Pod Insights The Real Estate Espresso Podcast Creating Systems In Your Real Estate Business with AI About this episode Today's show is an excerpt of a webinar held earlier this month on the application of AI in your real estate business. Enjoy today's segment on AI.
-------------
**Real Estate Espresso Podcast:**
Spotify: [The Real Estate Espresso Podcast](https://open.spotify.com/show/3GvtwRmTq4r3es8cbw8jW0?si=c75ea506a6694ef1)
iTunes: [The Real Estate Espresso Podcast](https://podcasts.apple.com/ca/podcast/the-real-estate-espresso-podcast/id1340482613)
Website: [www.victorjm.com http://www.victorjm.com/](http://www.victorjm.com http://www.victorjm.com/)
LinkedIn: [Victor Menasce](http://www.linkedin.com/in/vmenasce)
YouTube: [The Real Estate Espresso Podcast](http://www.youtube.com/@victorjmenasce6734)
Facebook: [www.facebook.com/realestateespresso](http://www.facebook.com/realestateespresso) http://www.facebook.com/realestateespresso%5D(http://www.facebook.com/realestateespresso)
Email: [[email protected] mailto:[email protected] ](mailto:[email protected] mailto:[email protected] )
**Y Street Capital:**
Website: [www.ystreetcapital.com http://www.ystreetcapital.com/](http://www.ystreetcapital.com http://www.ystreetcapital.com/)
Facebook: [www.facebook.com/YStreetCapital](https://www.facebook.com/YStreetCapital) http://www.facebook.com/YStreetCapital%5D(https://www.facebook.com/YStreetCapital)
Instagram: [@ystreetcapital](http://www.instagram.com/ystreetcapital)
Transcript 0:02 Speaker 00 Welcome to the Real Estate Espresso Podcast, your morning shot of what's new in the world of real estate investing. I'm your host Victor Manash. This is the weekend edition. We've got a special show for you today. Today's show is an excerpt on a webinar I held earlier this month on the use of AI in real estate. Enjoy today's segment on how you can apply AI in ways that maybe you haven't thought of.
0:24 Speaker 00 Well, good evening, everyone. We're here to talk about AI this evening, and I'm excited for this conversation because I've got a number of use cases that are, I think, going to be very important for you as real estate investors. Now, this is the second in our webinar series on using AI in the world of real estate. Last time, we talked about the importance of setting context for AI. AI tools are great at content, but they're dreadful at setting context. So unless you have the appropriate context set for your project, you're really going to get kind of random results. So we're going to dive into this tonight. We've got some great things coming up for you. Okay, let me share the screen here. And what we're going to start with, hopefully, the show's correctly. Here
1:05 Speaker 00 we go. So when we are talking to any language model, this one happens to be ChatGPT, what you see is a prompt that says, "Ask me a question." And when you see this, what you really have is a brand new employee on their first day on the job. They know absolutely nothing about you or about your business. And every time you open a chat window, this is what you've got. This is the image, the visual, that I want you to maintain at the forefront of your mind every time you ask a question. This is what you got. So it's not just about Google or OpenAI or Anthropic training the language models in terms of the knowledge of the universe. It's also about you training the AI tools about your business. That's part of setting context, and that's going to be most of what we're talking about tonight. And I'm going to show you some pretty cool things. But before
1:49 Speaker 00 we do, I'm going to show you something that's kind of simple, that is going to be a bit of a productivity hack for you. I know some of you have Android phones, some of you have iPhones, and we want to be productive on the move. But this image, of course, you cannot really do that when you're stuck in traffic. You've got to be hands-free. And let's face it, Siri is Well, maybe a little bit behind. So if that's all you got, then you're not really going to be very productive. Now, there is integration into CarPlay on the iPhone that is now become available such that both ChatGPT and Perplexity now have iPhone apps that work with CarPlay. So you can literally click on ChatGPT and you can put it in audio mode. And you can start asking it questions and being productive while you're driving.
2:33 Speaker 00 Maybe there is not a podcast or something or an audiobook that you want to listen to while you're driving. You can actually get to work by simply interacting with an AI tool. If you happen to have Google Workspace in your business, if your email is hosted by Gmail, then you have Gemini essentially bundled for free as part of Google Workspace. But Gemini is not yet available in CarPlay. However, you can kind of cheat and you can make it available as a widget. So if you search on how to set up a widget, you can set up all kinds of different widgets in CarPlay, like the weather and, you know, various other things. You can set this up so that you can interact with Gemini hands-free. It will recognize your voice and answer all your questions.
