Pod Insights The Real Estate Espresso Podcast AI Mastermind in Sonoma About this episode Today's show is a live talk that I gave a few weeks ago in Sonoma on how to use AI to gain leverage in your real estate investing business. I know we are talking a lot about AI on the podcast lately. We're doing so because it is so impactful on so many aspects of the business. Enjoy....-------------**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) 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) Email: [[email protected] ](mailto:[email protected] ) **Y Street Capital:** Website: [www.ystreetcapital.com](http://www.ystreetcapital.com) Facebook: [www.facebook.com/YStreetCapital](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 very special show for you today. Today's show is a live recording from a talk that I gave at a mastermind in Sonoma, California. It's a talk on the very practical use of artificial intelligence tools from the perspective of a practitioner who's using these tools on a daily basis. Enjoy this talk on how to use AI.
0:28 Speaker 00 Well, welcome. My goal here this morning, let me start with the notion of something called bonus time. Do you know what that is? Bonus time. Bonus time is what happens when you say, you know what, from at this moment, I got my money's worth out of today and everything that comes after that is a bonus. Okay, so my goal is to have somebody shout out bonus time before the end of my talk. So the rest of the day is gravy. Okay, AI has transformed the world already for those who are engaged in it. Not everybody is. And I want to impress not, I don't want to
1:08 Speaker 00 impress you, I want to impress upon you how accessible it is and the transformation that is possible for you. And I'm going to show you some real world, like real life examples. So this isn't a theoretical thing. It's from the perspective of a practitioner that's literally doing this on a daily basis. So we'll start with one that's really simple. You know that when you get on the airlines, you get they have the wifi on board and you normally pay for it, but they give you free text messaging. And that's useful sometimes, you know, message my wife and say, okay, you know, should be landing soon and all that kind of stuff. But if I really want to interact with something, if I want to do some research, and my hands are kind of tied unless I pay the eight
1:51 Speaker 00 or $15 or whatever the number is for the hour remaining in the flight, and maybe I don't want to do that. Well, turns out that you can actually get access to AI through a messaging app. So you can, because it tells you. You can message using iMessage or Facebook Messenger or WhatsApp. All of that works. So I'm going to show you WhatsApp. It's here on the computer, but you can use WhatsApp on your phone and start interacting with it. So I'm just going to type a simple query here. It says I'm visiting. Helps if I type correctly.
2:30 Speaker 00 I'm visiting Sonoma and I'd like to tour some wineries. Prepare an itinerary of wineries that specialize in biodynamic only. Means they don't use pesticides. See what it does. I have four hours. So I can do this from my seat on the plane and it'll put together a nice little plan here. Okay, so it says, "Visit the Benziger Family Winery." Then go visit the St. Francis Winery, then Donham Estates.
3:14 Speaker 00 And you now get a pretty decent plan and it tells you what to do in four hours. And now that's free where previously you would have paid, had to pay to gain Internet access. Now you can actually do all of that chat stuff that you would have done through a web interface, through a messaging interface, and it's kind of easy. So that's one little thing that might be of value. Okay, that's number one. Number two, let's look at a very practical thing from the perspective of someone who is looking to do things in the world of real estate. I'm going to start with, I'll come back to this example. I'm going to actually open a new tab here and do a new example.
3:59 Speaker 00 So put up your hand if you have, if you're using Google, Google Workspace. So is your email hosted by Google? So you're paying for, you're paying for all of these tools. They come free for, well, free within your Or whatever, ten, twelve dollar monthly subscription for that service. So there is Gemini, which is, which is a, a, a large language model chat like tool. It competes directly with, with ChatGPT. It's pretty comparable these days, especially Gemini three. And then if you scroll further down, and it's accessible through that tic tac toe board where you get access to all of the Google tools, like the calendar and, and YouTube and everything.
4:40 Speaker 00 It's all there. You keep scrolling further down and you get to this thing called Notebook LM. And Notebook LM is, is an AI tool, but it allows you to work directly with files on your computer. So I'm going to do a little live example here. I'm going to upload a file where this particular file happens to be a draft contract. It was very poorly written. Like it was literally a piece of junk, but I'm not going to embarrass the folks who don't pay too close attention. And so it up now, now uploads the document. It gives me a little summary of the document and I'm going to now give it instruction. What do I want to do with it?
5:22 Speaker 00 So I'm going to say, please find, now please is redundant. You're consuming extra resources, being polite, but, but I'm Canadian, so we'll, so we'll be polite and spend the extra micro cent on that. Exactly. Exactly. Please find the in, helps if I spell it correctly, inconsistencies in this draft agreement.
