WEBVTT 3 00:06:10.570 --> 00:06:12.310 Mendy Palace: Yes, so 4 00:06:14.320 --> 00:06:20.210 Mendy Palace: 1st is, I'd like to start off by thanking you all for joining. My name is Mendy Palace. I'm here with Aaron Zlodowitz 5 00:06:20.686 --> 00:06:25.930 Mendy Palace: big shout out to the real deal the real deal for organizing this talk and putting it all together 6 00:06:26.870 --> 00:06:34.279 Mendy Palace: in the next couple of minutes we're going to cover the ais that you could be using, you should be using and the ones that people are already using for real estate 7 00:06:34.550 --> 00:06:41.729 Mendy Palace: so feel free to post any questions in the chat. I'm going to be monitor monitoring that. And we're gonna also have a couple of minutes for questions. 8 00:06:42.940 --> 00:06:59.480 Mendy Palace: So jumping right in. When it comes to real estate, we're breaking down the types of AI that are being used into 3 categories. 1, st one is generative, generative. AI. So think like bots Llms like Chat Gpt, and anything that you'd be using for creation, summarization and etc. 9 00:06:59.860 --> 00:07:06.730 Mendy Palace: The second category that we're going to be breaking down is analytical. AI. So think processing data running high level reports. 10 00:07:07.010 --> 00:07:13.140 Mendy Palace: And the the 3rd category is automation. So arm's gonna go through. 11 00:07:13.470 --> 00:07:18.829 Mendy Palace: you know, on a higher level. Some some concepts about this, but those are the the 3 categories. 12 00:07:19.010 --> 00:07:25.530 Mendy Palace: So just before we jump in is, bear in mind that yes, AI is not perfect, and yes, AI can make mistakes. 13 00:07:25.670 --> 00:07:33.659 Mendy Palace: But if you're leveraging AI, and you're looking over the output versus creating the result from scratch, then there's no denying that those benefits are huge. 14 00:07:34.000 --> 00:07:37.387 Mendy Palace: So now, jumping right in, I'm gonna pass over the floor to 15 00:07:38.120 --> 00:07:45.220 Mendy Palace: our own to start going through some of the actual practical use cases for AI and real estate. 16 00:07:46.440 --> 00:07:51.756 Aaron Zlotowitz: Thank you. Mendy. And yeah, again, thank you to the everyone for coming. Thank you to the real deal team. 17 00:07:52.380 --> 00:07:57.080 Aaron Zlotowitz: okay. So I want to start with a. 18 00:07:57.290 --> 00:08:02.699 Aaron Zlotowitz: So this is a report. That. Okay, I'll explain to you. I'll explain what this is. 19 00:08:03.242 --> 00:08:12.000 Aaron Zlotowitz: We had a this is, we actually put this together for a client who wanted they they have. 20 00:08:12.170 --> 00:08:18.209 Aaron Zlotowitz: they they want. They basically take their T 12 every month. Every month they get a. This happens, takes our T 12 21 00:08:18.630 --> 00:08:32.929 Aaron Zlotowitz: for their for their buildings, sends it into their pro, sends it into an AI that reads through every single line. Item says, Okay, this with the jet with, give it, gives it the location, gives it the the unit, the the amount of units, the property type. 22 00:08:33.090 --> 00:08:48.290 Aaron Zlotowitz: and it goes through every line item and says, this seems like it's too high. That seems like it. That seems like it's too high, and it flags. It flags a bunch of items. Each one of those flag items that gets then gets sent to another AI, which is specialized for research. 23 00:08:48.400 --> 00:08:59.789 Aaron Zlotowitz: and it goes and it checks up in that location, checks up local comps. It checks up everything and looks. Okay. So let's say, over here. Trash removal was, is was $10,000 was almost $11,000, 24 00:09:00.600 --> 00:09:13.449 Aaron Zlotowitz: says, Okay, a unit, 30 unit, a 30 unit property in this area should not have a bill of $11,000 and actually goes in the checks and sees what's what the normal pricing is. And then it does that for every single line item it finds 25 00:09:13.590 --> 00:09:20.239 Aaron Zlotowitz: collates it all together and actually categorizes it by type. See, this is the operations and maintenance report. 26 00:09:20.746 --> 00:09:35.669 Aaron Zlotowitz: That they have. They have different people at the company who deal with different categories, they actually have it split the a 3rd AI, that actually takes all the results and splits them up properly, and then these get put into emails combined into emails and sent out to the relevant people. 27 00:09:35.780 --> 00:09:37.349 Aaron Zlotowitz: This happens every single month. 28 00:09:37.990 --> 00:10:06.753 Aaron Zlotowitz: So this is a real report. I just, you know, anonymized it for the privacy of the client. But this is a real report. It went through it, saw that their traffic was too high. Their payroll expenses for 30 unit property of $20,000 just doesn't seem like it makes sense. Their temp temp staffing is $9,000. It explains why, and then it goes into a slightly more detailed version of of what's going on. It. Actually, you know, it actually breaks down. Why, it's too high. 29 00:10:07.330 --> 00:10:15.199 Aaron Zlotowitz: And then there's actually a much more. This is really the back end, almost. This is the AI's actual thought process of how it came up with 30 00:10:15.350 --> 00:10:17.010 Aaron Zlotowitz: why, this is too high. 31 00:10:17.509 --> 00:10:37.969 Aaron Zlotowitz: So this is kind of just here more, as a, you know, as like a reference point. The main thing is these 1st 2 pages, but they get this report every single month on every single property, all, just all sent out automatically. You don't have to think about it just goes, and there's they don't have to have someone sitting there digging through. They could focus their attention on what matters and not have to. 32 00:10:38.130 --> 00:10:47.979 Aaron Zlotowitz: So, you know, have like pay a bunch of pay, a bunch of, you know, offshore people, or whatever to actually go and do this. This happens within minutes. This happens within minutes. And it happens automatically. 33 00:10:51.740 --> 00:11:00.720 Aaron Zlotowitz: okay, so that is so that's this. That's that's like, Mindy said. This is analytical. This is, you know, this is like a real like an analytical 34 00:11:02.560 --> 00:11:18.999 Aaron Zlotowitz: the next thing I wanted to cover the next use case, this is kind of, I'm gonna show more like this is a this is kind of, you know, a mix of what you could do now, compared to where we're gonna get to. So this is what we could do right now. 35 00:11:20.040 --> 00:11:38.909 Aaron Zlotowitz: this is where this is. Just chat. Gpt. Now, you'll notice I switched it to O 3. I'm gonna send this prompt. Now, basically, what I'm trying to do is is, let's say I'm looking to acquire some properties. So I wrote out, I had a actually had it helped me write this a very detailed, prompt to do research, finding me finding me prospects for my acquisitions. 36 00:11:38.970 --> 00:11:56.989 Aaron Zlotowitz: I'm going to post this prompt in the chat. If anyone wants to if anyone wants to reuse it. But while I'm going to explain what we're what it's doing and what where this is going, as it researches. So this is gonna I chose the O 3 model in particular, because it's going to take some serious time looking this up 37 00:11:57.110 --> 00:11:59.652 Aaron Zlotowitz: as at least take at least a minute. 38 00:12:00.900 --> 00:12:04.120 Aaron Zlotowitz: I'm going to here one second. Let me just drop this in the chat, and then I'll 39 00:12:04.830 --> 00:12:12.690 Aaron Zlotowitz: that's the prompt oh, sorry! It's too long. I will send it out after anyway. So 40 00:12:13.490 --> 00:12:24.400 Aaron Zlotowitz: so anyways, right? So where this is now is chat where you could do this, but it could really only pull from. Well, see, you can actually see where it's pulling from. We could only pull from the web and anything that's available made available on the web. 41 00:12:24.957 --> 00:12:28.260 Aaron Zlotowitz: The kind of pay gated 42 00:12:28.460 --> 00:12:34.534 Aaron Zlotowitz: the pay gated ones like you know your zoom info and and 43 00:12:35.240 --> 00:12:36.379 Aaron Zlotowitz: like. Yeah. The all the. 44 00:12:36.380 --> 00:12:36.850 Mendy Palace: Infrastructure. 