3:17 Speaker 00 The other thing you want to think about from a mobile perspective is are there things that you might have had going on on your desktop? And now you happen to have your mobile device with you away from your desk. Well, there are new capabilities available now also to help you with that. So if you happen to be using Claude within Claude within the mobile app, if you look partway down the left-hand side here, you'll see this thing called Dispatch. Dispatch connects your mobile device with the Claude Cowork app on your desktop computer. Whether it's a laptop, desktop computer, doesn't matter. And so you can literally start jobs that are running on your main computer from your mobile device. And you can manage those autonomously. In fact, part of the presentation for tonight, I started this a few days ago and happened to be driving, happened to have a long drive.
4:02 Speaker 00 And so this is one of the prompts that I actually put in to start some of the background work for today's presentation. So I'll show you that in a few moments. If your flavor is more ChatGPT and OpenAI's products, Codex is the desktop app and it has some things that it's better at, some things that it's Worse at compared with Claude cowork. But here too, you can access Codex from the ChatGPT mobile app. So you can start jobs running from your mobile device. So this is a way for you to be connected to the office, even when you're not. Maybe that's a, a, an anchor for some of you and you don't wanna be. That's fine. You can always turn it off. But if you wanna be connected, the capability is there.
4:44 Speaker 00 Okay, well, let's talk about one of the key problems with AI. And that is assessing accuracy. Accuracy matters in general, but it certainly matters in certain critical functions. Whether we're talking about legal, whether we're talking about healthcare or finance. For those of us in the real estate business, well, it kind of touches at least two out of those three. Legal and finance. So accuracy matters, and this is where context becomes important. You wanna have business systems that are nice and organized. And everything is as it should be. Everything is version controlled. And you're making sure that you're working with the right version of every document. And that is not always the reality. Sometimes this is what you got. Sometimes it's a plate of spaghetti and you have the wrong version.
5:26 Speaker 00 Within our company, we call that the disease of versionitis because you're working off of something that is not current. Maybe it's a different revision of a document. Someone sent you the wrong thing by email and you don't have a single source of truth. It's super important to have underlying control of your data structure in general. Because you'll screw it up if you're just using, you know, people to work on these documents. But if you now compound that with an AI tool that, quite frankly, doesn't know the difference. It will make mistakes with the same confidence that it does things correctly. And so then at that point, the effort to unwind that mistake is many times the effort. Then compared with if you had done the work correctly to begin with. Sometimes this is called the Prito principle.
6:09 Speaker 00 You really wanna minimize that management effort that goes into fixing those mistakes, problem solving. You wanna push as much as possible into things that are routine. So you wanna take those exception cases and push them to something that's routine. And this is where AI can help you tremendously. Now, there are some folks that say, well, I subscribe to ChatGPT. Someone else will say, I subscribe to Gemini and so on. And I don't personally have any religion around any of these. I use them all. The fact is that some of them are better than others. I'm going to give you an example of something that I tried. In fact, it was some of the research for today's podcast episode. Where in particular, I wanted to, I wanted to find out what the default rate was right now for loans that are underwritten by Fannie Mae for multifamily apartments. I thought that would be interesting.
6:54 Speaker 00 And maybe I could compare that to what was happening in two thousand eight, see if it's better or worse, the same. Is there any comparison? And quite frankly, Gemini couldn't handle the question. So I said, okay, let's try something else. At first, ChatGPT failed because I was not diligent enough in terms of how I asked the question. So then I asked it a better question and then it actually gave me some good data. So it's not just which tool you're using. You also still have to ask the right question. I then took it a step further and did a comparison work in Claude. And put together the numbers for both, the current situation as well as what it was back in two thousand eight, in the wake of two thousand eight. And when I was satisfied that I had pretty good data, I then asked Claude to write me a script for the Real Estate Espresso Podcast, comparing
7:39 Speaker 00 the current Fannie Mae default rates with what they were in the post two thousand eight period. And it did that. And you can listen to it for yourself and decide whether you think it did a good job. It really did it only in a couple of minutes. So as opposed to me spending twenty to thirty minutes writing one thousand words, it wrote a script. And part of the reason why it did a good job of that is because I have seeded these tools with this thing, this language model that we're affectionately calling Victor AI. Where we've uploaded the transcripts of gosh, close to three thousand episodes of the podcast. So it kind of knows a little bit about me. It knows how I speak, my cadence, the types of words that I use. I don't speak in horribly long, complex sentences.