5:53 Speaker 00 And so then it will go into the document and it'll produce a report saying, well, this, you know, this paragraph conflicts with this other paragraph. Take a few seconds. We're, we're almost there. And, and that's really a first line of review before you hand it off to a lawyer, even before you read it. It, I kind of think of it like the difference between a bicycle and an e-bike. You're still turning the crank and you're, you know, there's still a little bit of effort involved, but that's okay. It's just accelerating the process. It doesn't eliminate the need for you to turn the handlebars or do all the other things that you would otherwise need to do, but it is just giving you that extra little bit of assist.
6:36 Speaker 00 So first thing I'd identified here is that there's reference to some other agreement, but then there's a clause that says, well, this is the agreement in its entirety. So those other agreements should have been actually referenced and attached as exhibits. So it's found that right away as something that needed to be fixed. And then it goes on through it, owner liability. There's a whole bunch of these. This is this particular document's junk. So it was very easy for it to find things, but even any form of legal review, it's very useful. For example, if you have a memorandum of understanding and then you send that off to the lawyers to paper up, then the very first thing I would do is I would take the MOU,
7:17 Speaker 00 I would take the agreement, and I would put them into Notebook LM and say, well, did the lawyer actually implement what the MOU said? Or are there some gaps? That's a very easy thing for it to do, and it'll tell you that. So there's some significant time savers that are possible. Another one, I'm going to show you, this is an example, literally from this weekend. Someone in my company put out an email, which I wasn't happy with. I gave them a hard time, and I got a lot of defensive pushback from them. So I said, you know what? What we're going to do is we're going to take some of the material from our securities lawyer, because he's got a fabulous YouTube page, and I'm going to take the transcript
7:58 Speaker 00 from some of his videos, and I'm going to turn that into a PDF file. I'll strip out the timestamps, and I'll turn that into a PDF file and upload that into Notebook LM and say, these are now guidelines for SEC compliance specifically around how you communicate openly in the public with 506C exempt market offerings. Okay? And then I'm going to take the email. I took two of these videos, and then he had another document. So I had three, total of three documents that I uploaded as, I'll say, the rulebook. And then I had a fourth document, which was the original email that was sent out that I wasn't happy with.
8:37 Speaker 00 And I said, evaluate this email against Mauricio's guidelines and tell me how the email did. And so it did that. Scroll down here, and I apologize, this is a bit of an eye chart, but... For example, it said, you know, didn't like the fact that this was a little bit too promotional. It didn't like the fact that it actually didn't have a disclaimer. There were no disclosures, like there weren't any of the mandatory disclosures that would be required. All of these things, it picked up on right away and it had an interesting effect. Number one, it took that defensiveness away because it wasn't just Victor saying it. I was referencing the work of our securities lawyer and simply comparing the work of our securities lawyer with the work that was done internally.
9:21 Speaker 00 But it's even more powerful than that because now I have a tool where anybody in the company that's doing anything customer-facing can use this as a QA tool. They don't have to come to Victor and have Victor review the email. They can perform the QA themselves by uploading the file into Notebook LM and saying, "Does this comply or not?" So it's now a tool that can be used universally throughout the organization. And it's quite powerful because it takes me out of that loop. Does that make sense? Okay, so that's kind of very practical nuts and bolts. Notebook LM is super powerful. There's ways to use it that I'm continually discovering and rediscovering almost every day. Let's go next. Yes, good question.
10:06 Speaker 00 I don't know yet. I'll find out. Probably, I'm sure there's a way to do it. Now, we may automate this even further. We may actually not do it in Notebook LM for production. We may do it in Claude. There's reasons for both. The nice thing about, because our company's hosted on Google Workspace, everyone has Notebook LM already included in their email subscription. So there's no excuse, right? It's right there. Okay, just click on the Tic Tac Toe board and you're there, right? So there's no excuse for you not having access to all. I didn't have the $20 a month subscription or none of that. Like, forget it. You have access. So that's probably the number one reason why we're going to stick with Notebook LM
10:49 Speaker 00 because not everybody in the organization has access to the Anthropic tools, which might be a little bit more advanced right now. But this is good enough for production.