45 00:12:36.850 --> 00:12:51.819 Aaron Zlotowitz: Thank you. And like all these that they're pay gated, they don't re, and they usually they usually don't even have Apis. You can't even build something that would access them. But they all. But so right now, it's really just what you can find online. But 46 00:12:52.210 --> 00:13:00.989 Aaron Zlotowitz: where this is gonna go is if someone is going to make a real, a property like a database that has a good Api and connect to Chat Gpt, if you pay for access. 47 00:13:01.340 --> 00:13:10.920 Aaron Zlotowitz: You're gonna that means everyone else gonna have to do that because they're gone. Because why would I pay an analyst to go sit there doing it if I could do it myself with Chatgpt in a fraction of the time and the cost. 48 00:13:11.080 --> 00:13:12.830 Aaron Zlotowitz: So everyone's going to have to do it. 49 00:13:13.358 --> 00:13:18.439 Aaron Zlotowitz: So then you're gonna have. What's happening now is gonna be on. You're gonna see on steroids. 50 00:13:19.030 --> 00:13:27.529 Aaron Zlotowitz: But it's there's but and so really, we're already at the point where you're going to see in a second. You can get a very good report 51 00:13:27.870 --> 00:13:38.430 Aaron Zlotowitz: out of just in within a minute, 2 min, maybe 5 min. It might take that it might take that long but and that's only on web data is only going to get better and better and better. 52 00:13:39.430 --> 00:13:49.049 Mendy Palace: Just aren't. I see a question in the chat is someone's asking when you would be using? Well, this is when you would be using an O 3 versus, you know, 4 0, or the different models. 53 00:13:49.270 --> 00:14:02.829 Aaron Zlotowitz: Okay? Sure. So 4 0 is like it says here, great for most tasks, meaning, if you're writing, it's it's the difference is the real difference is, oh, 4. 0, doesn't do this. It doesn't do this. Sit there reasoning. And look, you can look. 54 00:14:02.970 --> 00:14:10.470 Aaron Zlotowitz: you can see it's thinking through. It's looking at results. It's it's it's going through this. It's basically in its head for lack of a better word. 55 00:14:10.570 --> 00:14:20.050 Aaron Zlotowitz: So that's for for writing. You don't need it to sit there doing that for writing. You just want it to be creative and just flow. So email drafting that sort of thing, basic back and forth. 56 00:14:21.390 --> 00:14:22.800 Aaron Zlotowitz: Basic web search. 57 00:14:22.910 --> 00:14:46.649 Aaron Zlotowitz: O, 3 is kind of like the polar. Opposite is their flagship like high end model. It's the most expensive one if you're paying if you were paying per use case as opposed to the subscription model, I'm on per use as opposed to subscription I'm on, but that's what O 3 is, for O, 4 mini high is kind of O, 3. But using a 0 4 mini, and O 4 mini high are the same model high. Just means you'll spend more time doing this. 58 00:14:47.010 --> 00:15:00.539 Aaron Zlotowitz: And O 4 mini is kind of the like. O, 3, but a weaker version of it. So it's cheaper, and you can use it from, and you can use use it more, but it's but it doesn't give you as good of a result. So anyways. So now. 59 00:15:01.280 --> 00:15:08.200 Mendy Palace: And I just see a question if you can just go over the the prompt and in terms of like, how how to prompt is there 60 00:15:08.400 --> 00:15:10.130 Mendy Palace: prompt optimization. 61 00:15:10.660 --> 00:15:15.290 Aaron Zlotowitz: Yeah, sure. So let's go. We'll go through the prompt, really quick, and then we'll go look at the results. So. 62 00:15:15.290 --> 00:15:17.249 Mendy Palace: How you came up with this specific, prompt yes. 63 00:15:17.250 --> 00:15:30.139 Aaron Zlotowitz: So, okay, so the general idea with the prompt is, you want to be as you want to give it as clear instructions as possible. We're not dealing with a magic trick. AI is not a magic trick. So if you're gonna go, use a if you're gonna give a prompt that 64 00:15:30.240 --> 00:15:36.819 Aaron Zlotowitz: if you were going to talk to an intern, and the intern wouldn't understand what you're saying. AI is not going to either, most likely. 65 00:15:37.180 --> 00:15:47.769 Aaron Zlotowitz: So you need. So I do. I came very specific. I'm looking 1st of all, I explained very clearly what I want. I want acquire commercial properties in New York City for my real estate portfolio. I need it, and now I want 5 of them. 66 00:15:47.920 --> 00:16:10.830 Aaron Zlotowitz: Now I went into the criteria. This is this is pretty. This is, I think this was kind of generic. This is what AI really came up with for the specific criteria. But the point the point isn't what the criteria are. The point is, you should put in what you what your criteria are, you put in? What kind of properties you're looking for, how you, what you should be looking for. If you're looking for the owners all these different, what details you want to get back? 67 00:16:11.360 --> 00:16:19.050 Aaron Zlotowitz: And then I also said, put in an organized report format. And then so then this is what it came came back with. 68 00:16:19.050 --> 00:16:32.210 Mendy Palace: If I can just add to that, I would say what what I find is a good, you know, hack to when you're trying to think of a prompt is to kind of ask Chat Gbt, what you should be asking it. So you can. You can say this is my result, and this is what I'm trying to get. 69 00:16:32.320 --> 00:16:44.590 Mendy Palace: Write me a prompt and kind of ask me questions to, to clarify what it is exactly that I'm trying to get to, and then going through, you know those steps, you should be able to get pretty solid, prompt out. That should give you a good output. 70 00:16:45.840 --> 00:16:52.900 Aaron Zlotowitz: Exactly. Thank you, Benti and then, right so now you could see here what I asked for. I actually get the 71 00:16:53.360 --> 00:17:22.209 Aaron Zlotowitz: I actually get with sort, with sources, I get results. I have 5 properties that I can go actually messed up the formatting a little bit. But there's number 2. It came back with a few properties. And then it gave me. It gave me a full report. You can actually, if you want to do this. You can, if you want to really get in all in, you could, if you would do that same, prompt into deep research. I'm not going to do it now, because it's going to take about 20 min 72 00:17:22.730 --> 00:17:43.729 Aaron Zlotowitz: to run. But if you could do into deep research and it'll do a much. This is a deep research is also it's it kind of it's like this. It's the same model that did this. But it goes. And it's given like 20 min, and like the ability to store its information and look back at it so it could go for much longer, and it reads through dozens, if not hundreds of sources. 73 00:17:43.850 --> 00:18:01.029 Aaron Zlotowitz: So it really actually takes like 20 min to run. But that is, that's that's like the next level of this. But you have to be very specific with your prompt, because if it doesn't have a clear direction, it's gonna spend 20 min down the wrong down the wrong going down the wrong way instead of 2 min or 3 min. 74 00:18:01.550 --> 00:18:03.010 Aaron Zlotowitz: But that's. 75 00:18:03.150 --> 00:18:15.830 Aaron Zlotowitz: I think, a good picture of just this is really where you could be using chat to be today and just have in mind of where this is going to be going to like you could be. Re, this is something you could. You would be paying an analyst for now you can have Chat. Gbt. Do it for you. 76 00:18:17.970 --> 00:18:27.149 Aaron Zlotowitz: The 3, rd a 3rd use case that we've done for that we've done. And this is more just really for data processing is I can't. I don't really have a demo for this 77 00:18:27.541 --> 00:18:42.460 Aaron Zlotowitz: so I'm gonna stop sharing, because the only the the demo we have is is client information only, and I couldn't. I wasn't able to really figure out a way to censor it. But they they want us that we built for them where they have their underwriting model in. It's a Google sheet 78 00:18:42.874 --> 00:18:50.549 Aaron Zlotowitz: and so they, you know, they have someone who sits there and takes the Om reads through it, finds all the relevant information, puts it into the Om. 79 00:18:50.710 --> 00:18:57.140 Aaron Zlotowitz: and then, you know, they maybe do a little research to add it in. They. 80 00:18:57.930 --> 00:19:12.959 Aaron Zlotowitz: So it's a little like they do a little research on, say, the property tax. This was a