How to use Claude in real estate: How should real estate agents use Claude?

3:42 Where AI fits in your work

Hello. who helped put this event together. An amazing event. I'm very, very excited to be speaking here. So, we are going to be going over AI in real estate. Um, I tend to look at AI in real estate a little bit different than a lot of educators and other people in the the space. Um, the title of today's class is Everything has changed forever, but not for we are going to get right into it. So, so by the end of this class my desired outcomes for you guys is to understand that AI is the beginning of a problem solving process right every I get crazy whatever you call it, right? There's these broad categories that all businesses have to perform regardless of what the actual product that you're selling is, right? And so to be able to

use an intelligence that's smarter than me and smarter than most everyone in this room, I think um the idea is that you have to have a starting point. It's just like calling a very very smart friend, right? Like you call them and you say, "Hey, I need help with this. I don't know where to start." Right? Most people don't know that they can even do that with the AI, right? The other thing that I'm really, really hopeful that you guys get out of this today is to figure out how to start using an AI operating system. What I mean by that is there's so many softwares that kind of scatter our attention whether it's Salesforce, GoHighLevel, Gmail, Google account, there's so many things in our ecosystem that contribute data to it, right? And

so the goal is to consolidate all of that information in one or several organized places so that you can point the intelligence at it to have it do your work for you. The goal isn't to go to the AI, prompt it like you would Google and have it give you a response back. That is time and effort and decision fatigue every single time you have to do that, right? And so the goal is to figure out, okay, how can I set up a system that allows me to point this super smart thing at it that's going to be aware of my brand. It's going to be aware of the things I like with my brand, don't like with my brand. Um, just as an example for that, everybody is very, very keen to tell the AI what

to do. Very few people factor in telling the AI what not to do, right? So, if you So, what can happen is you can say, hey, I need this solution and all of a sudden it's gone down a trail that you're like, I don't want any of this. It's way too much, da da da, right? And so, you want to tell the AI what not to do. What are the things that are your actual constraints? And so, we're going to get into all of that. So, I've got two rules for this class. The first one is I am an introvert with a lot to say, which means I talk very fast. I try not to. In my head, I'm very composed. I'm talking like this. I'm doing great. Give me about 45 minutes and the words per

minute ramps up exponentially. So, if you need me to stop, you want to ask a question, just raise your hand and be like, "Dude, slow it down." Um, the other thing is literally anyone can do this. Everyone in this room is more academically qualified than I am. I literally dropped out of a six-month personal training program after high school. So, the fact that you guys have real estate licenses and are doing things in real estate, you guys are more educationally competent than I am. I promise you, you can learn this.

So, the plan for today is we are going to go over where a is AI is now. And that slide has changed since this morning. Um, we're going to go over mental models. I'm going to go over some hard truths about AI, how the AI actually works. And that is important because if you know how the AI works, you know how best you can work with it, right? We're going to go over how to build systems. If there's time, I'm going to do a live demo, going to give you guys some homework, and then we will figure out what comes next. So, couple

8:56 Words to know: context, agents, loops

things that I need to go over. First is some basic vocabulary words. I'm standing right in front of the television. Um, so we've got prompt, which is a lot of people when they think prompt, they think, what do I type in the AI to get the AI to do what I want, right? Prompting can include image, audio, video. I fully anticipate at some point there will be smell-based prompts uh for like medical facilities, drug dogs, stuff like that. I fully see that coming. It's not there yet, but they use every other sense. So, um, so we've got context. Context is different than a prompt because context is the persist or is or are the persistent things relating to your business. So, as an example, if you have a color scheme, if you have a logo, if

you always avoid the word dream home, if you are uh focused on Airbnbs or you're focused on commercial, whatever kind of the details surrounding your business, that is persistent context. context window. A lot of people don't realize that AI has a limited number of characters that you can put into it before it starts to get uh exponentially Uh context engineering. So this is if you've heard of vibe coding or prompt engineering or all that kind of stuff. Context engineering is kind of what I was talking about earlier where you build your system and then you point the intelligence at the system, right? The goal isn't to build all these things outside of the intelligence. Everybody thinking about how to actually build systems that you can point this intelligence at is going to be one of

