04:34):
Spencer Burton (04:35):
And all of a sudden this changes and you no longer have a screening interview to get past that first round you have some form of an AI interview. How does that change the way you prepare?
Michael Belasco (04:49):
I think first and foremost you got to realize that you’re going to be recorded. You go in there and so you’re going to need to be a little more polished than you think because that is going to live on forever, many people are going to see it, there’s no word of mouth of how did it go so you’re going to need to prepare more. The way you do that is by getting coaching and prepping. This leads back to what we were talking about before where you just need to get your reps in and just understand what may be coming. Using these language models, using ChatGPT to train you is what you’re going to need to be focusing on. Those first-round interviews used to be all about how do I make that personal connection.
And we used to coach people all the time. And we still do we still do that. It’s the beer test or the plane test, whatever you want to call it. You want to impress upon this person that if you’re stuck in an airport or you’re in a plane for a long period of time or you’re out getting a beer, this person’s going to enjoy spending time with you. That somewhat goes out the window because there’s a lot of nonverbal cues that happened in those first-round interviews. I don’t exactly know how this all transpires, but for me it’s you have to be much more technically proficient and just have much more solid answers that are less about personal interaction or a warm smile or anything like that which is going to be interesting. I didn’t really think deeply about all that but that interpersonal-
Spencer Burton (06:26):
Well, imagine this. Take the tool Vowel, V-O-W-E-L.com, right? So they are a zoom alternative that’s AI-powered. How it works when you’re in a Vowel meeting is it records as the meeting is going on or a one-on-one. As the interview goes on it auto-produces a transcript, and then an AI is reading the transcript, summarizing it. Basically creating a meeting notes in real-time. You can call to certain segments of the meeting or the interview using this. Imagine a hiring manager using that and you are the candidate. And imagine the hiring manager now can clip pieces and share that with the subject matter experts who ultimately are going to make the decision, right?
And this is no disrespect to the hiring manager, they’re not experts in commercial real estate. They’re experts in finding talent that they can then pass to a team that does the second-round interviews. In this way, it’s almost like the first-round is your second round. Being able to demonstrate competency at that first stage is going to be key, right? They may ask technical questions in that first round that they otherwise wouldn’t have that will be seen because that interaction was recorded, transcribed, summarized by an AI. How you answer those first-round questions actually matter a lot more.
Michael Belasco (07:52):
Well, you know what’s going to be interesting is the classic question, do you have any questions for me, right? Do you prepare for that? Is the AI model going to be trained to answer those questions? Sorry.
Sam Carlson (08:07):
No, go ahead.
Michael Belasco (08:07):
And I wonder if part of the interview is seeing how well you know how to manipulate and deal with the AI interaction that you’re having because that is key to being prepared for this change that we’re all experiencing.
Sam Carlson (08:21):
Okay. I think we can agree that a person’s ability to work with AI is going to be a new job requirement if you will. And specifically within what’s happening in a commercial real estate transaction whatever it is. So one, you’re talking about proficiency which is … You touched on that. But two is, okay, if that person’s going to be coming in they need to be proficient in whatever subject matter it is, and they need to be proficient in the new tactics that we use to execute our strategy. Do they have experience prompting whatever it is … Whatever they’re using? Or their own model, their own whatever it is. So that is where you’re going to jump straight into, are you going to be able to produce an outcome faster and better than that person or not? I think we just get to the outcome and the result a lot faster by having AI see … Hey, how proficient would this person be in dealing with me-
Michael Belasco (09:30):
Sam Carlson (09:30):
In getting outcomes, right?
Michael Belasco (09:32):
Yes. Yes, yes. I want to shift a little bit if that’s okay-
Spencer Burton (09:36):
Michael Belasco (09:36):
Into the actual day-to-day of just being in commercial real estate whatever role you’re in. And Spencer, when you and I worked together there was this conversation about … You bring these experts, these great talent, talent from some of the best places in the world … Sometimes you find these hidden gems, and the hidden gems are everywhere, right? You bring these people in and there’s this constant problem with inundating these people with mundane tasks. Meeting minutes is a perfect one, right? Having this person sit there with their powerful brain, and making sure that they’re taking adequate notes, rather than critically thinking about the conversation that’s being had and being able to respond and offer ideas of which they’re more than capable of doing.