cool. I actually, I actually, as it turns out, New York has a an Api for all the Department of Finance data. So I was able to build where their 81 00:19:13.070 --> 00:19:14.056 Aaron Zlotowitz: where they're 82 00:19:15.000 --> 00:19:38.299 Aaron Zlotowitz: they put in the they drop in the Om into it. They get an upload button, they drop in the Om their model, then gets spit out a new, a fresh copy of their model gets spit out with all the information pulled out and written in exactly where it's supposed to be perfectly formatted, and it actually goes, and it pulls in the bit of information they needed that was coming from outside of that was the property tax that actually 83 00:19:38.720 --> 00:19:44.049 Aaron Zlotowitz: gets that from the Department of Finance's database and populates that in. So they get a fully done. 84 00:19:44.420 --> 00:19:52.739 Aaron Zlotowitz: fully fully done with the rent roll and everything in under 2 min, and with just literally one just drag and drop. That's the whole. And and they're done. 85 00:19:54.470 --> 00:20:03.050 Mendy Palace: I just see a a question in the chat is when you would be using, let's say, the the paid version versus the free version? In terms of 86 00:20:03.150 --> 00:20:08.530 Mendy Palace: output. Is there a difference what your usage are? Things like that within chat? Gbt, yeah. 87 00:20:08.530 --> 00:20:09.890 Aaron Zlotowitz: Okay. So 88 00:20:10.130 --> 00:20:22.660 Aaron Zlotowitz: if you're just using it as like a email spell check, then you could just stick on the free plan unless you're using it so much that you get told. Oh, you're you're reaching your users limits for the next few hours, and you keep running into that. They're like. All right, I gotta upgrade. 89 00:20:22.710 --> 00:20:41.509 Aaron Zlotowitz: You don't get access to any. You don't get access to the O. 3 model without on the free plan. You don't get you get a limited access to the O 4 Mini model. That's what the think deeper. There's like a think or reason button on the free plan. That's what that's referring to. It kind of moves your your request over to the O 4 Mini. But you don't get to really get that choice. 90 00:20:42.138 --> 00:20:48.550 Aaron Zlotowitz: You get stricter limits on, on 4, on the regular 4 0 model, and 91 00:20:48.800 --> 00:20:53.349 Aaron Zlotowitz: you get you get a lot fewer. Deep research requests per month. 92 00:20:57.790 --> 00:21:09.959 Mendy Palace: And then I I just going a little deeper into into an agent. How would you define an AI agent? What is that considered in terms of, let's say, these these tasks that you're setting up where you can have them be done automatically and 93 00:21:10.610 --> 00:21:12.129 Mendy Palace: just a little deeper into that. 94 00:21:12.130 --> 00:21:37.923 Aaron Zlotowitz: Okay. So an AI agent is a purposely vague term, because the people who made it who came up with it were trying to sell things but what it? What it's generally refers to is anything more than just, you know, sending a message to Chat Gpt, using AI for something the next level. So that property management, the property, the the property analysis was a it wasn't. It could be called an agent. The the 95 00:21:38.580 --> 00:21:44.660 Aaron Zlotowitz: the what's called again the the oh, the Om analyst analysis for the underwriting could also be called an agent. 96 00:21:44.800 --> 00:21:58.896 Aaron Zlotowitz: The deep research is an agent that's built into the Chat gpt app at this point, then all of them have it. Grock has that Gemini has that Claude recently got it, too? Rock has that it's everyone has it. 97 00:21:59.640 --> 00:22:08.250 Aaron Zlotowitz: but it that the agent is really anything that's not just you talking to a chat. But it's kind of giving a Chatbot the ability to do more on its own 98 00:22:08.610 --> 00:22:26.580 Aaron Zlotowitz: agency for lack of a better term. But it's it's a very broad category. I saw that people were asking the differences between the different models, the different providers. So between chat versus Grok versus copilot versus Claude versus Gemini, so 99 00:22:27.420 --> 00:22:36.139 Aaron Zlotowitz: Chatgpt and Gemini are probably the closest to each other in the sense that they have the same, more or less the same features available to both. 100 00:22:36.585 --> 00:22:40.929 Aaron Zlotowitz: There's I don't think there's much, really. There's not much that one has that the other doesn't. 101 00:22:41.396 --> 00:22:58.619 Aaron Zlotowitz: It really comes down to which one you really, if you're locked into an ecosystem. So you're you're using Google Workspace for everything. Your I, your company's. It is like, you know, we're using Google, you're sticking with, you're gonna stick with gemini otherwise it's kind of for lack of a term, a personality thing like they. 102 00:22:58.880 --> 00:23:12.210 Aaron Zlotowitz: They're the AI is the 1st technology that can be said to have a personality. In a sense, it's like, I don't use gemini because I just find it. I don't. I don't enjoy the conversation for lack of better term. It's just hard to. There's no, but not much of a real difference. 103 00:23:13.700 --> 00:23:26.040 Aaron Zlotowitz: Claude is better at coding, and it's better at writing. So if you're just doing writing, or if you're someone here is anyone here is more technically minded. Claude is amazing at coding. Actually, that's the only other one I pay for separately besides, for Chat gpt 104 00:23:26.650 --> 00:23:28.270 Aaron Zlotowitz: for that reason. 105 00:23:28.692 --> 00:23:44.397 Aaron Zlotowitz: But it does not. It's not as feature rich as chat, gpt as fewer as has fewer models. It doesn't have it recently got the research, but they don't have. There's they don't have canvas. There are certain things they just they don't have. So not a feature parity. 106 00:23:45.130 --> 00:24:01.899 Aaron Zlotowitz: Grok is kind of a wild card. I've been I I don't really use it because I don't. There's nothing it has that I need that not that other models I have don't but also they do kind of play with the system, prompt a lot, which is kind of 107 00:24:02.050 --> 00:24:04.509 Aaron Zlotowitz: what defines how it works. 108 00:24:08.130 --> 00:24:23.970 Aaron Zlotowitz: it does kind of defines how it works, but it kind of change. The process defines that works. And it kind of changes a lot because Elon Musk has a whim. So it's kind of I don't consider it to be necessarily dependable. And copilot is basically just Microsoft branded Chatgpt. It's actually running Chatgpt in the back end. 109 00:24:25.410 --> 00:24:34.260 Mendy Palace: And then just going back to to with agents, because that's, you know, a lot of the buzzword a lot of the things that that you hear about. You know, agents? The the general concept is 110 00:24:34.470 --> 00:24:43.419 Mendy Palace: what they're coming to do and what they're trying to eliminate is the manual, the repetitive tasks that you're doing, that you're setting up these agents to to kind of do for you. 111 00:24:43.837 --> 00:25:11.069 Mendy Palace: And then just I don't know if you can. You can just demo this with Chat Gpt, you know, showing how you can set up like a task where you know every morning you can have it pulling for you. You know the headlines from whatever you would want it to pull for you. You can have it pulling, you know the the latest stock prices, for for you know things that you might be interested, the treasury, the yield, any kind of information that that you would online yourself go and search you can have that task 112 00:25:11.390 --> 00:25:16.379 Mendy Palace: set up with for you pretty simply, just by going into Chat Gbt. And and doing that. 113 00:25:16.860 --> 00:25:43.240 Aaron Zlotowitz: Okay, so that's actually another difference between that between oh, 3 and O 4 versus 4. 0, I know their naming scheme is, don't worry is is horrendous. But oh, 3, and O 4, the ones that say reasoning over here. You can schedule this. You could schedule. You can schedule things like you said so I'm just gonna say this, let's say, every Monday morning I want to look. I want a report of 5 of 5 vacancy of 5 acquisition acquisition prospects. 114 00:25:43.410 --> 00:25:46.730 Aaron Zlotowitz: So I just say, run this for me every Monday morning at 8, at 8. 