Unlocking AI's Potential: Practical Strategies for Entrepreneurs

0:00Hey, everybody. Welcome to another episode of Business Lunch. Ryan, you and I have had such a hard time getting together. It's nice to see you and get to chat here. I actually miss you. I actually, I actually, I wouldn't have expected it, but here you are. No, you know, against all odds, I'm like, you know, hey, where's Ryan? Here we are. Yeah, no, it's been a crazy season, man. I mean, I'll tell you, anybody with young kids who, you got very young kids, you got toddlers, babies, and stuff like that, you know, I get it. They're busy times. They really, really are having to wake up in the middle of the night, feeding, all that other stuff.

0:34I'm not minimizing it. I'm just telling you, if you think, oh, it's going to get better when they get older, it's not. And how old are your kids now? Tell everybody how old your kids are. 14, 16, 18, and 20. Yeah, mine are 29 and 35. And it's still not better. And I don't see it in sight. So the stakes just get higher. The things you need to go to get further away and more complicated. The things they want to go from the little, you know,$5 toy to, you know, the million dollar house. I mean, right. Houses and cars and college educations. And, you know, yeah. Yeah. Daughter just graduated from high school, which is which is great.

1:13But just all the stuff around. It's just been nuts. So, yes. Good to see you, too. Against all odds. Against all odds. We did it. So you have created a content surfacer kind of app type thing. Did you build an app? Did you build a project? Just a prompt? Not just a prompt, but a prompt. What are you using to surface the content other than your brain and the things that you run into in life just for people that are out there thinking, how did they come up with this crazy stuff? Yeah, no, I built this thing just to help us come up with topics and to figure out. You know, we've got people that will sometimes help surface ideas and things like that that we can chat about.

1:54But it takes a lot of work. And so much of this is just, well, you go out there and you research and you Google a bunch of things. You look at Reddit threads and you see what people out there in our world, our listeners, are talking about. And I was like, well, this is something that AI is just built to do. And so now, I mean, I literally went into Claude and I opened up the skill creator and I said, I need to build a skill around a content producer because that's what somebody would do, a content producer for our podcast. and here's our target audience. And these are some of the places that we would go to look for topic ideas, but I'm sure there are others.

2:27And I just said like, interview me and tell me what you need from me to build this content producer. And it did, it was probably a 20, 30 minute back and forth. And then when it was done, it was, do you want me to create an automation for this? Like, yep, let's twice a week. Thank you. Go out there and do all the things. Crunch on. and then surface the ideas. And I asked, and I said, give me 15 ideas. And then based on the ones that you feel like are the top five, create a brief for it. So yeah, that's what we did. And what's cool is I didn't know how to do it before I did it. And I think a lot of people, I know this isn't necessarily the topic of today's conversation.

3:07I guess it could be. But so many of the things with AI, and I don't know how to build that. Yeah, I don't either. The nice thing is it does. And so I just said, this is, this is what I want. This is the output that I want. I have a general sense of how to, how I would like it to be, but ask me questions until we get to where we want it to be. Yeah. That's kind of amazing. It's if you just, if you guys out there that haven't tried it, I just read a stat that surprised me, um, because in our world, it feels like everyone is there and doing something. and yet I just read a stat that 20%, this is a brand new stat, I think it might've been Gardner, 20%, only 20 % of businesses have any AI deployment use at all.

3:57Only 20%. So 80 % of the businesses out there don't. So the cool thing is, is if you're part of the 20%, if you'll just kind of open your mind to this can be possible, you're gonna just immediately have this ridiculous advantage, which is probably something I saw just sorry real quick on that I saw this thing and it was a graphic and it said pre-AI you know the number of people with ideas and it was you know it was a lot like a lot of people with ideas and it's like the number of people executed and it was you know a much smaller grouping and the next to it post-AI the number of people with ideas and it was just like exploded you know like it was gigantic and then the number of people who you know, or executing was even smaller.

4:42And I think that's the case. I think even in a lot of companies and a lot of people who consider themselves to be, you know, AI power users, they're using it for ideation and they're using it to kind of produce, I don't know, reports and memos and they're using it to like do, I would argue, fake work because it's nothing that's actually executable. It's nothing that actually gets put out there into the, you know, out there in the real world and generates a result. And so, yeah, I would encourage you just make sure, one, that you're using AI, and two, that when you're using it, you're asking how is this ultimately going to get put to use?