Spencer Burton (10:35):
Or even being part of the meeting. So that’s the other thing, right? Sometimes you’ll have a meeting, it’s 30 minutes, and someone joins the meeting because there’s five minutes in which they need to hear something. There’s an AI attending for them that then summarizes their piece and they read it rather than attending.
Michael Belasco (10:55):
Like being the crowd in the meeting, you don’t need to be there you just need to be there for that one piece.
Spencer Burton (10:59):
Michael Belasco (11:04):
That’s one low-hanging fruit opportunity. And I think the more challenging stuff which is solvable … And I’d love to hear your thoughts about this. But there is going to be innovation in actual underwriting. Again, people who toil around in miles which is critical in your early stage to really understand the foundations of how financial modeling works and how the cash flows work. We call it a revolution, to where again, these brilliant minds aren’t spending late hours just crunching numbers. You might fundamentally understand what’s going on but you still have to crunch all the numbers and you’re not adding value there. So there’s going to be-
Spencer Burton (11:50):
Oh, it’s happening,
Michael Belasco (11:51):
Spencer Burton (11:52):
Michael Belasco (11:52):
Speak to that, speak to that.
Spencer Burton (11:56):
I don’t want to get too much into call it the proprietary of Stablewood, but let me give an overview and you can fill in the gaps yourself. If you think about real estate underwriting it’s an exercise first in aggregating information, right? So you can’t make an investment decision until you have complete information, or hopefully you have complete information. And ideally, you have information that the rest of the market doesn’t have. Because if you have information the rest of the market doesn’t have you can either gain alpha or you can avoid downside. That process of aggregating information, in the traditional real estate way, is a laborious process. And because it’s such a laborious process A, you need more people, you oftentimes need higher-cost people because they need to know how to find the right information, what information is relative. And because it takes so long and has a higher cost to it you can look at fewer deals.
And so I think about a traditional institutional acquisitions team. OM comes in the door, it hits a producer, director, VP, someone who’s seasoned in these things, and they say, “Okay, that looks like it fits within our box,” right? So they’re actually the first screen. That OM comes in the door. And maybe they get 10 om OMs in a day and one of them they go, “Okay, this makes sense.” Now what do they do with the other nine OMs? They throw them in the trash. I mean, they’re PDFs but you get the point, right? They throw them in the trash.
If you think about those nine OMs, they’re full of really interesting, valuable information that they don’t capture, right? But they can’t because it’s just their eyes, and brain, and a notepad. They pass it to an analyst, and what does the analyst do? The one out of 10 the analyst reads the OM. How long does it take to read a 40-page OM? It takes a while, right? They read the OM and they extract from that OM the data points that matter. They put it in some format that they can retain it. In a most traditional way Excel. Maybe on a notepad, maybe in a Google doc, or in a Word doc. They extract information and then they begin the process of accumulating other information that creates increased insights around that property.
So maybe they go to CoStar and they’re saying, “Okay, what rent assumptions should I use? What vacancy assumption should I use?” Maybe they go to comps and they do okay, let’s pull an insurance comp for this state. Florida right now is a disaster when it comes to insurance. But you get the point, right? Maybe they’re doing a property tax analysis and so they’re going to a tax assessor’s website and they’re pulling down the mill rate for that property. And then they’re pulling tax comps. The point is, in a traditional institutional acquisition shop you’re several days to aggregate information. I really enjoy family history so I like to learn about who I’m related to, right? My parents actually met in a family history library, right?
Michael Belasco (15:17):
Spencer Burton (15:17):
Back in the old days they called it genealogy, right?
Sam Carlson (15:20):
Spencer Burton (15:20):
So my parents met just researching ancestors. This relates to real estate. Up until, I don’t know, a few years ago the process of okay, if you wanted to find out who your seventh great-grandfather is you would have to find say a census record from maybe your sixth great-grandparent and that would show who their parent was, right? You’d find it in a library somewhere and then you’d find the census record that showed that they had lived in the same household and that sixth great-grandparent was a daughter or son of your seventh great-grandfather. And once you find that you go, “Oh, okay, interesting.”