115 00:25:47.395 --> 00:25:49.920 Aaron Zlotowitz: It'll give it a second. It's going to. 116 00:25:59.850 --> 00:26:00.530 Aaron Zlotowitz: Okay. 117 00:26:00.880 --> 00:26:04.720 Aaron Zlotowitz: I see I can go. I'll go. I'm gonna press edit, so you can see. And now 118 00:26:04.990 --> 00:26:17.720 Aaron Zlotowitz: it's going to it. Condense the prompt down, but I could go in if I wanted to. I can go in and edit this to be the full prompt. And it's gonna go. And it's going. This conversation will do that same. Prompt every single week at Monday, at 8 Am. 119 00:26:20.540 --> 00:26:22.420 Mendy Palace: That's a pretty neat. 120 00:26:22.580 --> 00:26:42.959 Mendy Palace: and then just getting a little. A little more technical is is our. If you can give a little under the hood where the Om creation where we demoed that you can take an Om and have that information automatically pulled into, you know, an underwriting model. What kind of is like the the under the under the under the hood of that, where you know how that's working, those workflows. 121 00:26:44.070 --> 00:26:50.018 Aaron Zlotowitz: So what in that particular case? So so what it's doing, what's what it's doing is 122 00:26:50.810 --> 00:26:54.880 Aaron Zlotowitz: So that was a mix of in that particular way. We did it. It was 123 00:26:55.370 --> 00:27:14.860 Aaron Zlotowitz: it was there was a it was a I built for it was Google sheets. Google sheets is actually a lot better for this because they have the ability to build programs on top of Google sheet. So I built a program on top of a Google sheet that takes an upload sends that Pdf upload to. I used Claude for this because it's it's excellent at analysis. 124 00:27:16.250 --> 00:27:36.031 Aaron Zlotowitz: Analyze the whole Pdf. Reported came back with a very structured output that I was able to then sort out and sent that back to the program. The program took that output and placed it. I pre programmed where everything's supposed to go, placed everything where it's supposed to go, and then separately, I had an Api call to the Department of Finance to get the taxes there was the. It's 125 00:27:37.340 --> 00:27:45.971 Aaron Zlotowitz: it's it's it's a bit. It's a very. It was a very. It is a very technical process to do that. That kind of agents is a real technical build out? 126 00:27:46.550 --> 00:28:09.030 Aaron Zlotowitz: yes, you could connect an agent to a proprietary system to report an internal data if that internal data system had proprietary system has an Api then you can build, then you can build a program. And AI that has the ability to do. It's called function calling, or tool or tool use. And you kind of the same. This is what Chat Gpt did for web search, or they all did for web search. They kind of 127 00:28:09.030 --> 00:28:24.130 Aaron Zlotowitz: told the model. Hey? If you need to do a web search, write the following words, and it'll write out internally, it writes out, you know, do a web search for, and then it puts in, quotes what it needs to search. And then a tool actually goes and performs that search and gives it back the results. 128 00:28:24.160 --> 00:28:41.879 Aaron Zlotowitz: So you can set some if you have. If you so you or if you want, you could ask chat, help you build this, or if you have someone on your team, you would do this for you. You could go and have them go. You can set up the tool you get the Api access just for those who are not aware of what an Api is. An Api is just 129 00:28:42.060 --> 00:28:43.989 Aaron Zlotowitz: a kind of is a key, that 130 00:28:44.310 --> 00:29:06.250 Aaron Zlotowitz: it's. It's a key that that systems are able to set up, that you know. How do like you can to give people access to data. There's a link that you can that you can send. You can send requests to, and it sends back information, and some of them can be authorized only need Api key. It's like, you know, only like a like an email password kind of concept before programs. You can only access our Api, if you have a key, and then they could use that key to Bill. 131 00:29:07.910 --> 00:29:18.289 Aaron Zlotowitz: you can use the key to Bill to to Bill, like, you know, chatty won't let you will, will, bill you, for every time you use their Api that kind of thing. So you can. If you get the Api key to that 132 00:29:18.851 --> 00:29:31.329 Aaron Zlotowitz: to that external source you can to to that proprietary source you get an Api key that only lets you into your part of the database. You can then have it. Use that as a tool use and go send and get request data from it. 133 00:29:32.400 --> 00:29:41.490 Mendy Palace: If you can just run through what are, you know, tools that you would be using to make this kind of thing? What kind of technical background. Do you need you know how the the barrier to entry getting started with that. 134 00:29:41.780 --> 00:30:01.820 Aaron Zlotowitz: You could use look like these. No code pro platforms like make.com zapier any of those to do any of to do these things. If you don't have any. If you if you're if you're starting out, it's the definitely the best way to get started with with time and experience of structure. 135 00:30:02.260 --> 00:30:19.009 Aaron Zlotowitz: with time, experience, time, experience of like how to structure these things. You could start use asking, you know, chats with your Claude to write you the code for these things and build more custom programs. But you could start out very easily. They have these platforms make.com zapier, where everything is very modular. It's like, you know, here, connect to 136 00:30:19.280 --> 00:30:41.340 Aaron Zlotowitz: my email, connect to Chat, Gpt, connect to this, and so made very straightforward, and you could even ask, Chat Gpt. Hey, how would I build this in zap here, and it'll give you a step by step of what to do and then there's some customization that we need to be done for a proprietary system that they don't have built in built into. But Chatgpt could very easily help you with that. If you can get pointed at the right documentation. 137 00:30:42.240 --> 00:30:47.309 Mendy Palace: I see a question. Which system would you would you recommend to connect for internal data reporting? 138 00:30:48.940 --> 00:30:53.449 Mendy Palace: So you wanna, you know, analysis with with your internal proprietary data 139 00:30:53.840 --> 00:30:58.859 Mendy Palace: can be like, let's say, within custom gpts, or you want a little bit of a higher level. Analytics. 140 00:30:59.280 --> 00:31:07.129 Aaron Zlotowitz: It depends it it there's the answer is, whatever. Any platform you'd like that has an Api that can handle it. You could speak to your customer service. Rep, 141 00:31:07.670 --> 00:31:33.670 Aaron Zlotowitz: I would do it over email and have chattypt kind of audit that if you're not, if you haven't dealt with this stuff before? So you can know if they're trying to. You know, they're just trying to upsell you. And they're not actually giving you, telling you the real, the real answers. But you wanna you should make sure that has an Api has Api access. And it's it's a well made Api that the things you want to access are accessible via that Api, and then otherwise, any system can can access any of these can can query Apis. 142 00:31:33.670 --> 00:31:39.470 Aaron Zlotowitz: So you don't have to worry about the platform, which platform, as long as that platform has Api access. 143 00:31:41.530 --> 00:31:53.629 Mendy Palace: And then, in terms of, let's say, processing documents where you want to to put a document in there to have it break it down like, you know, lease document things like that when it comes to processing Pdfs and and information which 144 00:31:54.060 --> 00:31:59.130 Mendy Palace: Lm, or which you know off the shelf tool. Would you use? That is the best for this specific task? 145 00:31:59.130 --> 00:32:25.170 Aaron Zlotowitz: If you're using chat, gpt, just stay on O, 3 use O 3 use O 4 mini high, or oh, 3, 0, 4, 4. 0, is gonna be again the naming scheme. I'm sorry 4. 