the most valuable skills you can have because like I said, every time you go and you prompt, you're you're spending time, you're spending decisions, you're spending um you're spending time that could result in an input that is wildly different from the last input you got based on the inconsistencies within your prompt. Right? And so if we have context laid out, it erodes a lot of those inconsistencies because you're pointing it at the same thing, right? So an AI agent that is a big big topic in the space right now. Um effectively an AI agent is the intelligence and it doesn't matter whether that intelligence is Claude, if it's OpenAI, ChatGPT, doesn't matter. It is an intelligence. It is an outcome that you are trying to get and it is what resources do you have to actually

get that outcome right because if you want the agent to send an email and it doesn't have access to your Gmail it it is not going to be effective right so you need to plan kind of what resources does this task or project need to talk a little bit about the frontier labs um and they are OpenAI which is ChatGPT they are Anthropic which is Claude they're xAI which is Grok and they are Google which is Gemini. All of them are slightly different. All of them at this point in history are intelligent enough to get done what you need to do. Now with that being said, the smarter the AI, the less capable technically you need to be to get useful outputs out of we're going to go over a loop or an

iteration. So a loop is three things. It is a trigger. What causes the action to take place? It is the middle stuff. What work is getting done? And it is an exit condition. How does the AI know that when it's done its work, it's done a job that's good enough to say, "Hey, I'm going to pass this to you. It's it's done, right?" And then we're going to go

18:48 Will AI take my job?

my favorite questions I get asked. Will AI take my job? I don't think so. Uh not yet. It can, you know, it can barely write an email that I trust going out to a client, let alone taking a whole job, right? So the the thing that I would advise you guys and the framework that I kind of analyze this uh in relation to is everybody's looking at will AI take my job in the wrong framework. Everybody looks at jobs as a whole. What you should be asking is what tasks from my job or what tasks that make up my job can I give to AI right now? Because if you can give away tasks like creating listing descriptions, um posting things on Zillow, posting things on Rep, whatever it is that is kind of your core

workflow, if you can give those individual tasks to the AI so that you can focus on the things that the AI does not do well yet, right? That is the goal is to figure out how our skill set can supplement what the AI is doing. And the important thing to note here is that everybody in here has a a minimum threshold of quality at this point that you can serve your clients with, right? Because the AI made that so that I I use the example of um people scamming. Uh historically to scam someone you would have to have web development credentials and you would have to know how to fake a phone number and you would have to know how to do all these things. Right? Now, they can just prompt a scam into existence. Super easy. Right? You don't

need to have the technical capacity that you once had to try to run a scam. It's the same with actually doing productive things. You don't need the technical capacity that was once required. If you want to send a, let's say, custom landing page to a client to say, "Hey, this is the images of the home in your style, in your decor, um, you know, with the key points that you like." Like if assets, and there's no time at all in liquidating those assets to your clients. So the important thing here is to understand that if you ever look down and you're like, "Huh, AI can do all of these tasks better than I can, you either need to find new tasks or start looking for new work." We're not there yet, and I don't think we're going to

get there anytime soon. But the biggest thing uh to kind of take away from this class, in my opinion, is the idea that AI can either build or destroy human connection. We've all gotten that email that we know has been written by ChatGPT, right? And everybody rolls their eyes. Some sometimes I even get emails where they forget to delete the part that's like copy and paste this into, right? And so I get those more than I would like to admit. Um, but that's and and we know what that does to the relationship, right? It's like, oh, I'm I'm on the back burner. This is AI. This is not their voice, right? when you guys get clients, they are coming to you for whatever it is that they were drawn to, right? And so we don't want to pull that

out into the AI and be like, well, now I sound more professional. Well, it's like, do you actually want that or do you want to sound like you? Do you want to be authentic? Because right now, authenticity is the only mode that we've got, right? Because the technical side is basically software at this point. So

22:16 Mental models for working with AI

be who you are. So mental model. So, um, no one dislikes theoretical teaching more than I do. Um, but I do feel like mental models are very, very important as it relates to working with AI just to have kind of a base structure on how to solve the problems, right? And so, we're going to go over some of those. So, I teach a framework called the OODA loop. Um, it was created by a fighter pilot named John Boyd. And I figured if it was right? So we want to observe, orient, decide, act. So observe. You want to identify what is true about your situation. What resources do you have to solve the problem? What is the problem?