11:00 Speaker 00 Okay, yeah. Yeah, that's a great question. If I did a generic search, probably not as good. Because the rules are the rules, but then what I want is the interpretation of a lawyer, a practitioner, who's been in this business a long time, the one who's guiding us on a daily basis, saying, this is what you can do, this is what you should stay away from. He is kind of the voice of authority, as opposed to something that is a homogenized AI thing that I really don't know what's behind it. It's opaque, I can't tell, right? So that's probably the number one reason why I would do it this way, is because I know the source.
11:40 Speaker 00 The valuable thing in AI, and I think a lot of people don't get this yet, is not the large language model, the fact that you can interact with a chat tool. That is not the value. The value's in the data, and that's an important distinction. Because AI tools are very good at content, they're very bad at context. So I gave an example last night when we were hanging out in the kitchen. I did a podcast, this is about eight or nine months ago, where I talked about, in that podcast episode, the dramatic rise in inventory in Cape Coral, Florida. And I handed it off to my team, who said, okay, I'm gonna turn this into a blog article. So they put that into an AI tool to take the transcript of the podcast, turn it into a blog article. And when it did that, it changed a few things.
12:25 Speaker 00 It took the massive increase in inventory in Cape Coral and generalized it across the entire United States. Not exactly the same thing, because it couldn't figure out how to maintain the context. And so that is one of the most important things, make sure that you've got the right context. This past year, we implemented something that we're affectionately calling Victor AI. And this is our own custom large language model, I'll show it to you here. And I'm gonna start, it's built on ChatGPT, so I'm gonna start a new search, eventually here, open sidebar, here we go, new chat. So when you're in ChatGPT, you get the particular language model, in this case ChatGPT 5.2. But if you look further down here on the left-hand side, I have this thing called Victor AI.
13:10 Speaker 00 And so Victor AI is our own custom language model that has been seeded with the close to 3,000 episodes of the Real Estate Espresso Podcast, plus the text of my book. It knows not only the universe of knowledge out there, but it also knows me through the lens of the podcast, which is still a narrow slice of me. It's not me in my entirety, but at least it knows about me to that extent. And so I can do things like, for example, write a 300-word article on, I don't know, riparian water rights and the voice of Victor Minash. And let's see what it does. I have no idea.
13:54 Speaker 00 So it will, it'll go off and it'll do that. But one of the things that it will do is it will, it captures the way that I speak and the way, because it's been listening to my podcast and it knows it intimately. Whereas I find that if I was to do the exact same thing, in fact, let's do, let's do that. Let's go into another tool and let's say, write a 300-word article on riparian water rights. I have perplexity open here. Let's do, let's ask perplexity to write a 300-word article and see what it does. So it No, no. No, this is this is open, just doing an open web search, completely generic, completely unrelated to me whatsoever.
14:38 Speaker 00 It provides, you know, the, okay, so it has its basis in British common law, including Eastern United States, Canada, Australia, New Zealand. So it gives a fairly generic thing. If we go back now to, to Victor AI, it's recognizing that I'm coming at this from the perspective of a real estate investor. Let me zoom in on that a little bit so you can actually read it. It, it's already taken what might have been an encyclopedia description that you would have got from perplexity and it's now slanting it from the perspective of a real estate investor because, because it knows I'm a real estate investor. So it's, it's already tailoring the answer to the things that would interest me. So it's
15:19 Speaker 00 Much more focused in its response. Now, I only asked for three hundred words. If I asked for a thousand words, I would have got even more. But it's it's given me something that is already focused on on on my level of interest. And then, of course, it would have written it in a very straightforward like in a very straightforward manner, which you often find with a lot of the generic AI tools is they tend to use a lot of hyperbolic language. Because it where is it getting its information? It's getting it from press releases and things like that, which tend to use a lot more flowery language. So it tends to imitate what it sees in its environment. So you tend to see things that are overly embellished
16:01 Speaker 00 when it comes to stuff that comes straight out of chat GPT or Gemini out of the box. And and that's one of the tells that you have that you're actually getting something from AI is it attaches more significance to things that are really quite insignificant. You'll see language like this was so impactful and it's some ridiculous little thing. So it tends to over embellish things, whereas when I ask it to write in my voice, it doesn't do any of that. I'll give you another example of that. So I did an example. I I wanted to try this as an experiment and I did this a couple of weeks ago
16:41 Speaker 00 where I actually used vector AI to write the script and and I basically said, write an eight hundred and fifty word script for this particular podcast episode and do it in my voice and all of that. And and so it did. What came back was pretty good. I might have changed a couple of words here and there. When I'm producing the podcast, I actually I script it. I write the script for the most part. Sometimes I'll get AI to help with a small piece. What I do when I'm recording is I'm actually reading like as if I was using a teleprompter and I'm recording into the microphone. The reason I do that is because when I'm recording, I can focus my attention on
17:23 Speaker 00 intonation, cadence, things like that. So I get a better quality recording.