0, is not gonna is not really gonna be up to the task, in my opinion oh, 3 is gonna be the best. But if you're you're there's there is somewhat of a limit unless you want to upgrade to the 200 month plan. There's a somewhat. If you run into usage limits, you can decide if it's worth it for you. But that's I would say. Oh, 3 146 00:32:25.570 --> 00:32:32.980 Aaron Zlotowitz: or O 4 mini high. You could try also, it should do a decent job. I use Claude for these sort of things. I think it's 147 00:32:33.300 --> 00:32:56.219 Aaron Zlotowitz: it is very again person that comes as person is very analytically minded for lack of a better term. And it's very. It kind of gets things clearly. So I find that's what I use for the for the underwrite, for the underwriting model. I use Claude because I find it to be very good at this sort of thing, Gemini. 2.5 pro. If you're on gemini Grok 4 should be good at this 148 00:32:56.410 --> 00:32:57.820 Aaron Zlotowitz: co-pilot. 149 00:32:58.010 --> 00:33:00.989 Aaron Zlotowitz: I don't. I don't know. I don't. I don't use copilot. 150 00:33:02.260 --> 00:33:12.640 Mendy Palace: And then I see a question we can get a little more into in terms of like Bill paying accounts, payable keeping track of invoices. What we've done in that space. How we've, you know, automated those processes. 151 00:33:13.350 --> 00:33:19.706 Aaron Zlotowitz: Yeah. So we recently actually just built for a client. For accounts, specifically for accounts, payable 152 00:33:20.230 --> 00:33:25.090 Aaron Zlotowitz: where they where they they have an accounts, payable email address. 153 00:33:25.240 --> 00:33:30.590 Aaron Zlotowitz: and every invoice comes in. Every invoice that comes in 154 00:33:30.800 --> 00:33:33.541 Aaron Zlotowitz: gets actually read by Claude 155 00:33:34.490 --> 00:33:39.850 Aaron Zlotowitz: and added to us and put into their they keep have a spreadsheet they use to keep track. It's put into their spreadsheet. 156 00:33:40.217 --> 00:33:49.430 Aaron Zlotowitz: It's been so far, so far. Do working very well. It's you know, the the higher level models, these very when you're dealing with very simple prompts, with high level models. 157 00:33:49.460 --> 00:34:04.259 Aaron Zlotowitz: they they're usually very accurate. Club sonnet 4 is a high level model. You can even add some reasoning if you want it to be extra short. And if you actually like, you know, you wanted you were saying, we don't want any human review. We want to be like 0 tolerance. 158 00:34:04.619 --> 00:34:20.219 Aaron Zlotowitz: Then, if cost is not an object, this, again, all, relatively speaking is still cheaper than even a Va you can have it. Do you can have it. Do multiple passes like, send in the get the result, get another result, and 3 like a 159 00:34:20.219 --> 00:34:37.559 Aaron Zlotowitz: send it in 3 times and have a 4.th One kind of look over the result and choose the majority, or you can do something like that. You could build a system like that has redundancy built in, so you can for sure have definite accuracy, because, you know, the rate of the error rate over here. If the error rate is, you know. 160 00:34:37.620 --> 00:34:46.360 Aaron Zlotowitz: 10%, 15%, you don't have to. If you have a redundant a 3 time redundancy. You're statistically going to get it right every time. 161 00:34:47.260 --> 00:34:59.800 Mendy Palace: And then just a question. I know this is something that has probably crossed everyone's mind, and it's something that you hear of. What would you say in terms of the security and the privacy in terms of putting personal information into these large language models? 162 00:35:00.040 --> 00:35:17.059 Aaron Zlotowitz: Sure. Okay, so this is actually, this is actually a good thing. So I'm gonna demonstrate, I'm gonna show on Chat Gpt, how to deal with this and then there's kind of, and then it's kind of platform by platform of how you deal with this. So this is the chat. This is in Chat Gpt. You go into settings. 163 00:35:17.210 --> 00:35:32.739 Aaron Zlotowitz: You go into security and sorry you're right. Data controls my bad, and you go over here. You see where it says, improve the model for everyone. You want to make sure that this is off, because if this is on, you're giving them permission to use what you upload 164 00:35:33.220 --> 00:35:39.950 Aaron Zlotowitz: beyond that point once you've shut this off, and there's an equivalent of this in every AI Claude happens to be. He's off by default. 165 00:35:40.010 --> 00:36:05.269 Aaron Zlotowitz: but ever they all have this. This option, deep seek is not reliable. Besides, the fact is Chinese, and you know, whatever the security implications that has besides, for that they also they don't have the option to turn this off, and you use deep seek on the deepseek site you. They are using your information. There's no way for you to even say Tell them not to. But everyone has that. Everyone has that concept. Once that's off. 166 00:36:05.370 --> 00:36:25.840 Aaron Zlotowitz: there's no real difference between chat to be between Chat Gpt, which is run by Openai versus Google, which you're trusting the information or Microsoft, or wherever it is. It's just, however, you evaluate that. Do you look at their security, clear their security evaluations? You look at what? What their official policies are. All these different things. 167 00:36:25.840 --> 00:36:41.499 Aaron Zlotowitz: Do you? Do you go to them? Do you? Do you trust them? Why do you trust them so? Whatever, however, you determine which software vendors you trust? Openai is no different anthropic is no different as long as you once they say they're not doing it. Then they're in the same boat as everyone else. 168 00:36:43.110 --> 00:36:45.270 Mendy Palace: Okay. So thank you, Arne, we're gonna 169 00:36:45.410 --> 00:37:03.070 Mendy Palace: stay on after for questions. But just, you know, to closing remarks is that it's like we say right now, AI is at the worst place that it'll ever be, you know. It's only getting better. It's only going to be more powerful. The use cases are endless. The way to start using it is literally by by just starting to use it jumping in. 170 00:37:03.210 --> 00:37:09.100 Mendy Palace: And if you all you know, found this valuable. And thank you. 171 00:37:13.110 --> 00:37:13.800 Aaron Zlotowitz: Alright! 172 00:37:13.800 --> 00:37:15.420 Aaron Zlotowitz: So we're. 173 00:37:15.420 --> 00:37:18.369 Mendy Palace: Going through some questions that I saw is. 174 00:37:18.750 --> 00:37:32.090 Mendy Palace: when is it going to get to the point of taking the place of you know. Say, the paid tools. You know, Costar property shark, loop net where you know they're actually paying for that information. You know, they're doing the the manual research of finding that information out. 175 00:37:33.760 --> 00:37:38.529 Aaron Zlotowitz: So I I think that depends on really cause these they all make their a lot of their 176 00:37:38.650 --> 00:37:43.024 Aaron Zlotowitz: their information that makes them valuable is is is 177 00:37:44.280 --> 00:38:01.929 Aaron Zlotowitz: public. It's all public information technically, but it's not very accessible, because let's say, it's either it might be available in file cabinets in some municipality somewhere, or it's it's available online. But it's not scrapeable, because it's in like some old, really old fashioned database that's not set up for the modern web. 178 00:38:01.980 --> 00:38:27.420 Aaron Zlotowitz: So then, it's until that point they're not really going to be able to. There's not much you can be able to do because they're having the man hours to go in and do it. So the hands on until you know Tesla has the optimus robots actually showing up. There's nothing you're showing up to do it. There's nothing to replace it with. You need humans hands on to go scan it in or write it down, or whatever they're doing. If it's 179 00:38:27.800 --> 00:38:37.259 Aaron Zlotowitz: scrape from a website and actually have to do manual thing there, there are ais that can use that can use websites. And so but they're not amazing. 180 00:38:37.280 --> 00:39:06.500 Aaron Zlotowitz: I don't have a forecast, for when they will be, because they've been around for well over a year now, and there's been virtually no progress on on their capabilities, maybe a little bit here and there, but nowhere near the other, the other AI progress we've seen, and so I have no forecast on when they'll be able to do that. But eventually, presumably eventually, they'll get it to the point where it can go through municipality website or the municipalities will update their websites to be, let's say, have an Api like, again, the the New York Department of Finance, anything that's on the New York Department of finance website. 