Right? Figure out exactly what the outcome is that you're trying to get. Orient. That's where you kind of figure out okay how far can I get on the solution like what what do where do I need to go what resource all that kind of stuff you're figuring out exactly the direction to go then you need to pick one then you need to send the AI in AI is the act part in this and then once that gets done you will go back to observe and you will repeat that a lot of people believe AI is like they type So what we need to think about is how we can iterate forward and how we can use kind of these techniques to solve the problem not through one iteration but two three four what context do we need

to give it what information do we need to give it to really solve these problems at scale. So thinking in first principles we always want to think about what exactly is the outcome that we want. So, if we want to sell a house, we want to go back to well, how do we do that? We need listing appointments, right? We need different things that go into selling a house. So, we want to go fundamentally back to how do we use AI to create those structures that are going to help us?

Um, and that could be referral, that could be cold outreach email, that could be running paid ads on Meta, that could be whatever it is, whatever your traffic generation system is for your business, your lead gen, AI can help you scale it up. And I always recommend that people start with something that they're strong at. The reason for that is if you're learning AI and something you're bad at, you have no idea how to quality control the AI because you don't know if you're getting good outcomes from the thing you're bad at. If you are installing it to optimize for a strength, you can very clearly validate if it's giving you good outputs or not, right? And so if you're optimized for a strength and you learn AI through maximizing something you're already good at, you then don't have to learn it when

you go back and try to resolve things you're not so great at, right? Because you understand, okay, I've used it once to build this system. I can go back and now I can start to focus on things that I'm less proficient at. So the next best decision and this is kind of kind of talking about iteration. Um but the goal is always to take the data that you get from the AI output and then use that to make the next best decision. Sometimes that next best decision is okay it's time for me to act. Sometimes it's okay this output isn't quite good enough. I need to make these changes.

So skill gap. This is kind of what I was talking about where if you're not where you want to be with something, there is a gap in the skills that you have, right? Like if you are not getting the outcomes you want, it is a gap somewhere in the skills that you possess, right? And so understanding where that skill gap is and what you need to be working on is super duper helpful. And AI can actually help you do that. Um, and I'll get into that like how to have AI interview you if you want and it will give you a bunch of kind of useful tips on how to do that.

So, circle of competence and again this is kind of what are you good at? What do you want to get good at, right? Because what you're good at is are the things where the AI should start and then as you continue to get more proficient that circle will grow and grow and grow and grow and you can use the AI to scale more and more parts of your business. figure out how to automate the things that don't require you so that you can focus on the things where you are genuinely adding value to the company, that is where kind of the prioritization can really be scaled up with AI. Um, one thing I don't ever recommend is having an AI automate end to end something that is irreversible. So, if you have a sensitive client email to

send out, I always make sure that it stops at drafting so that a human can review it. Whether that's you, whether that's someone on your team, whether that's someone who just does a good job with customer service, right? You want to have a human in the loop somewhere on those irreversible decisions. Um on the So context is almost always better than trying to get cute with it. So, if you have emails that you're like, I this was like peak writing for me, like I crushed this email. Add it to the AI and say, "Hey, pull my writing style from this email. I I I did a really good job um you know, illustrating my point. I did a really good job using language that I like. I did a really good job X, Y, and

Z." Put the actual context in there as examples. That will be something that saves so much time because it's already created for you. You don't have to go create new data. You already have things that are very very useful. And then this is probably the biggest thing that you guys can can learn is pattern recognition, right? because all of these problems are a little bit of a different shape, but if you can get to the same starting place to solve them, they can all resolve in roughly the same roughly the same decision-m process.

Right? So, if you understand that, hey, I've got this AI and I've got this problem. Here's how I take this problem to this AI. Whether it's you're going to brainstorm solutions, whether you already have a solution in mind and you just need to piece it together, whatever that is that kind of your problem, whether it's admin, content, cold lead gen, retention, operation, finance, all of that is roughly the same start to solving the problem, right? And so if you can get to a place where you're like, you know what, all of these problems are wearing different clothing, but I can take them to the same store, that that's that's gold, right? If you can figure out how to always get to the

29:33 Three hard truths about AI

same starting place. So here are the hard truths with AI. One, AI is not going to save you. So if you are like, ah, I just need AI to do all my stuff. It it's not coming here. It's not a magic uh, you know, magic carpet that's going to take you away. So you still have to either have some level at least right now and this likely will change in the future but you still have to have some level of skill or you have to have someone on your team that has some level of skill to really get the you most useful outputs out of it. You're probably the constraint and I mean that uh personally for myself as well. So, if you are thinking like, "Oh, it'd be really cool to be able to do this

digital project, this app, this Google form, this um custom booking calendar, and then the other thing is things break. Don't ever think that just because the job is done, it's done forever. So, you always have to iterate right? And so we're at a kind of a place where the business rotates around the intelligence at this point. And so the more that we can get the intelligence locked into what we want and the more we can remove friction to that intelligence executing for us, the better.