17:29 Speaker 00 As opposed to me in a dark room, oh shoot, what am I going to say next? If in front of a live audience, it's very easy for me to talk. But if I'm by myself in a dark room, I get stuck. So I, I found just through experience, I do a better job if I have a script. That's how I produce the show. So I got the script. I said, well, let's see if I, there's this text to voice technology out there. And let me see what it can do. So I took that script and I uploaded it into a tool called 11 Labs. I'm going to play you an excerpt from that.
18:06 Speaker 00 Welcome to the Villarreal State Podcast. Your morning show, it's new to the Villarreal State dusting. I'm your host, Victor Manash. Mainstream headlines are saying home sales have slowed. And on the surface, that's true. January's existing home sales number came in down 8.4% from December to a seasonally adjusted annual rate of about 3.91 million units, according to the National Association of Realtors. Okay. Now, I cheated a little bit. The first five seconds, I actually ripped off from another recording. That was actually me. After that, that was text to speech produced by 11 Labs. You could not tell the difference. You could not tell the difference. So what I did is I took, I don't know, half a dozen podcast episodes, put them all together in one audio file,
18:51 Speaker 00 uploaded about 30 minutes of me talking to 11 Labs, and that trained it on how to talk like me. Okay? I also, towards the end, fessed up and said, by the way, this is an experiment, that's not me talking. Send me an email if you could tell.
19:10 Speaker 00 I got, uh, one person said, yeah, I could tell it was, I could tell it wasn't you. It sounded a little bit too mechanical. Okay, fine. But then I got a bunch more of those. One just even came in yesterday and said, wow, if you hadn't told me, I would have not known at all. I would have not known at all. Took it one step further. This is kind of cool as well. And it said, you know, wouldn't it be cool if, you know, because there's a very large Spanish speaking population in the United States, wouldn't it be cool to create a Spanish version of the podcast? So I took that same episode, translated the script into Spanish, and took it back into 11 Labs and said, okay, see what you can do. So this is 11 Labs with Victor talking in Spanish.
19:59 Speaker 00 Bienvenidos a Podcast Real Expresso, tu voz es de nuevo en una versión mobiliaria. Soy Tofino Rittorino. Los titulares de los medios generalistas dicen las ventas de viviendas sean desaseguradas. Y en apariencia eso es cierto. I sent this actually to a few native Spanish speaking friends just to get their feedback. And and what they said was, well, one of them said, well, holy cow Victor, I knew that you spoke a bit of Spanish, but man, you're rocking it. That was awesome. Right. And then another native Spanish person said, yeah, there were a few phrases that weren't quite right, which meant not the issue wasn't the text to speech.
20:40 Speaker 00 That the issue was the quality of the translation from English to Spanish had a few gaps. So that was something to work on is to make sure that the translation was done correctly. Right. So this is, this is, this is what's here now. Now, that's not the most impressive stuff, actually. All right. So I'm going to take you something that's even more impressive. Anthropic has a bunch of tools under the brand name Claude. Claude is a, a language model like, like ChatGPT or, or Gemini. And then there's this other thing called Claude Code. And what Claude Code is, is a bunch of tools that actually make it easy to write software.
21:25 Speaker 00 Because writing software is actually, it follows a set of rules and it's actually, you know, easy to, easy to imitate. And many software development companies have in fact started writing code by simply directing the tools to write the code. So the key is to provide good instructions. This tool, and, and I want you to, you know, we are as humans, we're used to being linear thinkers. We're, we're not very good with things that accelerate. We have a hard time with that. And this is an example of acceleration. So within This particular tool is called Claude Co-Work. And you can see on the top, there are three buttons.
22:10 Speaker 00 There's Chat, and I have the chat window open. There's Claude Code, and I'm gonna click on that. And this opens up Claude Code. So if I want to access those tools. And then there's this button in the middle called Co-Work. And Co-Work puts a graphical user interface that combines the Chat and Co-Work and access to your local computer. That's not even the most impressive part. The most impressive part is that this got developed over the Christmas holidays in 10 days, from concept to shipping in volume. And the reason it came to be is because they found out that some of the employees at Anthropic were using Claude Code
22:55 Speaker 00 to do things like scan a directory full of receipts and put that into an Excel file to submit their expense report so they get reimbursed on their travel. And folks were just doing this organically. And they said, "Well, wouldn't it be cool if we could make that a little bit more accessible, a little easier to use, and put a graphical user interface so you don't have to do this from the command line?" They said, "Let's try it." Now, bear in mind, in most other companies, you could not get the marketing focus group together to even talk about the feature set in 10 days. They went from concept to shipping code live in the field in 10 days. That's mind-blowing. That's absolutely mind-blowing.