181 00:39:06.500 --> 00:39:17.130 Aaron Zlotowitz: except maybe Pdf uploads. But the the data that in the data, with most of the data within them is you can run an Api request for. So there's no, there's no real barriers. 182 00:39:17.500 --> 00:39:23.738 Aaron Zlotowitz: anything that you could find that way. Costar has no mode on that. There's nothing with them. What are they going to do? 183 00:39:25.860 --> 00:39:49.339 Mendy Palace: I find something that that's a tool that's pretty basic, that that's really powerful. And it's for sure. You know, beginner, getting started is custom Gpts where where it can be, let's say, on a on a company level, where you're putting in your your company, you know, policies and procedures you're putting in. Let's say your your lease documents and you're asking it questions. You know you could set up those reminders. When are the leases coming due? Be having it. Answer. You know. 184 00:39:49.930 --> 00:40:17.489 Mendy Palace: customer service tickets where that's that's it's a pretty basic setup. So then, Chatgpt, it's just basically what you're doing is you're putting in a prompt once it's living there and then you're you know you want to. Let's say, from a document. Anytime you put a a lease document in. You want it to pull, you know. When does this lease end? What are the different things within that lease? So that's something that you can have that prompt built in. And then all you need to do is just drop the document in there, and it's gonna give you the output as if you just put that whole prompt in. 185 00:40:18.950 --> 00:40:24.550 Aaron Zlotowitz: Let me show. Let me show an ex. Let me show an example of how you can make those. 186 00:40:28.300 --> 00:40:36.210 Aaron Zlotowitz: one second. Here we go. Okay, it's going a bit slow. I actually, I'm just gonna do their creation. So let's say, we're looking for one for 187 00:40:36.740 --> 00:40:48.220 Aaron Zlotowitz: answering customer service queries based on the attached 188 00:40:48.430 --> 00:40:52.140 Aaron Zlotowitz: document. So you don't even need to know how to make this. I just said what I want to do. 189 00:40:52.290 --> 00:40:55.689 Aaron Zlotowitz: And then Chatgpt will actually go and make that 190 00:40:56.087 --> 00:40:59.570 Aaron Zlotowitz: when this is finished I'm going to go into the con. It's done. 191 00:40:59.760 --> 00:41:02.149 Aaron Zlotowitz: I'm gonna go into configure over here. 192 00:41:02.820 --> 00:41:03.980 Aaron Zlotowitz: Said it was done. 193 00:41:05.020 --> 00:41:05.790 Aaron Zlotowitz: Second. 194 00:41:09.480 --> 00:41:12.682 Aaron Zlotowitz: there we go see customer service assistant. 195 00:41:13.330 --> 00:41:17.811 Aaron Zlotowitz: just. It's given instructions. Only work off the attached document. 196 00:41:18.390 --> 00:41:25.700 Aaron Zlotowitz: and then I would go over here, and I would press upload files, and I would give it our customer service like rules. And then it could answer those questions. 197 00:41:26.204 --> 00:41:45.955 Aaron Zlotowitz: I could give it choose which tools as access to? Can I search the web or not? Recommended model if I want, if I wanted to? And then you could publish this. If you have a teams, count, you can publish this internally within your within your within your organization. If you have a regular count, you could publish it either for everyone or for 198 00:41:46.580 --> 00:42:02.250 Aaron Zlotowitz: or just, for if you want to, just for just for your own private use, or only access with the link those but you can. So I have a few of these I made for myself. More programming related, because, you know, that's what I deal with all day. But that's the concept. 199 00:42:06.480 --> 00:42:12.780 Mendy Palace: Okay, so there's, I guess we'll go through if there's any more questions. Now is the time to put them in? 200 00:42:13.020 --> 00:42:19.139 Mendy Palace: If not. I wanted to say. Thank you. I'm putting our contact information in the chat where you can. Reference that 201 00:42:19.740 --> 00:42:20.260 Mendy Palace: the. 202 00:42:20.260 --> 00:42:24.900 Aaron Zlotowitz: See if I can, in multiple pieces. Let me see if I can get that prompt in in a few pieces. 203 00:42:25.090 --> 00:42:25.930 Aaron Zlotowitz: Oh. 204 00:42:31.430 --> 00:42:34.313 Aaron Zlotowitz: that's still still too big. 205 00:42:36.160 --> 00:42:44.760 Aaron Zlotowitz: can I attach files? Yeah, I can attach file. You know, I'm gonna make it into a file, attach it. Give me a second. Okay, so, men, do you have any? We have a 206 00:42:45.360 --> 00:42:46.560 Aaron Zlotowitz: we have anything else. 207 00:42:53.710 --> 00:42:59.000 Mendy Palace: See a question. Is there a direct prompt that you can share that works from Om to underwriting? 208 00:42:59.970 --> 00:43:04.819 Aaron Zlotowitz: So, not exactly because the real, the, the. 209 00:43:05.030 --> 00:43:08.810 Aaron Zlotowitz: the the real details for the prompt 210 00:43:09.080 --> 00:43:15.969 Aaron Zlotowitz: would be the real deals with problems is kind of which specific bits of information you're looking for is kind of very important but 211 00:43:16.100 --> 00:43:21.640 Aaron Zlotowitz: meaning. It would just be the the overall, the structure of it would be. What is 212 00:43:21.890 --> 00:43:26.780 Aaron Zlotowitz: you know? Here I'm you're you're my underwriting. So this is the general broad strokes. 213 00:43:27.069 --> 00:43:46.340 Aaron Zlotowitz: You're I'm looking to. I'm looking. You're an underwriting assistant. I need you to go, and I need you to read this. Om, that you will be provided I need you to pull out, and then you would list the information and put it into. And because, depending on what you're doing with how you're gonna do. Further, you say I'll put into a Json structure or some other 214 00:43:46.340 --> 00:43:57.999 Aaron Zlotowitz: some other way of how you want it, how you want it! How you want it sent and how you want it returned. Because if you have building into a program, you need it structured properly. So the the code can read it. 215 00:43:58.333 --> 00:44:21.709 Aaron Zlotowitz: But that's really all there is to it, and then you have the code would kind of do the rest of the entering it into an underwriting model. I wouldn't really recommend saying, Oh, you know. Build me the underwriting model from scratch as you go, because that's going to be much more finicky, because if it messes up on the code it messes up. It has to write the whole code to build the spreadsheet, and if it messes up one word in that code the whole thing fails, so it does it, and it's 216 00:44:22.170 --> 00:44:45.540 Aaron Zlotowitz: dozens or hundreds of lines. So it's not a I wouldn't recommend to do it that way. I would recommend you build it as part of a larger program, and the prompt is, just keep it simple. Also, the the more reliable to be is the more straightforward you keep it. So you just put it in and you send it back. You put it, you put in. You have a program that just sends in the Om with the very specific details and sends it back, and then you process it separately. 217 00:44:47.910 --> 00:44:59.479 Mendy Palace: And then, just in terms of with brokers. What? What brokers are using it in terms of like lead scraping, you know, looking for for specific properties. Kind of those those rules where it's it's doing all that work automatically for them. 218 00:44:59.760 --> 00:45:09.783 Aaron Zlotowitz: So I just so I just shared the pro prompt the prompt, the oh, sorry. That was actually the that was the that was the sample report happens to be my my mistake. 219 00:45:10.230 --> 00:45:15.520 Aaron Zlotowitz: But sorry that was. Where did I put that 220 00:45:18.980 --> 00:45:20.639 Mendy Palace: If anyone wants to come on video. 221 00:45:20.640 --> 00:45:20.980 Aaron Zlotowitz: So right. 222 00:45:20.980 --> 00:45:22.280 Mendy Palace: Ask any questions. 223 00:45:22.930 --> 00:45:29.469 Aaron Zlotowitz: So that prompt was, I just sent that prompt. So that's you're just again. It's the idea of be very specific in what you want 224 00:45:29.940 --> 00:45:41.810 Aaron Zlotowitz: and what you're trying to get out and like, if you explain. If you had, if you had just hired a researcher to work for you, what would you tell them to look for? They've never worked in real estate before, but they're very good at research. 