31:25 Channels, connection and what an agent is

So how AI works. So So these are a couple things that you can do with AI. So you can do cold outreach, in-person connections, paid search or like owned audience, right? And so you I don't recommend having AI So I I don't like that. Um but it can help create scripts. It can help you um with the CRM. It can help you if you say, "Hey, this call is being recorded." it can take that transcript and go find more valuable information and add that to the CRM. So, there's a lot of things around cold calling that AI can can help you with right now. I would not recommend doing the actual calls with a voice agent yet, though. Some people do and you know, knock yourself out. But, um, and then the same with in-person

connections. You can create at your open houses, you can create QR codes that allow people to scan that QR code, take a picture of the room, and it can then put that room into their personal style, right? And so if you have one person that's like, I love antique furniture. I want, you know, gothic style. da da da in, it's like, I want clean, modern, da da da da, they each scan that QR code, they can put in their style and it will generate that room in their style so that they can see, hey, this is actually what this would look like if I lived here.

That is the full reality of it. So, the closer you can get to connecting people to their problems, the more likely you are to make human connection with them. And that is again what the whole point of this AI is, right? Because again, AI can be used to turn you into a robot. AI can be used to help you focus on the things that are building human connection. So that is what we're trying to do. Paid search. So if you have a website and you either want to run, let's say you want to run Facebook ads and you've never run Facebook ads before, AI can literally walk you through creating the ad, creating the headline, creating uh the data analysis for it, all of that. It will go through and it will it opens up capabilities

that most people don't even realize they had. And this is why I mean everyone here is the constraint, right? The intelligence is no longer the thing that's holding us back. And then if you have an owned audience, which is the most valuable audience for you guys, um, which I would recommend, this is kind of a marketing side trail, but if you don't have an email capture or a lead magnet on your website, I would highly recommend you get one because if you can start getting in front of people with newsletters, if you can start getting in front of people for um, just sending out deals, stuff like that, things that will provide value, I highly recommend it because distribution is going to be the most difficult thing in the coming years because everyone can build everything. So to get people to

care about what you're building, to care about what you're doing, it's going you're going to have to provide a lot more signal than the noise that's out there. And there's going to be a lot more noise. You guys are I I don't know if you guys are seeing it yet, but I'm getting more cold emails. I'm getting more like just cold DMs on uh my socials, all that kind of stuff. So it there's going to be more noise than ever before coming into inboxes into kind of everything. So um again and that's where authenticity is important because if you have an owned audience that have come to you because they trust and value what you're saying you want to use the AI to expand on what you're saying not change it right so the question I always ask is does

this build human connection so as an example um if you're using AI is like a that, you know, using AI to um create mass blast text messages, but you can use AI that when something comes in, it can generate a, hey, this is going back out, it's going to start that first engagement because usually that first um hey, how's it going? Doesn't have to be the most personalized thing in the world. you can just get it out to them so that as soon as it hits their inbox or as soon as it hits your inbox they're getting responded to and now they are much more interested to work with you.

if you can have a reason to stay in front of your clients and this is um the int and this is one of the biggest things about installing the intelligence into your CRM because if you have a CRM Here's a funny story. Um, so the one of the things that I use AI for is I, uh, in general do not give off warm and fuzzies when people talk to me. Um, whether that's customer service or like just, you know, consulting, whatever. I'm very good at giving, you know, direct, straightforward, honest advice, but often times people are like, I don't like that. Um, so what I did was I took um the people in my life who were the absolute best possible people at customer service. I took their insights and I created a script that makes me

nicer. So, it doesn't change who I am, right? But it gives a warm and fuzzy feeling. It kind of kind of uh smooths the edges, right, to me being like, "No, that's a terrible idea. Why would anyone do that? I shouldn't say that." So, it will say, "Hey, this seems like we could make improvements, right?" So, like it it changes how my communication is and it's it's actually made me better at being able to communicate like, "Oh, this got a really nice response. I should try to make my communication sound more like this in general." Right?