23:38 Speaker 00 So let me show you a few things that this can do. Well, actually, I'm gonna use some ones that I've done in the past year. So it can go into your computer and you can tell it... So I just told it here, "Organize my desktop files." And then it said, "Okay, you've used up all your resources. Come back at 11 o'clock." "Okay, fine." So then I came back and said, "Try again." Then it said... And then it tells me what it's gonna do. It's gonna organize the files. Because in this particular moment, my desktop folder was a bit of a mess. So it said, "Here are the folders I'm gonna create. Would you like me to organize it according to that?" And then it just does it. And then it tells me, you know, how many, you know, how many files to put in each folder
24:22 Speaker 00 and all of that. So it does it. That's not the most impressive part. I said, you know what? I haven't written an application for the Mac ever before. Let's give it a try. Let's start with something simple. Could you write me a game of tic-tac-toe? And the first thing I did is, is, is I said, just tell me what you're gonna do. So then, and then it, you know, gave me some choices. It's, I think I did that in the chat window, but it doesn't matter. It said, do you want me to use, and I said, use open source tools because I don't want to have to go buy any other software to do this. It said, use Xcode because that's Apple's free software development environment. Use Swift UI, which is their libraries. Have you installed Xcode? So I went and installed Xcode and I said, yes,
25:07 Speaker 00 I've installed Xcode. And then it said, you know, if you do this through the command line interface, it's free. I also asked it, tell me how much is this gonna cost in term, like how many computing resources am I gonna use? It said, well, it's gonna take between one and two hours and it'll cost you somewhere between $2 and $5. And I said, okay, fine. Go ahead. In actual fact, it took way less than that. I'm gonna say it took under 10 minutes. And, and then when I tried to run it, it, you know, told me what the structure was gonna be and all of that. So then I started to run it, started to run it and I said, I'm getting three error messages in Xcode. It's not working. So then it said, I can help
25:48 Speaker 00 you fix those. Why don't you upload the error messages? Maybe take a screenshot and tell me what's happening. I'll see if I can figure it out. So then I typed in, here are the three error messages. And then it immediately said, I can see the issues. Let me fix all three. And it did that in under 10 seconds. I now have a working game of tic-tac-toe on my computer that I wrote. I mean, I didn't write it. Claude wrote it. But I have my own custom version of tic-tac-toe. And I gave it some unique specifications. I said, I'd like the banner on the top to be in blue. So I have a tic-tac-toe game with a blue banner. Okay?
26:29 Speaker 00 Now, it's a trivial example, but what it tells you is that the cost of writing software is vanishing to zero. It's vanishing to zero. Even if you don't want to develop software yourself, you might hear this term called vibe coding. That's what this is. So you go into a tool like Claude, like Claude Co-Work or Claude Code, or there's a new one that OpenAI has just introduced in the past week that a lot of people are messing around with. There's a bunch of them out there, but people are out there writing software using this technique. They're saying, "Write me a program that does this." And it'll interact with you. "Well, tell me more. What do you actually want it to do? What are the specifications?"
27:13 Speaker 00 There are, I'm going to take you to some plugins, and I want to be mindful of time here. I don't want to go too long. Let me go to, here's my pro plan. I'm going to go to Settings. Here we go, Settings. There are connectors. I can connect Claude Co-Work to all kinds of different things. So for example, I have already connected it to Gmail. I have already connected it to Google Calendar. I have already connected it to Google Chrome. So there is in Google Chrome. If I switch to that, you'll see that I have a little icon, that little star icon in the in the upper right-hand corner. That is the Claude plugin.
27:56 Speaker 00 There's an extension now to Google Chrome. So I can access all those tools directly natively within Google Chrome, and I can get Claude Co-Work to do things for me on the Internet, on the web, from the command line interface in Co-Work. There's a plugin. Then you go into something called Skills. So let's go, or let's go to Extensions here. You'll see that I have three extensions already installed. I have an extension to Excel, an extension to PowerPoint, and an extension to PDF tools. So I can take a whole bunch of numbers and say, "Stick it in Excel for me," and it will do it, and it'll produce a native ...cell file.