The good old-fashioned way … I remember my dad doing it just on paper, right, he’d write it down and that’s how you find out who it is. What they have now is an AI that does that searching for you. It searches all the trillions … I don’t know how many there are. Billions or whatever the records of census are and say … It will search this and say, “Okay, I think that this record relates to a potential relative of yours, right? That AI does all the work that in the past you would’ve had to go to the library, search through to find that census track. It does that-
Michael Belasco (16:40):
Spencer Burton (16:41):
For all of your relatives. Come to real estate. Imagine a tool that could search the entire world of data at your fingertips and say, “Here are all the data points that relate to that one property.” And instead of the analyst spending two or three days doing it, it was instant for them, right? That’s one example of how this will just accelerate or elevate what we in commercial real estate do.
Michael Belasco (17:11):
I am adding not a point almost an … I don’t even know what you’d call it.
Sam Carlson (17:19):
Michael Belasco (17:20):
An observation or just a remark. The competitive advantage of larger real estate shops or shops that are established is … Obviously, there’s capital to invest in resources, right, to invest in data, and that effectively gives a lot of these shops a competitive advantage. What’s happening now is the data’s becoming-
Sam Carlson (17:51):
Sorry, go ahead. I’m having this thought but go ahead I’m not going to interrupt yours.
Michael Belasco (18:01):
This concepts of the big data’s been around forever, that’s not new, it’s just becoming … In the real estate world, it’s been slow to adapt but it’s becoming more and more available. I don’t know how this plays out. I don’t know where I’m going with this one.
Sam Carlson (18:18):
Well, let me just jump in here because I think when you look at the way that a small company operates compared to the way some of these huge companies operate, there’s that analogy of how do you turn around a huge cruise ship? One degree at a time. I think that we are going to start seeing smaller firms have a huge competitive advantage over larger firms, mostly because they’re going to adopt a new way of doing thing and a new tool faster than the big companies. The big companies have people, have legacy, have tradition, have … It’s always been done this way. Have operations, SOPs. Everything built on top of people and all that stuff. You get a small family shop come in and they can do … Name how many deals come in their door. They can do that and all of these things by sourcing automation plus AI. Automation plus AI. Go ahead.
Michael Belasco (19:32):
Before shops would need to hire subject matter experts, right? For example, people who know how to code to be able to then gain access to these tools. You mentioned no code. You won’t need that subject matter expert so it almost … It levels the playing field. You’ll be able to find the resources you need and build them with off-the-shelf things that are coming to you daily and you’ll be able to start competing. So I think what’s happening now is a leveling of a playing field for everyone, not only in commercial real estate.
Spencer Burton (20:07):
It’s interesting. I go back and forth between whether this benefits the big companies more or the smaller companies more. I agree with you that companies with legacy issues, legacy tech, legacy processes, tradition is a significant barrier to adoption. Now, whether they can use it or not is a different thing, but they have data, data.
Michael Belasco (20:37):
Spencer Burton (20:38):
And data is far more valuable today than it was six months ago because of AI. And if you can train a bespoke AI model on your data you win. They have more data than the small companies do. But to Sam’s point, can they actually accomplish it? Most of the big institutional real estate firms that I’ve interacted with, they’re too cumbersome, there’s too much bureaucracy. And no offense to any of our peers. I think they’d be the first to admit it.
In fact, I was speaking to a good friend this past week, I won’t name him, who works at one of these big firms. He was lamenting how long it takes to produce an investment committee memo. And I was describing the investment committee memo. Well, we automate our investment committees memo so they’re automatically produced. He was describing this process of 10-plus people, and dozens of hours, 40-something pages all manual to produce an investment committee memo. And he’s like “If only we could adopt what you were doing.” He’s like “It’s just impossible though.” And they’re trying to do it incrementally. It’s like okay, this one page out of 40, let’s see if we can automate that one page. And the barrier to that is so difficult because everyone’s done it that way forever, this is the technology they have, they got to go through multiple levels of approval to change any of this.
Michael Belasco (22:11):
The barrier to entry typically is data and then the ability to extract valuable information from that data. A lot of these smaller shops couldn’t afford or just didn’t have access to data, right? I am thinking of a lot of third-party providers right now, their data is essentially public information. And their value add is that they’re able to go out and exhaustively extract all of this public information, aggregate it, and then package it up, and then sell it to real estate shops. Well, now because the data’s all public, maybe there’s a little finagling, so to speak, but at the push of a buttons a lot of these small shops could duplicate those efforts. So it goes to software as a service potentially going away which we had mentioned. These third-party data providers who are really just repackaging public information, they’re at risk. And then the opportunity opens up for these small shops to be able to then get a level playing field, you could say, to then have the ability to extract the value.