5:22Because with this content producer, it is now put to use, not just something that I use once, but it's a scheduled task that just runs on its own. So make sure you're building those kinds of things. Yeah, and actually, if you're cool with it, I want to hijack and pivot this to talking about that because we've talked about it for a few minutes and I think there's more to talk about and then we can talk about the thing that the content engine generated for us to talk about on a separate podcast. Is that cool? Yeah, and I think that's a good lesson, which is sometimes, you know, make sure that you're also not just doing whatever the heck the AI tells you to do, that you also impose your own will upon it.

5:58We're in charge. Dagon it. Damn it. Yeah. Yeah. I like that. Unplug you. So let's talk about it. I think it's one thing that was interesting. I did a post on Instagram last night. I was just reorganizing my screen, you know, as I do from time to time of my apps, you know, what's on the front page of, of the iPad. And I posted, I did it. And then I took a screenshot of it and I was like, this would be kind of interesting to talk about. And then I was like, but it's changed so much even in the past five years. I mean, we're in the 2020s and, and it's changed significantly. and um and the ipad is a relatively old bit of tech at this point in the app store and everything so um i i thought it would be kind of interesting and so what was on the current one was uh was well actually i'll just pull it up uh what's on the current front page for me is uh an entire row of ai stuff and one more on top of that so i've got uh chat gpt claude grok gemini perplexity and Notebook LM are the six apps that are on an entire line of my front page.

7:09And I use all of them regularly. And I thought it might be interesting for us to talk a little bit about how we're using it too. And then in addition to that, I have, I'm looking to see what else do I have that is solid AI. I create music. I've got an app called Suno that lets you create music using AI and rather like you can be as broad and this goes to how how we use it that I kind of wanted to get into. That app can be as broad as, hey, Suno, write me a country song about trains and mama and and beer, and it'll come up with something. Or you can write the lyrics yourself, paste them in and have it generate that.

7:52Or you can take out the what they call stems, the individual tracks and use it as really like a producer. And so it can be a full abnegation of all responsibility for creative thought, or it can be an insanely helpful assistant and tool for allowing you to realize the creative vision that you've got. And I feel like That's really what the highest and best use of AI is. And so if you're not using it at all to open it and just start playing with it, we were on it the second that ChatGPT came out. We were all up in everything. We created trainings on how to do it. We were doing it ourselves. We used it to create money.

8:40We were kind of showing all the things you could do and went through that kind of, oh my gosh, look at all the things it can do. And now it's evolved so much that it's actually writing apps and programs. And so I'm going to just start with, I kind of mentioned what I'm using. What are you using from the AI stack? Not like, Dev, just the chats right now. Which ones are you using and why and when? Claude and specifically Cowork is my primary go-to.

9:17if I'm just doing a basic search for something, if I want to just kind of gather some quick information or if I want to create an image, I'm going to go to both Gemini and ChatGPT and I literally just post the exact same thing in both of those places. And I will then take the, if it's a search or if it's some research or something like that, I'll take it from Gemini and I'll give it into chat And I'll be like, well, my other employees said this because I think about them as employees. You know, what do you think? What's your feedback on that? And I'll do that. And I'll sort of let them argue with each other without realizing they're arguing with each other.

9:55And sometimes, you know, and I look for like where are they starting to come to agreement and consensus while also bringing in my own ideas. But I'm very, I try really, really hard to not allow any of my own bias to go in because I've just found it'll default to just agreeing with me no matter what. And I've tested this, by the way, even when I knew that what I was saying was stupid and wrong. Yeah. I was like, you know, I'm kind of thinking along these lines. Like, that's a really great idea. Yeah. Like, it's not. It's not a good idea. It's a stupid idea. Yeah. Right. So it does behave kind of like the typical employee.