28:39 Speaker 00 It's very, very powerful. Very powerful. I did another experiment this past week. Here we go. I actually haven't pressed the button on it. If we have time, maybe I'll do it here. My wife runs a clinical family therapy practice. Her, her website is familytherapy.ca. And then she has this other website that we've completely forgotten about. It's her personal website. Her name is Natasha and her website is natasha.com. And I looked at natasha.com and said, "Oh my God, this is like years out of date. This is junk." I then said, "I would like to create a new website for natasha.com. The existing website's out of date. Identify the sources you would use to design the new website."
29:22 Speaker 00 So then, this is what I came back with. Great, I've got a clear picture of who she is, her practice. Here are the sources that I would use to recreate the website. And then it asked me a few more questions. I don't know if it might be risky to do this live, but I could ask it to design a new website for me. And I'm gonna bet that in prob, in under five minutes, I'll, I'll do it during the break, but in under five minutes, it's gonna produce a new website. So, I'll pause it here and then let's just, you know, kinda have some, some discussion. And then after the break, I'll come back and we'll see, see what it does. That's a, that's a great question.
30:01 Speaker 00 In fact, I was at a friend's house on Sunday and his daughter is literally applying into computer science program. She has three acceptances so far. And so, over lunch, I was asking the question. I said, "Well, what's your favorite programming language?" She said, "I think it's Python." "Okay, cool. What do you intend to do with your CS degree?" I said, "Why did you choose computer science?" She wanted to follow in her dad's footsteps. "Okay, fine." You know, he was a senior executive at Amazon in their AWS division and was at Google. He's done lots of very cool stuff. And so, she wanted to follow in her dad's footsteps. Now, I didn't get into the conversation that said, "You know, it's not what you think."
30:47 Speaker 00 But yeah, it's a great question. I think the way it changes is you are no longer a coder as a computer scientist, but you definitely need to understand you become effectively a team leader. The latest incarnation, by the way, in Claude Co-Work is with Opus 4.6 is it can now launch up to 16 autonomous agents in parallel. So here I was just doing one thing at a time, but I could actually get Co-Work to fan out across 16 agents simultaneously, and I could even designate one of those agents to be the overseer of all of those other activities, to coordinate those guys' work. There's a school of thought right now that
31:32 Speaker 00 says the way to work with these AI tools. I was watching a podcast earlier this week where the individual said, "I have organized the directory structure on my computer according to different departments in my mind. So this is finance, this is quality assurance, this, like, all these different departments, and I now have agents working in, effectively digital employees, working in each of these departments, and my job is to manage my digital employees." That is the skill set. So the skill set is not a technical skill set, the same way that you would develop an employee in each of those departments. That management
32:16 Speaker 00 skill set is the same management skill set that you need to be an effective user of AI tools.
32:21 Speaker 01 I've heard if you don't know anything about coding, you can still...
32:25 Speaker 00 The environment's gonna change within each semester. That's how fast this is moving.
32:29 Speaker 01 You shouldn't chat reflexively.
32:33 Speaker 00 Yep.
32:34 Speaker 01 No...
32:35 Speaker 00 Correct.
32:36 Speaker 01 You want it, all of that speech tactics.
32:39 Speaker 00 If you remember, I said that it's good at content, terrible at context. The reason to have your own private memory space is to set that context. That doesn't mean that now the CIA and the NSA know all about you. Well, they probably do, but you you now have your own private workspace that where the context is defined. I think there's value in that because otherwise you're just getting generic things that don't apply to you. And and so then there's a disconnect. And so when you start to interact with an AI tool and it gives you something that's, oh, this is junk, at a certain point, you'll stop using it. In fact, there was a very interesting
33:21 Speaker 00 study done by Microsoft and think Harvard University in the fourth quarter of this year. It's a very interesting report where they talked about the rollout of AI within corporate America, where on average, people have been getting about six hours of training on how to use AI tools. And what they found was there's an eighty twenty rule. Twenty percent of the organization, after I forget, three or six weeks, were still using it at a high level. Everyone else had abandoned it. And it's because they basically just got frustrated that they weren't getting the results. So at a certain point, they said, you know, I can do this better and faster myself. And then they just leave it
34:02 Speaker 00 aside. They they never rounded the corner. They never topped the hill to determine what it really took to be effective with it. And that is the majority of corporate America today, is they they get in it and then they step back. And then the way that they're using it, in many cases, is still very simplistic. It's in that question answer type of mode. So, you know, if you want to figure out, you know, which wineries to visit in Sonoma, that's fine. That's question answer. But it's not very sophisticated. It's not going to really move the needle, you know, in terms of your your your productivity. It saves you a few seconds of time because you didn't have to now research which wineries, you know, use biodynamics or whatever, or research wineries that don't use sulfates in their
34:45 Speaker 00 wine or whatever. It can save you minutes, but not weeks. Where this is really powerful is when it saves you weeks, not minutes.