Spencer Burton (23:19):
I think if you are producing proprietary data, that data’s far more valuable than it was, and therefore it’s … And it’s easier now thanks to AI to accumulate proprietary data if you’re disciplined about it. This is to the argument that the smaller shop is going to win in this race because … Assuming they’re disciplined. If they can accumulate proprietary data they’re going to be able to use it better than the bigger firms can even if the bigger firms want to. Just the legacy challenges.
Michael Belasco (23:58):
I don’t know that-
Sam Carlson (24:01):
Consider not just shops having disadvantage but positions. You look at somebody that is hiring analysts, or whatever the case may be, at some point in time wouldn’t it make more sense to outsource all of your analyst work to one place that has maximized the potential of AI? And you just say, “Hey, instead of us hiring these people and trying to do this ourselves, they have nailed it, have a proficient system. We send all of our” … “Everything to that firm to analyze.” It comes back and it goes to the next level, whatever they pass back to us.” That is what the disruption could look like. Which by the way, anytime there’s disruption there’s opportunity. You have the firms that are just going to be … That have the flexibility and the speed to adopt these new things, but these positions are also … You can say, “Well, they’re vulnerable. My future as a career or professional and this thing is vulnerable.” Maybe this is the most opportunistic time to get in real estate that there ever has been.
Michael Belasco (25:17):
You know what’s interesting, you talking about this, is that there’s been … There’s a debate when a firm or a shop develops some amazing tool, there’s always a conversation about whether we offer this to the market or we use it as our own competitive advantage, right? And you mentioned it, right? You perfect the underwriting process, and you have this tool, and you can offer that to the market. In my mind when you said that I was like man, would you keep that in-house and make that a proprietary thing that gave you a competitive advantage? And when I thought about it’s like well, no, because everybody can now have access to potentially do the same thing. And if you crack the code early enough there is an opportunity out there to go and do that.
Spencer Burton (26:03):
I mean, I think one of the low-hanging fruit … So there’s this concept, Sam, of a lease abstract. I think you’re somewhat familiar with it. You got to lease, the lease is 100 pages or 20 pages, somewhere in between there. As a landlord or as a prospective landlord, you want to understand the summary of that lease. What’s the rent? What are the landlord’s responsibilities relative to the tenant responsibilities? Are there termination options? What’s the term of the lease? How has the rent changed over time, right? So there’s all these tidbits that you want to pull out of it. In fact, on the website we have a lease abstract. On Adventures in CRE, we give away for free in the same way, right? … It’s an Excel file.
Traditionally how it’s done is there is an individual who’s trained to abstract leases. Some firms use overseas talent to do that, to make the cost more effective. But in essence, someone reads that entire 60-page lease. Let’s say there’s 40 data points that you want to pull out of that lease, and they’ll look the lease and they’ll fill out that Excel file or whatever that may be.
I’ve been playing around with abstracting … Having one of these natural language models abstract at least, and I would argue it’s as good as your overseas talent. In many instances as good as your analyst or in many instances as good as your seasoned professional. And so your first pass of a lease abstract automated. And so the days of paying $300 for a company to abstract your lease for you I think are going away.
Sam Carlson (27:54):
Can I ask a question real quick? So let’s say you build that tool and it does exactly what you’re saying, I’m just thinking tactically, wouldn’t you then just say, “Okay, fill out this spreadsheet.” And then have a column next to it. “Also tell us where you got that information.”
Spencer Burton (28:10):
Sam Carlson (28:11):
So we can go back through validate the information. There’s probably going to start to occur in these contracts when you’re doing a deal, this lease has been evaluated by AI. If any of these assumptions are incorrect we reserve the right to revert or whatever it is. You know what I mean?
Michael Belasco (28:30):
Let me ask you, in that lease abstract process-
Spencer Burton (28:35):
And this isn’t hypothetical, by the way, where this is real.
Michael Belasco (28:37):
I’m sure you’ve gone through a couple iterations. Knowing you, you’ve got it down if you want to do it. I want to ask you, how long does it take you to do a lease abstract now with what?