225 00:45:42.060 --> 00:45:51.620 Aaron Zlotowitz: What would you tell them? So you would go and explain to them. Maybe they worked in. They had like a little dabbled in real estate, so they know the they know the terms. But that's about it, so you would go, and you would send that to them. 226 00:45:54.540 --> 00:46:01.000 Aaron Zlotowitz: and then you would send it in as the prompt, and then again you would use the higher level model to actually sit there and do web search. And all this stuff. 227 00:46:05.654 --> 00:46:10.150 Mendy Palace: And I just saw some question, Is this gonna be recorded and sent out after the answer is, yes. 228 00:46:10.150 --> 00:46:12.460 Aaron Zlotowitz: Is being recorded, and yes, it will be sent out after. 229 00:46:15.940 --> 00:46:17.570 Aaron Zlotowitz: Okay. 230 00:46:21.080 --> 00:46:25.329 Aaron Zlotowitz: Yes. So okay. Is there any? Are there any other questions? 231 00:46:26.595 --> 00:46:40.669 Aaron Zlotowitz: Of a top quality management. Yes, we do. Yes, we do that. Yeah, as as my as my dad said we do. That is what we do? We do help companies set this set up these sort of systems. 232 00:46:44.200 --> 00:46:51.823 Aaron Zlotowitz: I think that's all. So if that's everything. Oh, scrape public records for recently traded 233 00:46:53.160 --> 00:47:10.160 Aaron Zlotowitz: isn't traded. If I'm if I'm not mistaken, traded puts that out. So if you could just have something kind of you know, pulling their Instagram or pulling their website. That should get you the information you need, at least for New York, I guess. Really, the answer is just identify the sources. 234 00:47:10.665 --> 00:47:30.800 Aaron Zlotowitz: Identify. What sources would have that? Or try asking chatgpt with like, what what sources would have this, and then see if it's able to access them, and if it is then great, you're done if not, see if you can set up some sort of like some sort of Api access, or a scraper, or something like that. 235 00:47:31.720 --> 00:47:38.779 Mendy Palace: And I see I got a message privately. There's people that want to ask questions. You can just unlock the to unlock the mute and the video. 236 00:47:38.780 --> 00:47:42.259 Aaron Zlotowitz: I will unlock the the video is unlocked. 237 00:47:42.520 --> 00:47:44.810 Aaron Zlotowitz: Let me unlock the mute. 238 00:47:48.740 --> 00:47:52.270 Aaron Zlotowitz: All right, allow unmuting is now allowed. 239 00:47:54.540 --> 00:47:57.490 Mendy Palace: Okay. So now, for all those that don't want to be shy. 240 00:48:06.180 --> 00:48:07.269 Top Quality Management: So the system would base. 241 00:48:07.270 --> 00:48:09.330 Charles: That's fine behind. So. 242 00:48:10.350 --> 00:48:13.109 Aaron Zlotowitz: I think top quality management was first.st 243 00:48:14.470 --> 00:48:28.185 Top Quality Management: The question is so basically, for an example, if you want to set up a system that would pull the information of all of the lease renewals, pull the information of follow up, send out the renewals, follow up with the tenants that didn't respond. 244 00:48:28.630 --> 00:48:37.789 Top Quality Management: All of the renewal process, which is currently so. Time consuming all of that can be set up. I guess, in certain rules that they would 245 00:48:38.180 --> 00:48:41.259 Top Quality Management: do all of that on its own sort of is that correct? 246 00:48:41.260 --> 00:49:05.460 Aaron Zlotowitz: If you, if you can define what needs to be done, and there, and anything that doesn't need to be done doesn't need to be done. So you'd have to check your with your systems, with with the systems you use. But anything that doesn't need to be done like mouse, click and drag typing. Then I mean, if it could be done using the Api then, or you're using an Api, then yeah, it can. It can be automated. 247 00:49:06.370 --> 00:49:20.023 Top Quality Management: Got it alright and like follow up on tenants. Make sure that you know we received the response to 60 days, 90 days. All of these rules. You get all of them put them into the system so they could follow the rules that 248 00:49:20.940 --> 00:49:22.409 Top Quality Management: that you put into place. 249 00:49:22.690 --> 00:49:47.139 Aaron Zlotowitz: Yeah, you could have it so it can. You could have it set that, you know, have a database of tenants that it checks against when the email comes in. If it matches on the database, then it reads the email, and if it has the proper document, puts it into a into a storage or checks off a box that oh, we got that document and then you just have the. The system has an automated thing that if the box is not checked after a certain amount of time, it automatically sends out a reminder email to them. 250 00:49:47.270 --> 00:49:52.770 Aaron Zlotowitz: The the details would have to be obviously is, depends on the system. But that conceptually. Yeah. 251 00:49:53.140 --> 00:50:00.170 Top Quality Management: Got it, and does it integrate with for an example with rep manager, or any of these management? Softwares that are out there. 252 00:50:00.390 --> 00:50:15.010 Aaron Zlotowitz: So this is what I was saying with you have to check about their Api. I'm not familiar specifically with rent manager, but you'd have to. You'd have to see what their if they have an Api, what their Api allows for. Some have like an Api, but it's really expensive. I forget which one. 253 00:50:15.440 --> 00:50:16.460 Mendy Palace: Think was that for? 254 00:50:16.460 --> 00:50:16.960 Mendy Palace: Yeah. 255 00:50:17.340 --> 00:50:22.450 Aaron Zlotowitz: No, no, yeah. Folio is a few $100, I think. But what's the the the yards. 256 00:50:22.450 --> 00:50:23.050 Top Quality Management: The one that I. 257 00:50:23.050 --> 00:50:32.900 Aaron Zlotowitz: Already has a $5,000 court charges, $5,000 a month for their Api access. So that's you. Gotta be absolutely certain that it's worth. It's worthwhile before you're gonna pull the trigger on that. 258 00:50:33.950 --> 00:50:43.460 Ira Zlotowitz: 5,000 used to be expensive number in general, but if you're thinking about, if you could replace an employee or you have 3 people on your team, and only 2. That's where it starts making sense. 259 00:50:47.170 --> 00:50:50.659 Charles: I have a question in regards to, you know, when you think of 260 00:50:50.820 --> 00:50:57.970 Charles: Chat GMT and AI, it's questions you're asking it. Through that thought through through their platform. How do I connect 261 00:50:58.080 --> 00:51:21.340 Charles: these to my current softwares like? Again you mentioned a Apis integrations, I think, like that. But what, when when it comes to Bill paying, when it comes to other other aspects of it? Is it connecting it to my Gmail account, or or Microsoft account? Are there certain? Are they connected? It's possible to connect to them that they basically scan important emails or stuff like that. Or it's just a question based platform. 262 00:51:21.660 --> 00:51:24.769 Aaron Zlotowitz: Oh, you 100%. You could connect it to your emails, you can. You can. 263 00:51:25.290 --> 00:51:32.430 Aaron Zlotowitz: Everything has to be set up with rules, because some, you know again, it can't the same way that if you don't tell your employees what to do. 264 00:51:32.430 --> 00:52:00.649 Aaron Zlotowitz: then they're not gonna be able to do anything either. So yeah, but you're you have to. You could connect it into your email. You can. You can have it checking your emails that you have it checking your emails every every minute. Say, even like, is there a new email? Oh, there's new email. Who's it from? That's all these different? Who's it from? What? Okay? So if it's this from this person, then we're gonna prompt the AI. With this question from this person prompted with that question, you could, if you're dealing with a small enough data set, you can have the AI decide who it's what it's going to even think about. 265 00:52:00.966 --> 00:52:14.590 Aaron Zlotowitz: But I wouldn't recommend that at like enterprise enterprises like, you know, next level. But like, I wouldn't recommend that at a large scale. If we're dealing with hundreds or thousands of of possibilities. Then it's already it's it's going to get overwhelmed. 