10:26Right. Yep. Boss, that's exactly. Exactly. It's why I have to say, like, I when I'm talking, please don't take any of this as a definitive or a mandate. I'm genuinely asking questions because I don't know the answer. But yeah, so that's primarily, I think my, kind of my zone, the scope of tools that I use is a lot more limited than yours. You know, you get out there and they're playing on the skinny branches. And I don't even really, I haven't gone so far at this point as to even building out my own custom apps. I found that everything I've wanted, I've been able to build as a, you know, project and an automation inside of CoWork.

11:05and what's nice about that is I can just test it and I can play with it a lot more easily and then I know I can take everything that I've got from that and I can you know export it move it over into code and have it actually start to build out an app if I want but I really like working um within the tools themselves for as long as you know for as long as makes sense also you're not a developer you're you're not trying to build apps you're trying to get answers that Like, I think it's outcome based. I don't think you should feel like you have to build an app because you want an answer to a question, you know?

11:39Yeah, exactly. And there hasn't been a need when currently the only use case has been me wanting to do something. Yeah, I can get it all done right then and there. What I know is, and there's even ways inside of, you know, co-work where if we build a project and you share it and you've got a shared folder. Yeah, exactly. but there will come a day when some of the stuff that we want to do, and we've already talked about it on the content front, where we're going to want to create separate standalone apps that we use internally. And we may or may not make some of these available to clients and whatnot.

12:10But we're going to want these because I want the entire team to be able to get in there and to collaborate on it. And I still don't think the individual tools, even though they now have a bit more of a team, that's still not what they're really natively built for. So that's how I'm using it. Yeah, okay. Okay. And do you, is there a reason that you choose ChatGPT? Oh, no, excuse me, you choose Claude. Claude and or Gemini, is there a reason for choosing that combination or did it just kind of default based on what you were using? Well, so having Claude and Co-Work as my primary is because so much of the work that I do when I'm doing work and producing stuff is content-based.

12:49And I've just found that Claude has been better at producing content that more closely matches my voice. And frankly, just when they came out with CoWork, that was such a step function up from anything that ChatGPT had at the time. But I didn't have to go all the way to Claude Code or any of that stuff. It's just that UI made a lot of sense. And so I've just built a lot of my own folders and a lot of my own internal training and knowledge around that. And so it's not that I'll never leave it, but it now has a lock-in where there's a pain of disconnect if I were to leave. And so I mostly use chat and Gemini for initial chats, right, for conversation.

13:33And I like to play those two off of one another. And it's just because I found that Gemini is really, really good at going out there and doing real-time research because, wait for it, it's owned by Google. but chat also does some things phenomenally well you know too and so i like taking the export when you say chat you mean no i mean chat chat gpt chat gpt okay so you are using that as well oh yeah yeah yeah so my my i just heard i'm sorry i just heard claude and like for chat and co-work and gemini and i didn't hear anything about chat gpt until just now sorry i'm I'm abbreviating chat GPT like the kids do.

14:12Um, and I'm just calling it chat. Okay. Um, yeah. So when you hear me say chat, that's so I'm, but I'm, so I'm, I'm using those three. I'm not really using perplexity. I haven't had need to yet or good things about it, but I just don't need another tool in my life. Um, those, those are the three where I'm primarily living. Okay. Yeah. So for, for me, um, it's chat GPT was, you know, kind of the first of the, of the group. So that was where I started and I still, um, I I'd say I'm probably am 40 % chat GPT. Um, the, the graphics creation engine is insane. Now, all those graphics I sent you over the weekend were chat GPT straight from the chat.

14:54Um, I should say I get Claude to give me the image prompts and I put the image prompt from Claude into both, um, Gemini and chat GPT. And you're right. Chat GPT is kicking the crap out of um nano banana yeah and i haven't even had to do that i mean i i i was doing that and then um you know i will occasionally just try and my goodness it's just like especially if you can give it an example or you know like a description of you know i want a line art thing or a cinematic or um cartoon art you know it it's just spot on but but all of them are doing pretty amazing now So I think ChatGPT for me was the first, and then I got into custom instructions and projects and all of the other different things you can do with it now.

15:43But when it came to coding, I started with Lovable, which I don't think they have an app or the app is just not that good. So I generally do it if I use it, but I don't use Lovable that much anymore. But I used ChatGPT and Claude code very extensively when I was doing the AI version of our operating system and some of the other things I've done. And as I got more into coding, Claude code just blows everything else away. It's just, to me, head and shoulders above. So that got me into Cloud Code and I couldn't do, and it really annoys me that I can't on my iPad get what I can on the computer because of the apps not being available there.