Spencer Burton (28:51):
Personally or with the AI tool?
Michael Belasco (28:53):
With the AI tool. How long does it take now?
Spencer Burton (28:55):
Michael Belasco (28:56):
Spencer Burton (29:01):
Michael Belasco (29:01):
Spencer Burton (29:03):
And the reason it’s 30 seconds is because I have to upload it … Upload the lease file to ask your PDF and that takes 10, 15 seconds. And then I have to grab the doc ID and drop it into ChatGPT-
Michael Belasco (29:14):
And press your prompt.
Spencer Burton (29:15):
And then a prompt … Press my prompt.
Sam Carlson (29:16):
So, hold on. So there’s two personalities listening to this. One is like uh-oh-
Michael Belasco (29:21):
Sam Carlson (29:21):
And the other one’s like oh, amazing. I think we appeal to the sensibilities of both sides because they both have merit. But I think there’s an oh-oh and there’s a a-ha.
Michael Belasco (29:36):
Just exposing the reality, right, it’s true. And this is what we talked about in the last episode. There is a plurality of emotions, and feelings, and thoughts about what’s going on right now. But I wanted to run that … Bring that home. How long did it take you to do something that people made a career out of, right? Or at least a bulk of their career? Well, it takes them 30 seconds. I don’t know what that does. There’s other functions, other things that you can do.
Sam Carlson (30:04):
Evolution. Evolution. We’ve evolved to the next high value proposition and we’ve become better. We get more deals, better deals, whatever it is.
Michael Belasco (30:16):
But to your point, it’s hard to say to some person, “Hey, look, you got to evolve buddy, time to evolve.” You have to, it’s true, it’s just a hard … I don’t want to understate the emotional feeling that could be happening. It’s not permeating that much in commercial real estate yet as opposed to some other industries but it is here, and it’s coming, and-
Spencer Burton (30:40):
Oh, it’s happening.
Michael Belasco (30:41):
You should be preparing now.
Spencer Burton (30:42):
So, Sam, you described two people listening. I think there’s a third which is the yeah but person, right? Again, no offense to any of the three, I think all are natural reactions. The yeah but I have interacted with a fair number … A fair amount as … One thing that I’ve always been passionate about in my career is efficiency and improvement. I was part of a shop where we had, in my mind, a very inefficient process for developing materials for approval. And so I developed this tool that largely automated a component of that and there was the yeah but crowd. The yeah but had really two issues. The first is, well, how do you know it’s accurate? And so I think there’s a lot of people going yeah but what if it gets it wrong? The lease is so key.
And I think, Sam, you described one solution which is you got to validate it, right? There needs to be some validation. In the same way that if you have some six-month analyst briefing your lease, you probably want a more seasoned person to also validate it, right? The first is, is it accurate? The second is, yeah but how do you train the next generation? And this tool that I created in this other firm, one of the pushbacks I got was yeah but this is an important tool, this is an important process for our younger people to learn the industry.
Because they have to do this manual work they learn what a lease actually is. How do you respond to those people? They say, “Well, yeah but.” That means no one will ever read a lease. Don’t you want your people to read leases? How are they going to know what makes a good lease and a bad lease if they’re just turning it over to a machine to brief the lease for you? How do you respond to those people?
Michael Belasco (32:30):
Well, first I would ask the question, is that valid? The principle key that you want to hit with your response back is, are these people going to learn how to read a lease? I’m answering this question. If we found a more effective way to get the work done then that leaves time for a more efficient training on what a lease is to where these people don’t have to spend months reading leases, they can have a highly focused time period where they’re training about key things to look for in leases and that could take a week. You could develop a supplementary tool that talks about what exactly you want to check for. And these people could be reviewing.
Spencer Burton (33:12):
Or in other words, do you actually need a read a lease? I think the day’s coming where you have two AI attorneys negotiating the lease thing.
Michael Belasco (33:20):
Wow, we’re laughing but what if.
Spencer Burton (33:23):
The business people, what we care about are the business points. All the legal it’s like, we’ll turn that over to the attorneys. I mean, imagine the day. The first question you’re really asking … The premise is, oh, you need to actually read leases. Do you need to read leases? And I’m not suggesting you don’t I’m asking. I’m posing the question, is that skill actually important for a real estate investment?