And so, you can take what your your weaknesses are and you can put them into the AI. You can take your friends, family, even people like if you have like internet people that you follow and you're like, "Oh, it would be really cool to have it sound more like that." You can use whatever influences you want to create the outputs that you're going for. They don't have to be your outputs. I say I did say that authenticity is the most important thing. That being said, I think there's always ways that you can kind of level up. Um I am trying to be nicer to people. Uh and so that is authentic. Uh and it's helpful. So, so basically an agent and this is the thing that actually gets work done, right? So, there are some things that

get work done, some things that are um useful information. So, an agent is the intelligence. It is the resources to do the task that you have set out for it to do and it is a very specific and clear outcome that you want. I also highly recommend with AI that you try to make your outcomes focused on input goals, not necessarily output goals. Right? So agents are going to be the thing that you guys are going to be handing off work to over the next uh foreseeable future until something better comes out, which I mean that could be tomorrow, but theoretically it's going to go a little bit while a little while for the agents.

40:19 What belongs in the conversation

So again, one of the things that is super duper important is what is actually in the conversation. Because if you put all your stuff in the conversation, it's going to blow the context window. And that's going to cause something called context rot, which basically means that the stuff at the front of the conversation starts to get forgotten. And so if you have a hundred files in there, it is going to pull all of those files and a lot of them are going to be things that you probably don't need, especially at each phase of the workflow. So there are, as an example, in this one, if you were writing a uh let's say listing description or a social post, you would pull in the listing at Maple Grove, you would pull in the listing at West Haven,

and you would pull in your past posts that did really well. And so it would pull those things in and then it would create the new post based on that security. So this is a fun one. Um and

42:21 Letting an agent into your email

Um and so one of the things so I get asked a lot like how I don't want to give AI access to all of my emails. What do I do? Right? So there's a couple solutions. One is using what's called an MCP server. That is a model context protocol and you can connect that right to the actual desktop app. So what that does, it's a native connection directly to the desktop app. The way the MCP works is it will like if you're searching for information in your email, it will query that and it will only pull in the emails that are relevant to your search.

That does mean that some sensitive emails may get pulled in and so you do need to be mindful of that. There are a

44:04 Room questions: which AI to use

favorite part of the um event and that is how to build the systems. Um are there any questions before we get into actually looking at systems and like how to build them? question.

What you're talking about is I need to learn more about that.

Okay.

So, is there a good place to look online, good people to look for to teach you about creating email like

for your agent,

servers and stuff like that.

Yeah. So, we're going to go over a lot of that right now. Yeah. And effectively like you can take that to AI and say like hey where are people learning about this information because AI does have training data and so if it's not in the training data it's it does web search to find it and so you can always ask the AI like where are useful conversations around this particular topic happening and it can go find it on Reddit it can go find it on X. It can go find it on like some of those sources that are not like mainstream um websites. Thank you.

This is a very I mean I feel like this is such a basic question.

What do you got?

Uh and you may be getting into it. So when you say you can go to AI, am I download am I using a specific app? Like what does that even mean?

Yeah. So that is whatever AI you currently use. So you whether you use ChatGPT, Claude, Gemini, Grok, an open source model, whatever AI you use. So I'm just using that as kind of a broad term. I could replace it with go to chatgpt.com and then

so but but how do you even choose one of those? I mean that's just a choice of I mean like do they have different positives and negatives about user capability and easeability and I mean

that's let me go break down. Yeah. So every model will be intelligent enough to solve the problems that you guys are solving right now. um it will all they will all do a fine job of it. So you can choose any of them and you're not going to be like, "Holy cow, this thing's terrible." Now, that being said, some of for Claude, Claude has been become kind of the de facto for doing work. like they have really leaned into like the knowledge work side of things. Um, and so a lot of people use Claude Cowork.