34:55 Speaker 01 Sorry, so you've been working when you decide this year.
34:58 Speaker 00 Hmm?
35:00 Speaker 01 How do you call that edge you can give it and move it?
35:02 Speaker 00 Move it. I actually just watched a podcast on that. So there is a way to download that history from ChatGPT and actually re-upload it into Claude, into Anthropic. It actually showed you how to do it. If you want, I'll I'll send you the link to that that podcast. Yeah, actually. Yeah.
35:24 Speaker 02 Yeah, I, I, so all, what I'll do is I'll do like projects and so I'll create a project if I'm working on a new deal, I'll create a project. So like cherry tree and it has, you know, at first I just started with a thread and some underwriting debt. And then when I finally got to a point where I said this is bigger, I said summarize everything into this in a project and then give me any source files that you need to tell me that I need to upload. And so then I create, you know, a new project. I put in the project instruction and I upload any source files that I have. So I might upload, you know, an underwriting spreadsheet, a rental, a T12, any of that stuff. And then from there, I'll create threads inside of it and it knows to reference back up to questions. So when
36:06 Speaker 02 I, you know, ask about tenant eviction clauses, it knows I'm in trouble. It knows I'm in trouble.
36:12 Speaker 00 We're in due diligence right now on a, in a particular project in Commerce, Georgia. And I told the team, I said, I want us to put together a really solid due diligence plan. This is a sale leaseback scenario, which we have not done before, where we're very used to performing due diligence on the property. What we're not used to is performing due diligence on the tenant. So, because that's now a new dimension. So I went into Victor AI. Well, first thing is I said, you know, here are the top three risks that I see and that I think we need to talk about and put together a risk mitigation plan. We
36:54 Speaker 00 either accept the impact of those risks coming true or we have a mitigation plan. Okay, fine. I was getting pushback from the team. So I said, all right, let's see this risk management plan. So I typed into Victor AI, prepare a list of risks that need to be confirmed through due diligence in the case of a sale leaseback with a commercial property having a triple net lease with the existing owner. And so it captured the three or four that I had come up with on my own, no surprise. And then it actually found a few more. Couple of them were trivial and don't apply, but it came up with 10 in total. I sent that to the team and that's what they used as the starting point for the due diligence.
37:34 Speaker 00 And you know what? I got no pushback. It's amazing because somehow if it's coming, it's packaged in a very complete manner, there's less resistance than if the boss is telling you, "Go do this." There's an interesting dynamic from a human perspective that just in terms of how people engage with the business process. It's fascinating. And this was in the last 24 hours. This was yesterday.
38:07 Speaker 00 Great question. SEO with chat right now is broken. I can give you some examples. I've taken two websites recently. Now, maybe if I gave the tool complete access to, you know, postmaster tools and all of the tools that a human would have access to, like behind the covers, it would do a better job. But when I unleash the chat on a website and say, "Tell me, do an SEO analysis of this website," it flags a lot of problems that aren't there.
38:47 Speaker 00 Like, for example, it'll tell me, "Oh, your robot.txt file is not right, and I'm getting 403 errors, and your website is not searchable, so Google's not seeing it at all. You need to fix that before I can even do any analysis," and it's not true. So, because then I literally took that report, handed it off to my web team, and said, "Here's what it's saying. What do you think?" And he said, "No, no, no. It's all wrong." So, it's not quite there yet. It's flagging a lot of false positives, but I'm sure it's a solvable problem. I haven't taken the time to actually harness it properly. I think it probably could do a good job. I just haven't figured that out yet. Yep.
39:30 Speaker 00 I will often do the same analysis in more than one tool and compare the results. Uh, there, there is a tool out there called AI, uh, what's it called? Box, Box AI, I think. It's, uh, put together by a gentleman named Jaden Schaefer. And this one interface has access to forty different tools for like nine bucks a month. Now that doesn't necessarily give you a lot of resources, but it gives you access to the paid version of those. You may need to up the subscription to get access to more, but if you want to compare different AI tools with the same query, uh, it, it will do that for you.