Michael Belasco (33:45):
I think yes. I think leases are not templatized, for the most part. There’s legal creativity in leases. You mentioned all these bullet points of leases that you need to extract. What you’re really trying to understand for … Other than the nuts and bolts is what is the risk of this lease? If there’s a lot of creative things and you’re assuming this risk, it’s … In certain asset classes … Well, most asset classes, it’s everything, it’s your cash flow. So I personally believe you should. At this point, you need to read a lease. Now, is there AI out there that’s coming that can identify some issues better than humans? The answer’s possibly yes.
Spencer Burton (34:36):
Michael Belasco (34:37):
Probably yes. Possibly yes. I’ll leave it at that for right now.
Sam Carlson (34:44):
I’m not sure. We’re doing lots of loops here. But I just keep coming back to the copilot idea. The person that is wielding the sword of AI in any capacity is going to have to be … The people that are really going to thrive are the people who are just good at the fundamentals, know how to coach, and prompt, and work within the tools provided. If it’s reading a lease, evaluating a lease, they’re going to be able to do 100 of them instead of one. And that’s a value add for anybody. I don’t know why any business wouldn’t just flip out in excitement being able to do that. Our capacity now with the people we have we can evaluate X amount of deals per month. Now you can do X amount times 100.
Michael Belasco (35:42):
You know it’s interesting-
Sam Carlson (35:43):
Why wouldn’t you want to do that?
Michael Belasco (35:44):
Whenever there’s a major crisis at an organization somebody’s head rolls and that’s how you solve the problem, right? Well, if you implement and it’s automated everything like whose head rolls when there’s a massive breakdown or something? Maybe the person who implemented is the first to go.
Sam Carlson (36:06):
I’ll tell you-
Michael Belasco (36:06):
There needs to be accountability and responsibility is what I’m getting at, and that is what I think creates the biggest risk to the yeah but crowd in a institution like that is that there needs to be personal accountability all the way up the chain.
Sam Carlson (36:18):
Well, I’ll give you one example of how we’re using this in my software company, and then maybe let’s wrap this up. When you’re managing, optimizing marketing campaigns of things that a person, we call a media buyer, they will do this. They’ll go through, they’ll evaluate something that’s happening, and make some very predictable adjustments to that marketing campaign to keep it producing results. Jumping in and out of different marketing ad accounts can take time, evaluating it can take time. You’re talking about an hour per account on average.
We’re using AI right now to evaluate common issues. And then that same media buyer could look at the metrics and hit a quick action. And the quick action could be do this, do this, do that. Now the quick action that it makes available is going to be executed by a person, but instead of spending an hour they’re going to spend 30 seconds. That just meant that that person in one hour can do 120 times more actions than they could do. So now one media buyer can now handle, I don’t know, 500 accounts instead of 50.
Spencer Burton (37:37):
Let me give a real estate example now that I think is relevant. And let’s move now to the underwriting side, real estate financial modeling side of this. We teach that real estate financial modeling is … What you’re doing is forecasting the future, right? And so in the same way that a meteorologist will forecast whether it’s going to rain tonight, rain tomorrow, or rain the next day, what the temperature is going to be tonight, tomorrow, or the next day, what do they do? Well, they build a model and they have inputs. The quality of those inputs determine the quality of their outputs, and the quality of their outputs determine their credibility. And in real estate, the quality of your outputs determine your profitability and your profitability determines how much money you make both in a career and as a professional. Your inputs are key is the point.
There’s this exercise right now, right, where one of your key inputs is going to be, let’s say, cap rate, okay Reversion cap rate. And that reversion cap rate, how do you derive it? Well, it’s a function of today’s market cap rate. And how do you arrive at the market cap rate? You pull comps. How does a traditional shop pull comps today? They email the broker, they go to CoStar, they go to whatever their preferred data source is and they put a comp set together. And it’s a somewhat laborious task. By the way, this is just one input of, in some cases, hundreds of inputs that go into a model. And so there’s this laborious task of polling comps. That’s for fun, it was actually for real.