Um, and the cool thing about Claude and these AIs is they give you access to things that you did not like possibly even think you could solve before. So, as an example, I am not like an IT guy per se, like a hardware like I'm going to build my own computer type thing. However, I can take the uh intelligence and say, "Hey, I need a virtual IT guy that is going to analyze my system and make sure my processor is running, make sure my um system is running effectively." And you can do that with all of um but Claude Cowork does a really really good job of organizing, optimizing, and doing kind of basic building. So to answer your question, it kind of depends on what the primary thing your business is looking to do as

far as driving forward. Um, what I recommend honestly is take the same prompt, like pick a prompt that you think is like this is like a really good example of something I want solved. Do each uh AI, run it four times, see which one gives you the best output, right? And then go back and just say, okay, this is the one that I'm going to use for this particular thing. Right? So, the best thing that I can say and something that we've kind of uh get talked out of as adults is being creative and playful and having fun with our work, right? And so, the thing you can just move on. So, I would say play around with these things. take the same thing and do all four and give it different tests that you think would be

useful within your business and that would be that would be kind how I resolve it.

Out of curiosity, how much true variation do you see in AI from person to person? I say that like as example right if I'm looking on Facebook often times I can clearly tell somebody use AI to make a post or to create graphic. Y

is there truly a lot of variation and how how long it take to build that?

Yeah. So I would say that the the variation for most people is very little. Um because the the way the AI works is when you ask it to design something, it takes its knowledge and it's basically taking the conglomeration of knowledge that it has and it's applying those design principles. So if I say the word luxury, it's and I have a vision of what luxury means in my brain. You say the word luxury. The AI has a vision of what it mean what luxury means to it in its brain. So if we all say luxury, the AI is going to produce roughly the same amount of results. But if instead of saying luxury, I say I want clean lines, I want sharp contours, I want and I actually unbundle my language and tell it what luxury means,

you end up with like much better visuals than just kind of that hodge podge like, oh that's 100% AI, right? And so like thinking about it's just kind of kind of like we talked about with our work, right? If you can get your work down to its atomic unit, which is the task. So what task am I doing right now? It's the same with AI. If you can unbundle your language to get to a point where you are being more descriptive and I I the story kind of do that with your AI. If you don't, the AI has an idea in its head of what it wants, right? And so the more specific and the more clarity you can provide, the better the AI output is going to be targeted to what you want.

Any other questions?

Just creeped me out that you just said AI has an idea in its head.

Yeah. Yeah.

And it's computer better, worse.

I don't know.

I mean, that'll be true when the robots get here, so I'm just future dating myself. I'm just I'm just preparing for that.

Anything else?

53:10 How to build a system

Okay, this is how to build a system. Building systems are not that difficult. Um but people do want to be like oh you have to be an expert to build a system. You really don't at this point. So the things that you need to identify to build a system is you have to have a problem for which that system solves. And something that I've been kind of harping on for people is a lot of people are married to their processes, right? They're like I I love this process. It's what it is.

Processes are only valuable if they serve an outcome, right? And so the things that we need to be married to are the outcomes at this point because the tech has changed the process for so many things that we don't need to have some of the steps that a former process needed to include. Right? So think about what part of this process can I just put into the AI natively. So, like instead of going to Gmail, can I just add the Gmail connector to my AI? Instead of going and looking at my calendar, can I add that connector? Do I feel comfortable adding that connector to my AI? Some people do, some people don't.

You have to kind of feel that out on your own, like what you're comfortable use, right? And so the process needs to serve the outcome. Don't ever forget that. that is the most important thing right now. Um, and that is why I always tell people clarity of outcomes is probably one of the best skills you can have, right? Because if you understand exactly the outcome you're trying to pursue, you can steer the AI when it goes sideways and it will go sideways, right? Because um, one of the skills that I think is going to kind of rule the world in the near future is communication. Um and there are all these kind of different communication chains that we have in society now where it's um there's human to human, there's human to AI, there's AI to human and

there's AI to AI and there's any permutation or chain of those events, right? So you can have human starts, it goes to an AI that goes to another AI that goes to another AI that goes verified to the human that goes back to the AI, right? And so there's going to be these communication chains and the thing that we need to pay attention to and it's the same thing as a human to human like it's the same thing if I give um a web developer or if I give an assistant a task the communication breaks in that handoff where I am giving it with the things that they're assuming and the things that I'm assuming that's where the things break. So you have to have that as clear outcomes as you can.

You have to have an understanding of what resources you have available. And so I highly recommend everyone puts more value on the planning process of your projects than the execution at this point because if you can plan it well, the AI will go through and execute a majority of it and then you can iterate forward. So the goal is to spend a majority of your time as a human planning, let the AI execute, and then as a human, you go back and quality And so really, really take your time, plan it out, understand what resources you have. I recommend that everyone have some type of resource management system however you want to do it. So get your CRM in order, get the tech tools that you use in order. So do you use Slack?