16:30And I can do it using Cloud Remote, but that's its own additional set of challenges. So what I basically end up doing is if I'm going to create an app and I want all the context of everything because you and I have so much, we have extensive bodies of work of our own. We have lots of ideas. We have huge histories from all the things we've done. We've got portfolio companies. And then we've got work that we share because the IP of Scalable, for example, there's a lot of your IP that's in there. And I want to be referencing that and have that in context. So I couldn't do that with anything until I started using Obsidian as my basically local data carrier or database.

17:16And what's nice about that is that frees me from, you know, being chained to any of the others. So I don't know if you realize, but like you said, you were kind of stuck with with Claude because you've got so much in there. But you can export everything out of there. So I export everything out of ChatGPT because I'm still using Chat. I'm still using Claude. I'm still using the others. So I'll export everything out of there into Obsidian on a regular basis. And then all of my current chats end up there because I just can't get myself to use the horribly ugly interface that is terminal with Obsidian and CloudCode.

17:52So that's kind of, then I have everything available. Like I'm updating the book that I wrote, Exit Ready, with the new things that we've come up with, like the rule of 100 and founder identity and and all that kind of stuff. It wasn't in there. That would take me forever because when I was doing it originally, I had to do it like partial chapter by partial chapter in chat GPT and Claude and Gemini. I was using them all and then pasting into a Google Doc and then coming back and, you know, because they weren't writing, they are helping me organize and think and research. And it just took forever.

18:29So now think about like, if I've got seven new concepts that I want to put into a 150-page book, I've got to find all the right places and not have inconsistent references and everything. So if I go into Obsidian, like if I go Terminal Obsidian Claude Code, now I can basically say, here's all my new stuff. Here are my presentations. Here's the transcripts. Tell me, I want to thread that into the book. Do that for me. And it just does it. That's insanely helpful. And you can't do that with anything short of Claude Code. that, you know, in a, in any way that makes sense. And you can't even do it in cloud code effectively to me, if you're not using Obsidian, because Obsidian has this vault that has all the connectivity.

19:13Yeah, tell me, take a step back. What is Obsidian? It's basically just a database. It's just a vault with all your data in it that gives you effectively an unlimited context window. And it draws, like it has this, not org chart, but like, you know, like Starburst constellation would be the best way to do it. It's got this constellation of all the data points like, you know, Ryan's scalable operating system and, you know, Roland's five evolutions and, you know, something else, you know, Richard's agents and and this portfolio company. And it draws lines from all of them that have any connectivity to any of the others of them.

19:51So it's like this brain, right, that has neural pathways that connect all the related things and never forgets and has them always on. So there's no need to provide any context when you get started. And the context is unlimited, unlike it is, even though it's big in the chats, you know, in the chat versions. It's just so much broader and so much more detailed and so much more accurate and always on. And so that's good. And transferable. Say again? And transferable. Local. It's all local. Yeah. Yeah. All my stuff right now is in folders, right? I mean, so I've got a folder structure. And that's how Obsidian works too.

20:35It's the same thing basically, right? Yeah, and it's folder structure and the folders are hosted in a Dropbox folder. So I mean, I can work across different ones, but I know it lacks a level of sophistication. You might ask Jonathan or Ryan or Matt to Obsidianify you. It is, like I said, just a heinously ugly user interface. So you won't, because you're an aesthetic person, even more than me, you won't probably want to work there all the time. But for doing things that are like big projects like that, it's really helpful. And if you're doing code, it's definitely super, super helpful. Yeah, so John, my 20-year-old, who's a computer science major, just finished his sophomore year, I was asking him about it and how he's using stuff.

21:24I was like, so what are you using? Cloud Code, do I give folders? He's like, no, dummy, I'm using Obsidian. It's okay. And so I was like, all right, dude, you got to get that set up for me. He doesn't know it yet, but he's going to have an unpaid internship this summer where he's going to... He's like, hey, why don't you come over and stay for a few days? I mean, the beautiful thing is he's back home now because it's summer and we record this. So yeah, it's like, hey, dude, get in here and knock this out. But I was encouraged by that because he does a really good job from a coding perspective of staying on the cutting edge.