8:21 Speaker 00 I try and speak in a very straightforward manner. So it has managed to capture my voice and it's capturing it in the script. So that's something for you homework, if you wish to see what you think. And the comparison between the conditions today and two thousand eight are in fact quite striking. So I think you'll enjoy that episode. 3062 today. Okay, we talked a little bit about that brand new employee first day on the job. And when you do something in any language tool, it's basically going to figure out how to solve your question, how to answer your question before it answers the question. And when it does, it's kind of making it up as it goes along. It will try this, and if that doesn't work, it'll try that.
9:02 Speaker 00 Once you have figured out, once you have trained your favorite tool, Claude OpenAI, doesn't matter whichever one you like to use, you want to take that knowledge, that formula, if you will, and capture it as what's called a skill. You're going to tell the tool, okay, so what you just did, capture that and do it the same way every time from now on. And what that does, number one, it saves you a whole lot of money that you would otherwise spend, relearning every single time this gets executed. But not only that, it gives you a much more predictable result. So in Claude, it doesn't matter, we can do this in Claude, we can do it in Codex. There are these things called skills. And so I'm going to describe a little bit of what about what they are. It's basically a set of instructions and it's just a standard tax document.
9:46 Speaker 00 And there, today, there are massive libraries of prepackaged skills that are out there. You can download them. There are literally thousands of them that you can download. Many of them are useful out of the box. So for example, if you want to get a morning briefing, you just go into cowork and you type slash morning briefing, and it'll look at your calendar, it'll look at emails that have come in in the last 24 hours, tell you what you need to respond to. If there are certain topics that are important to you in the news, it'll summarize those. And you will get that as an email first thing every morning as part of your morning briefing. There's many of them. In fact, Anthropic, the makers of Claude, have packaged bundles of skills for different types of businesses.
10:27 Speaker 00 So for example, there is a finance group of skills that will do everything from bank reconciliations and generate reports out of QuickBooks and all kinds of different things, create business dashboards, all within that finance package. There's many of them. There's a productivity package, there's a design package. You can explore those. You'll find literally hundreds and hundreds of them. Some of them can be triggered, for example, based On time of day, or so that's kind of like the alarm clock trigger. Or sometimes it can be triggered based on an event, kind of more like a doorbell. When the doorbell rings, then something happens. So these are all things that can make a part of the skills universe. So how do they work? When a user types a message, then basically Claude detects the intent of that and then figures out how to
11:11 Speaker 00 do this in a repeatable way. And so it will do it for the first time and it'll probably ask you a bunch of questions along the way to make sure that it gets it right. And then when it's done, it'll capture it as a skills file. And then, of course, if you are not happy with it, if there's something that wasn't quite right, you say, I need you to modify the skill. This wasn't quite right, go fix this. And then update the corresponding skill file and it will do that. So as I mentioned, there's all kinds of different skills libraries that are available. They're available also in ChatGPT, in the Codex application. And so I encourage you to get to know these as well. I'm going to show you something that was really born out of necessity. This was the result of a conversation that I had with someone in our finance team.
11:54 Speaker 00 And we were late paying property taxes on one of our properties. It's not because we didn't get the bill, we did, and it got forwarded to the right place. But anyway, it just got missed and the conversation went something like this. It was like, Victor, property taxes are the bane of my existence. You know, they're all on different timetables. Sometimes it's once a year, sometimes it's payment in advance, sometimes payment in arrears, sometimes it's three times a year. Like they're all different, different deadlines for the tax assessment. It's just a lot to keep track of. So I said, hmm, I wonder if this is something that could be easily solved. So I went into my, all of my favorite AI tools and asked this question.
12:33 Speaker 00 I said, I own a number of properties across several states and communities, and they all have different schedules and due dates for property taxes. If I provide you a list of these properties in an Excel spreadsheet with the addresses, could you in fact create a skill to add the property ID to the spreadsheet? And then research the tax assessment and create a global dashboard for the whole company with all of the due dates for everything in the portfolio? And then ask me some questions. Questions if you know what you need to do. So I gave it a list of addresses. This was my test case. I gave it five addresses from different properties. This is not our, nowhere near the entirety of our portfolio,
13:13 Speaker 00 but just a few sample addresses. And then it asked me a bunch of clarifying questions, so I answered those. These were my answers. I said, I expect about 20 properties. The spreadsheet will have the address and the city. They're owned by different entities. It's in different states. I want all of the relevant data, including the assessed value of the property taxes, the installment dates, and the installment amounts. Put it in an Excel spreadsheet for the whole year. And then you can use browser automation, that's important, which we'll get to in a moment, to scrape the county assessor websites. So I first tried this with Claude, and sorry for the eye chart here. But what happened is Claude got stuck. It figured out how to read the Excel spreadsheet.