Do you use Gmail? Do you use Claude? Do you use Cursor? Do you like what are you using from your technology stack? And once you have all those resources laid out, you can look at it and say, "Okay, I want a morning report." Uh this is an automation that I run. I want a morning report that tells me based on my emails uh over the course of yesterday and my calendar for today what is my priority based on my project management software and it will go through and it'll be like hey you've got this date coming up that you probably need to look at you've got this you've got this right and so now I don't have to spend my time planning my day and looking at my calendar and doing all that stuff it's given me what I need

to know um and again that's why having those resources is important because there have been times where I have been negligent on building out my resources and it missed something that was entirely my fault. Right? It's not the AI's fault that it didn't know that I had this meeting that I didn't put on the calendar. I didn't put that meeting in my capture area so that the AI caught it and put it on the calendar for me. That's something else that I highly recommend. One of the most simple AI agents that you can build is something called an entropy management agent. Um, and so basically entropy is the more energy you put into a system, the more chaotic the system becomes, right? And so the the more data that we get, and I very important, this is right. there's

all this data just bombarding us every day. And so if we can have a system where that data ends up in one place that we've captured and then the AI looks at what that data is, it looks at your system and it understands, hey, this needs to be logged here, right? And so if you have one place that everything goes and you have an AI agent that you can point there and say organize my life for me, it will take that from that file and it puts it where it's supposed to go. I don't have to spend the time going through my client folders and then finding the client and then finding the specific folder that it goes into. Now you may be saying, how in the world does that work if it files something in the

wrong place? Right? So the thing that's important with that particular system is that I also have a log for that system. So every time it moves a file, it also fills a log for me saying, "Hey, I moved this file to this place." So if I ever need a file and I go either to look for it manually or the AI can't find it, I go to where I think it should be and it's not there, I can go to the log and say where like where is this file and it'll say like on September 28th I moved this file from this place to this place and this is why I did it right? So, it's important to build your systems thinking about that kind of stuff. And people get overwhelmed and they're like, "Oh, I I

want to build this perfect system." Da da da da. Pick one small thing. The biggest uh kind of destructive habit that I see with AI is because it can do everything. People want to do everything with it and they end up building a bunch of half bridges that don't actually solve problems, but they spend a lot of time building those half bridges. And so now they're like, "AI sucks, right?" And for for that it it probably does, right? And so think about what is required for you to e execute a task end to end like from the the start of the task to the end of the task. What are the things that need to happen? What tools do I need? What is the exact outcome that I want? And literally you can take that to

AI and say hey this is what I need. Help me build it. So the systems are just find the problem, pick the outcome, inventory your resources, build the plan, iterate the plan, build the system, iterate the system, right? And so we always thinking with AI, build and iterate because the build is no longer the thing that takes a bunch of time. Like it used to be where I would have to hire a web designer and I would have to wait two weeks to get a proof back and then I would have to tell them all the things wrong with it and all that kind of stuff. Right now you can say create an image of this potential website that I want and then you can literally have it create the image, create the image, create the image until you're happy with

it and then you say hey this image looks great. Turn this into an HTML fully functional website. and then it will go from that uh image into a website, right? And so that historically was something that would be basically impossible to execute with that level of speed. Um it's very very real now. And so velocity and kind of as fast as you guys can move is one of another one of the moats that I think is is very real in kind of the modern economy.

1:02:26 Set up the workspace and Context Cards

All right. So the thing you want to do is first thing you want to do is prepare your workspace for the job. So what you want to do is you want to identify what files you want in there. So as an example, I always recommend building uh what's called context cards for your effectively what you need to think about is what files need to be a part of this project, what resources need to be a part of this project, right? And so for example, if you have a context card that talks about your own business, you have a context card that talks about the specific client. So as an example, you can have their design choices in there.

you can have their um kind of their choices as it relates to where do they want to live, where what what is important to them, what is not important to them, what are things that you should never say to them, what you know, all the things that relate to like really um basically a a micro CRM um file basically. And you can also pull these from your CRM. You could say, "Hey, go grab this person from uh GoHighLevel and pull them into my uh system here."