266 00:52:14.650 --> 00:52:18.030 Aaron Zlotowitz: But then you would have more of a rule-based system. 267 00:52:18.120 --> 00:52:23.770 Aaron Zlotowitz: and then you would. You would have to have a rule-based system, and then kind of the AI would be built on top of that. 268 00:52:25.490 --> 00:52:26.219 Charles: Thank you. 269 00:52:32.800 --> 00:52:36.469 Isaac: Isn't there an issue with Chat gpts? 270 00:52:38.561 --> 00:52:41.860 Isaac: Where they were exposing data to strangers? 271 00:52:42.540 --> 00:52:51.249 Aaron Zlotowitz: Okay. So this was this is early in the beginning, actually. So so we covered this before just the the that if you are worried about that. 272 00:52:51.635 --> 00:52:54.220 Aaron Zlotowitz: I'm gonna share my screen again. Just you go and share 273 00:52:54.220 --> 00:52:56.609 Aaron Zlotowitz: a video that we're gonna send out after. 274 00:52:56.610 --> 00:53:15.750 Aaron Zlotowitz: Oh, yeah. Yeah. So we'll do it like that. Yeah, so you'll see there's there's a setting you can, or you can look up yourself. There's a setting in there to say, you know. Don't send my data. Don't use my data to train. This happened. This really entered the public consciousness like back when Chat Bt 1st started, when it was just. It's called Chat. By the way, the name Chat Gpt is so clunky because it was starting as a 275 00:53:15.840 --> 00:53:41.860 Aaron Zlotowitz: it was starting as a it was starting as a like a research project. It wasn't a real, it wasn't, wasn't a product. So chat gpt is just chat. Gpt is the technology behind it. That's why it's clunky like that. But the point is, it was you couldn't. There was no chat history, there was no, there was no accounts, there was nothing. So they were a research project. They let it train on every conversation that went through. They want to see what happened, hey? If we trained it on every single conversation. What would happen? 276 00:53:42.140 --> 00:53:56.798 Aaron Zlotowitz: Someone from I believe it was. Samsung actually went and used it with real company data. And then someone else asked it a question about Samsung, and it just spit that out, like, you know, because it was in its database. But it doesn't do that anymore. 277 00:53:57.270 --> 00:54:18.129 Aaron Zlotowitz: because they became. They realized that 1st of all it was messing with the model to just because anyone who said anything, some people could just say ridiculous things, and it would get built into the model. So they realized that. But also they realize that they're an actual product of people have privacy concerns. So they built the ability they put in the ability to say, Don't you don't share my data, and then it's like, if you trust them, you trust them. If you don't, you don't. 278 00:54:18.130 --> 00:54:30.769 Isaac: Right. But a few weeks ago there was a there was a court order that ordered Chat Gpt to save all chats. And there was a big, a big like a scandal, and basically even even things that you delete and they remove. They were actually saving. 279 00:54:31.140 --> 00:55:01.110 Aaron Zlotowitz: So. So that was so. So that was that really again? But that that's only other points. That's there's no different from Gmail. What if Gmail, what if a quarter Google got a court order that said all Gmail emails have to be have to be stored, even people, those are deleted, even all search results have to be. We kept. Even if they were in incognito, we have to keep them. It's just a matter of it's it's data like any other data. And you're dealing, giving it to a cloud provider. So it's the same risk as anyone else. There's no. 280 00:55:01.250 --> 00:55:14.049 Aaron Zlotowitz: it's not because it's AI. It's any different, because again, it's not going into the general model pool and then getting spit out elsewhere. It's kind of that's just a that's just a governance and and legal issue. It's nothing. It's nothing to do with the fact that it's AI. 281 00:55:15.890 --> 00:55:28.089 Isaac: No, not about the fact that it's AI. But like, I've seen these issues with things being leaked even recently, meaning even in the past few months there have been stories coming out with personal information that people put in, and then somehow got leaked out. 282 00:55:29.280 --> 00:55:45.079 Aaron Zlotowitz: Again. That's again. That's just if you don't trust open AI as a provider. That's that could be something. These stories you don't want to use them. Then go to. You could go to Google. You could go to anthropic. But there, it's not inherent to the technology. That's what I'm trying to say. 283 00:55:45.560 --> 00:55:46.400 Isaac: Understood. 284 00:55:47.990 --> 00:56:01.540 Mendy Palace: And then I just see a question in the chat the the Daily like Chat Gpt checks where you have it, you know, looking for for headlines and things like that? The question was, do you have to go back to the website to check the results? Or does it come to your email address. 285 00:56:01.540 --> 00:56:07.669 Aaron Zlotowitz: Well, if you have the app, I don't think they have email notifications. But if you have the app on your phone, it could come up as a push. Notification. 286 00:56:09.170 --> 00:56:15.019 Mendy Palace: I found that that it could send you. I mean to the email that you're signed into. But it would be a link that you 287 00:56:15.280 --> 00:56:19.950 Mendy Palace: takes you back into chat. Gpt, yeah. 288 00:56:22.210 --> 00:56:24.980 Aaron Zlotowitz: Okay, any other questions. 289 00:56:31.990 --> 00:56:53.189 Aaron Zlotowitz: Okay, if if that's everything, then thank you all for joining. And oh, there we go! We set it up to read my emails. You want to set up to read your emails. So it's not the chat of the app that'll do that. It's kind of you have to set up more of a somewhat of an automation you go into. make.com. You go into zapier. You could go into N. 8 N. 290 00:56:53.320 --> 00:57:02.050 Aaron Zlotowitz: And you can connect these. You can. They have email connectors, they have Gmail connectors. They have connect for outlook, Gmail, all of any other 291 00:57:02.240 --> 00:57:11.640 Aaron Zlotowitz: they have for Zoho also. And I'm saying that's basically all the providers that you're gonna use, or you know, you could set it up with. I think imap or any other, if you use like something more niche 292 00:57:12.006 --> 00:57:32.239 Aaron Zlotowitz: and they can check periodically at a frequency of a minute of up to like, you know even every minute check for new emails. And then you could have further steps. Oh, when an email comes in. Now, here's what you should do. So you could send it to chat to Bt. You could send it. You could do other things with it. You can just you could have send it to chat to be and have it tag that email. 293 00:57:32.380 --> 00:57:36.310 Aaron Zlotowitz: You can have it right in a draft response. There's a there's a lot you can do 294 00:57:36.803 --> 00:57:40.240 Aaron Zlotowitz: but it's really any of these, any of these systems. 295 00:57:40.900 --> 00:57:46.029 Mendy Palace: I can add, there's there's the off the shelf tools that are are very good in terms of email organization. 296 00:57:46.210 --> 00:58:04.309 Mendy Palace: And let's say, email drafting within your email where it's not necessarily pulling it out of your email to to read it. But within your email, it's reading your emails. It's it's inboxing. It's creating different folders where it's doing all that intelligently. More than you know. Apple mail. And Google is is doing currently. 297 00:58:09.020 --> 00:58:09.800 Aaron Zlotowitz: Okay 298 00:58:11.050 --> 00:58:39.989 Aaron Zlotowitz: the recording, we will be sending out. I saw someone ask that we will be sending out the recording the redial guys will be sending out the recording afterwards. If anyone here signed up signed up through like, you know. Got it, then got it from me or Mindy or my father. And they wanted to get they wanted to get that information. They want to get the recording. You just reach out to reach out to us. I think Mindy put a contact information in but yeah, so you could reach out to us and we'll send it to you. 299 00:58:44.750 --> 00:58:53.790 Aaron Zlotowitz: Okay, and unless anyone else has any questions, I think that will be it for for the evening. Thank you. Everyone for coming, and have a great night. 300 00:58:54.690 --> 00:58:55.510 Mendy Palace: Thank you.