13:57 Speaker 00 But when it came to going to the county assessor websites and the county tax collector websites, all of these websites recognized that this was not a human trying to interact with them. It was in fact a machine. And it said, nope, we're closing the door to you. You cannot get in. So that can be a problem. So I said, all right, well, if Claude can't do it. And it, by the way, it did give me another solution, but it was complex and it required writing a script and opening a command line. And I'm technically capable enough to do that. But I said, you know what? I'm feeling lazy tonight. Let me try a different tool and see if it can figure it out. So I abandoned Claude temporarily and decided to try Codex, and this was the exact prompt that I gave it. I said, you know, if I provide you a list of property addresses in an Excel file,
14:40 Speaker 00 can you generate a dashboard containing all the properties, one row per property, showing the property taxes owing for that particular property, along with the installment due dates and the amounts, and then, you know, output it into an Excel file for those jurisdictions having multiple taxing authorities, like for example in Texas, you've got a municipal utilities district. You've got the county and the school district. If you need to, go and add more columns for each taxing authority. And so then it Went ahead and it did it. It did a pretty good job. It didn't do a perfect job. I'll show you what it did. So you can see with these five properties that it got, the first two bang on, the ones in Philadelphia with no problem, the ones in Louisiana, it struggled with.
15:23 Speaker 00 The one on the bottom in blue, I had to give it a little nudge, give it a little help. It couldn't figure out that the name of the street in this particular case, the address was number 400 and the road is WM Perkins Road. It said Perkins. It got rid of the WM and then it got stuck. So it didn't know what to do that. So I gave it a little nudge. But it got stuck on the ones in Louisiana. The one on Weaver Road, it got stuck on. I didn't bother going in and fixing it. But you can see that it got most of it. The little tweak that would be necessary to get this to figure out how to do this reliably across the entire portfolio is pretty small. And so then at this point, I have an entire dashboard for the whole portfolio.
16:03 Speaker 00 I can obviously give it a longer list of properties and it would be able to produce that dashboard for the entire portfolio. And I think that's actually valuable. It's part of that picture at the very beginning where you've got an organized filing system. I mean, think about doing it not just for property taxes. Think about insurance policies, your GL policy, builder's risk, flood insurance, all different dates, different renewals, all of that stuff. You would really want to have a single dashboard for your entire organization so that your financial team has a single place. All of the annual reports for your corporations, those filings in Delaware and Wyoming, and all those different states, make sure that you paid your whatever annual 200 bucks and you don't miss that.
16:45 Speaker 00 And now all of a sudden, your company is no longer in good standing. You want to do the same thing for all of those. These are all part of getting organized and frankly, these AI tools can do this. It can make light work out of all of that and really help get your business organized. So then after I did this, I then said, okay, take what you've learned and write a generalized skill. So then it took that and says, okay, I've created a skills file, which is now reusable. And now we have a formula that's going to be useful for every single time that you do this. And this is just an example. There's many of the ways that you can use skills to create power in your organization. Do things on an A's needed basis where you invoke the skill manually. You can base it on a particular set of conditions. You might run a skill weekly.
17:28 Speaker 00 For example, I'm using skills right now in my wife's business where she has close to 20 staff working for her. And we're making sure that all of their systems are in fact being run. And so we're using it as a QA tool. We're using it for financial reconciliation and that's something that runs nightly. And it's something that runs every two weeks with the bi-weekly payroll. To make sure that all of the timesheets are in and all of that kind of stuff. The clients that were supposed to be billed. If they haven't been billed, why not? There's follow-up. All of those sorts of things are now getting automated. At least the intervention is getting automated. Where before it required manual exception handling by management, which of course is like 10 times the effort compared with the success case. So those are some of the thoughts that I wanted to share with you tonight.
18:14 Speaker 00 There's many more examples that I could show.
Previous Navigating Today's Market with Buck Joffrey Next Lessons From The FIFA World Cup Page updated November 14, 2026 · transcript and summary are machine-generated