The channel. So, is this going to be a cold email? Is this going to be a LinkedIn post? Is this going to be a a warm email? Is this uh um is this a text message? So, like what way is the information actually getting distributed? Because each one of those has best practices that go along with it, right? even uh social media to the point where um Instagram, LinkedIn, YouTube, Facebook, they all have different things that perform well as it relates to captions and all that kind of stuff, right? And so if you have your channel uh card, it basically says these are the things that I want if I send a text message. I always want to end with my name and phone number. I always want to make sure that text messages are

under 200 characters. I always want to make whatever it is that is important to you for your communication style. Um home. So if there is a house in the deal, so like if there is um you know, if you're working with them and there's a specific home um in the deal, you can pull that in. And again, these can be broken out. So you can take this client away, put a new client in there, and change all the design choices so it matches to them, right? That is the goal. The goal is for every time someone sees an interaction with you to be like, "Wow, they really care, right?" Like they really care about me. They're really thinking about me. They're putting in effort to ensure that it's not just generic kind of the same repetitive stuff that they see

from everyone else, right? It's not super hard to gain an edge with this stuff, but you just have to understand how to use it to evoke emotion in people rather than use it to just get get productivity done, right? Like that's the goal. So, persona again, are you sending this to a lender? Are you sending this to a client? Are you sending this to another realtor? Um, are you sending this to a hostile realtor? Are you sending this to a friendly realtor? Right? like you can get any any way you want. And then the guardrails, what are we never supposed to do? Um, and that includes like your fair housing stuff. That includes um like, oh, I hate the word dream home. I never use cozy.

Like whatever it is that you don't want that AI to say, throw it in. Um, and then all of that together makes a production system for what you're trying to accomplish. Because once all this is done, all you have to do is point the intelligence at it and say create this asset and it will. Now,

1:06:30 Fix the system, then run again

you may need to iterate the system. And that's something that I want to talk about because a lot of people what they'll do when they get a bad output is they'll go right to that output and they'll just work through that output until they get something good and then they leave and go on to their next task. Right? But the goal is to uh as one of my company names implies build things that build things. The goal is to make the system function so that you don't have to go through all that stuff the second time. Right? So what you do after you get the output that you like, you say, "Hey, this output is amazing. I had to go through x amount of prompts to actually get something I'm happy with.

What can I change in my system so that I get this the first time?" Make the changes to your system and then start the system over. You have the thing you need right now, but start the system over to see how close the system gets to the thing that you need, right? Because the second you have that system and you can trust it to build the things you need. It's like I did it right and then you can move on to the next one. Yeah.

So you keep saying you know when this happens tell it that or say this to it. How do you say and tell those things?

Good question. So, so when you run the system, you will be in an interface typically like this. And so this will have a um this will have a chat interface. And actually, let me see here. So like GPT and you know you just kind of drill down

pretty much. Yeah, you just So what I actually did is I forgot or I have the QR code somewhere and it just didn't post on the thing. But what I did just here is I said create a QR code for the landing page uh da da da. And so I can then tell this, hey, put that QR code on the landing page once it comes back to me saying, hey, I built this QR code. It's X, Y, and Z. Rock and roll. Um, and so the system should be designed to ask you feedback at key moments where a human is needed, right? Because like I said, one of the things you never want to have happen is having to do damage control because the AI did something stupid,

right? And so you always want to put a human in the loop on irreversible decisions. So like I always say that like read the output and then you can say like go back to this like read the email that it creates. Go back then to this AI interface and say that email was really good but you can definitely tell it sounds like an AI because this this and this right and so and again you want to try to um unbundle the language as much as possible. You don't want to say like oh this sounds too robotic. What does robotic mean? Right? like what what does that mean to the AI? Well, it's not it's not warm enough. What does that mean? Right? Like am I addressing the person by name? Am I using actionable verbs in the the the text? Am I like

what specifically are the things? Cuz so many people will like put in here that doesn't look like luxury to me. And it's like, okay, well, no one knows what luxury looks like to you. So, you have to define that, right?

The biggest thing you want to spend time on is planning and quality control. Execution is effectively solved if it happens digitally at this point. um to the point where like you can say like, "Hey, I need to put in an order for t-shirts for my business. Go do that." Right? That is digital, but it's creating a physical product, right? And so, you can do all that kind of stuff. You just have to understand that the AI is capable of it. Now, and honestly, the the best tip that I give people and the thing that probably gives me the most positive feedback from people I work with is, "Hey, go try that.

That that's it. That that's like my biggest tip is hey go try that because iterating until you're good. Um