I got my personal property tax bill for 2023 and it was excessive, it went up 40%. It’s ridiculous in Colorado. Where in some states they cap how much it can increase every year, there was no such cap in Colorado. I had this big increase in my property tax and I’m like, I’m going to appeal this thing. Well, to appeal it what’s the process? Well, traditionally what I would do is I’d go out I’d pull some comps that would support the value that I think is a reasonable property tax assessment value for my property. Instead, I’m like, I’m going to try ChatGPT’s new … Some of their new plugins, and they have a plugin with Zillow, okay, right? And I know Zillow’s not relevant to commercial real estate, but Zillow can be a proxy for really any data provider.
And instead of the manual process of going through an MLS or going through Zillow or realtor.com and pulling comps and identifying which are the right comps, what I simply did is I went to the AI I said, “I’m trying” … “I need to appeal my property tax. I need comps for this address that sold in the last six months that support a value of X.” Or, if I didn’t have a value I could say that would … That are comparable to this property at this address.
In a matter of seconds I had a list, a longer list than I needed of comps. And then I asked it to organize those comps in a way that I could then present to the taxing authority and it did that. In a matter of minutes I had a property tax appeal prepared or the support that I needed. Now that’s just one input.
Sam Carlson (40:57):
We want to know if it got … What the bill is.
Spencer Burton (41:04):
They valued it at Y, I appealed and proposed a value of X, and they met me in the middle. And now I’m at this point where okay, do I fight more for X, or do I accept the middle? My time such it’s not that much more I’ll probably just accept it. But the point is I did save myself some money by going through that exercise. But as it relates to real estate-
Sam Carlson (41:23):
Wait, wait, hold on, this is an AI show. I mean, just have ChatGPT fight with these guys, go back and forth, it’s not your time.
Michael Belasco (41:29):
What do you mean your time?
Sam Carlson (41:30):
It’s ChatGPT’s time. Exactly.
Spencer Burton (41:32):
Well, that’s how the government gets you is the next step is I’d actually have to do an in-person appeal.
Sam Carlson (41:38):
Spencer Burton (41:38):
Where the first step is an online appeal. So it’s like okay, do I really want to go in? And I could, right? I could have ChatGPT or whatever AI model produce some material that I can then go in person. And my time’s probably not worth whatever the tax amount was.
Michael Belasco (41:54):
They knew their target market here and were like “We know how to take this to the next level.
Spencer Burton (41:59):
No, they’re smart they know what they’re doing. The point is, is it relates to anything, right? It’s coming, you need to get your set together. The first iteration of your comp set can be created or assembled by AI.
Sam Carlson (42:14):
Okay. Let’s wrap it up here. I got one final question I’m going to give to you.
Michael Belasco (42:19):
We have more examples.
Sam Carlson (42:20):
I’ll give it to you, Michael, and then you, Spencer to finish. Real estate’s a relationship business, yes or no?
Spencer Burton (42:27):
Michael Belasco (42:27):
Sam Carlson (42:28):
Okay. Does that change? If so, how? If not, talk to me about that idea.
Michael Belasco (42:35):
This is one area I think AI is exciting because it gives you the freedom to prioritize your relationships. And that goes from top to bottom, from in any position. Now, this could go in many ways. Again, we’re at the infancy of this AI revolution, but theoretically where this is heading is that there is, like you mentioned, a guy who’s reviewing his accounts could do 120 when he was doing three prior. You can 10X your productivity and then you have that extra ability to go out and network and foster those relationships more. So I think this opens up an amazing opportunity to be able to grow your network a lot more just because it allows you to automate a lot of the tasks that you couldn’t before.
Sam Carlson (43:30):
All right, Spencer, final word.
Spencer Burton (43:31):
Oh, 100% agree, that’s exactly right. To me, the relationship piece is one of the most important, if not the most important aspect of commercial real estate both in terms of you develop unique proprietary insights oftentimes by talking to people. You source great deals through talking to people.
Michael Belasco (43:50):
Spencer Burton (43:51):
You build trust talking to people. You find the right people for your team by getting out there. To unlock your time, to spend it building relationships, and/or thinking critically about investments, and not spending your time briefing leases or assembling comp sets I think just makes all of us better.
Sam Carlson (44:17):
Awesome. All right. That was a fun episode. I mean, the two AI episodes back to back really exciting. I think there’s a lot of cool things coming with AI. Thanks for watching, we’ll see you on the next episode.