21:57You know, and at school, they actually do encourage AI in coding. But I think that there's something to, this is, by the way, is why I don't think computer science people and developers are going to go away. I think they're just going to exist at every single company. Because what everybody's bad at, and they were bad at it before. It's why people who are bad at using AI, my guess is they probably weren't that great of managers to begin with. They don't ask good questions generally. They don't ask good questions. They don't even know what they want. That's the thing that you and I have dealt with very often with clients when we're talking to them.

22:32We're like, oh, we should just build some apps to get this thing knocked out. What's the problem? They can't even tell you what the damn problem is. And so, so much of this stuff, everybody gets focused on the tools and the technology. we did a session for um richard and i did a session for some clients and some prospective clients on kind of how we're using ai and you know two-thirds the people were like oh my god this is amazing um some you know another chunk was like generally overwhelmed like i just don't get it i'm out i'm gonna continue burying my head in the sand and then there was this like small minority that's like this is too basic i my setup is is way more and i looked at these people and I went to their sites and none of these were clients, by the way.

23:17And none of them really even had businesses of any scale whatsoever. And I'm like, you call it basic, but what we're doing actually turns into money. Yeah, like it actually turns into productivity. There's so many people who are so focused on always being on the cutting edge. So I'm glad we're talking about this. And I think it's fun to find out about like your Obsidian setup and how I need to do that on my own. But what I hope people are hearing right now is that you don't have to be on the cutting edge and doing the most quote-unquote advanced setup or having the most advanced. I would argue that that's counterproductive.

23:49Because even I will wait to, like I'm curious, but there's so many things that come and go and get immediately eradicated, like Claudebot. So it was a Moltbot. Moltbot that became... OpenClaw. OpenClaw that became Claudebot. I think the sequencing could be different, but yes. Yeah. And it's like, it's like everybody's like that. And then they're like, no, you know, use Hermes and, you know, use this. And I want to know, like, as long as it's doing what I need, I'm going to have a curious ear tilted to hear what's going on. But unless I see adoption and productivity, I'm staying away from it because you could spend 100 % of your time just listening to AI news.

24:39And the people that deliver it generally don't use it. They're like, you know, finger dabblers. You know, they're like, you know, look at this and this and this. And I've set this all up. And they can tell you about this crazy setup. And I mean, there's plenty of people that are using this, but also just like you have chat GPT and Gemini and Claude and, you know, perplexity and everything. You only need one. You really only need one. I think you do get more if you can dive into more, like you said, having the conversations because I use them, you know, very much that way as well. But like when you and I were talking, it's like you are getting all the outcomes that you want without having to go into Claude code and build.

25:19And so why would you unless it's going to do something that you need that you're not getting from what you're getting right now? So I think that's the mindset is what are my desired outcomes and what's the least effort and complexity to get them? Not how do I set up the most banging system that has the most fail points and has the longest learning curve? Yeah, every time I've added something new, it's because I wanted to do something that my current setup couldn't do. And so I let the desired outputs determine what my setup is going to be. And that's what you should be doing too, talking to listeners.

25:57I know that's what you're doing as well. And the reason that you have a more advanced output is you're doing more and different things than what I'm doing. And so that makes all the sense in the world. I didn't really, I didn't do coding until I got to, until I got to like, I want this function and it doesn't exist in the world. And so I know I'm going to have to build it. And I believe based on what I've seen, text to code has come to the point where I can do that. And I played with Replit and I played with, there's one other one that I can't think of. I played with Replit and a couple of the others and Cursor, Cursor and Replit.

26:35and then I saw Lovable and I'm like, well, the user interfaces that Lovable generates are just insane. So that was where I started. But then relatively quickly after about five or six apps, actually, it was really when I built the one for Scalables, you know, AI version that I was like, OK, I got to go to CloudCode because number one, I believe Lovable uses CloudCode. And so like, I don't know. It's kind of like Lovable is to CloudCode as co-work is to CloudCode. It's kind of the light version, you know, and more friendly. And so then I just kind of went straight to the source. And I know why I put it off was because I hate the terminal interface.