40:14 Speaker 00 And I think that could be quite powerful, actually. And I, I haven't been doing it that way. I actually have like four or five that I use regularly. And so it's, it's a funny thing because we don't know whether the result we're getting is good. All we know is that's good enough for me to accept. Now, this is actually very interesting. In any business, the number one function that is the most ignored is quality assurance. It's an afterthought in most businesses. I mean, imagine for a moment if the software that Microsoft shipped wasn't tested. None of it would work. It barely works now.
40:56 Speaker 00 And the only reason it barely works is because they can't make a test matrix big enough to test all of the corner cases. That's one of the reasons why Apple's quality is so much better than Microsoft is because Apple has constricted the platform. So the number of permutations and combinations of test cases is much more confined, whereas Microsoft is trying to borrow the ocean with every different platform and they can never test it enough in our lifetime to test it thoroughly. Google has the same problem with Android, by the way. So QA is absolutely vital. I mean, imagine if you got into a car that didn't go through QA, would you get behind the wheel? Forget it. And yet we ship stuff to customers that hasn't gone through QA. Now I have a discipline within our
41:39 Speaker 00 Organization where even if I produce something, I send it to somebody else for review. Because I can't check my own work. I know that. I have blind spots. Everyone does. Can an AI tool perform an effective QA? I think the answer is yes. Can it generate, can you prompt an AI tool to generate the test cases to evaluate the quality of what's being shipped? This is why autonomous driving is going to be so much better than humans. And it's because your experience as a driver is just your collective experience. But in truth, you actually have very little experience as a driver.
42:19 Speaker 00 Because the time that you spend going 10 miles an hour on 101 is not furthering your skill as a driver at all. What you need is you need a lot of practice. You need to train as a driver the same way we train pilots in the simulator. Where they test, where they actually simulate engine failures. They simulate stall conditions. They simulate icing. They simulate all these different things. You want to be able to say, you want to be driving down the street and simulate, you know, a driver opening the door into traffic and a bicycle swerving around the door and then coming into traffic. You want to see that from every possible angle over and over and over again. That's how you become a great driver.
43:02 Speaker 00 Now, AI is not just taking the experience of one driver. It's taking the collective experience of all drivers. And so, layer on top of that, the ability to generate more test cases to further train the self-driving algorithm, it's going to be a way better driver soon than the best drivers. Ah, okay. I forgot. I was going to show a whole bunch of stuff there. Okay, write down land.ai. Land? Land, L-A-N-D dot AI. So if you want access to records about a property, it integrates a whole bunch of stuff that are, it may otherwise be accessible, but it just really consolidates it.
43:46 Speaker 00 Title information, surveys, NOAA records, FEMA records, all of that all in one place. Another one is called Pax dot IV, P-A-X, Pax, Pax IV, yeah, P-A-X IV. I think they just started in Utah, but I think they're branching out to other states.
44:08 Speaker 00 Does a similar sort of thing. One of them, I forget which one, will actually allow you to produce custom maps of your property and embed renderings of your concept design on the map. So it can be a very powerful tool when you go for your pre-consult with the zoning board. You can actually show them your concept on the topographic map in like in minutes that would have required you to spend a ton of money in engineering just to get, you know, 30, 50 grand in engineering just to get to that point. You can now do in a very, you know, very short period of time.
44:50 Speaker 00 I think that one's land.id that does that one. There's another one called placer.ai. So imagine that you want to do due diligence on a commercial retail opportunity. And you say, you know what? I think this is going to be a great spot to open a shoe store. So you go out, you commission the third-party market study and get all of that. You can go get your traffic study. They put a little tape on the road and all of that and get traffic counts. And six months later, you'll have all the information you need. Placer.ai, what it does is it uses the same technology.
45:32 Speaker 00 So you know how when you're driving in your car, it takes the anonymous data from all of the cell phones that are in the cars and figures out, does a moving speed estimation on them and says, well, if these cell phones are only traveling at 10 miles an hour, I'm going to mark this freeway as red. Or if you're doing, you know, 45 miles an hour, I'm going to mark it yellow. So it's using that same technology, anonymous position information, to determine traffic counts. So it will tell you what the foot traffic is going in the front door of Walmart. Cannot tell you how much you spent, but it'll at least tell you what the foot traffic is. That, for anyone trying to do due diligence in retail, is extremely powerful. Yeah, my pleasure.
Previous Innovative Property Management with Chris Lento Next A Second Brain Becomes Your Corporate Memory Page updated November 14, 2026 · transcript and summary are machine-generated