Outpace Everyone With AI: Why Owners Must Lead AI Adoption w/ Alan Pentz & Tonya Berenson

1, 2, 3, 4. You're listening to Owner RX with Alan Pence, the podcast for business owners who want to scale without sacrifice. Each episode delivers battle tested strategies from the Owner RX Playbook library, showing you exactly how to build systems that run without you. Here's your host, Alan Pence. All right, we are back on the Owner RX podcast post Thanksgiving. Good morning, Alan. Fake European. Thanks. Thanksgiving, it was actually lovely. We celebrate on Saturday with some friends who do a lovely spread. Gotta say, the turkey's good. It's got like giant legs because it actually runs around wild, I think, like the legs. So it's a real. You guys had turkey and it's a real turkey. It's a real turkey. I mean, seriously, it looks very different than the Butterball we see at home. It looks almost like a dinosaur. Looks like a dinosaur, but it's delicious. Okay, well, I want some photos. I will, I will send you a picture and you'll see what I mean. Like my kid was like eating the leftover drumstick. It's like this big. Looks like eating the leg of a Tyrannosaurus rex. I love it. I love it. Yeah. All right, well, what are we talking about this week? I wanted to hear a little bit about legacy infrastructure of AI because I know this is a big thing, it's a problem and all the challenges it presents. And we've been talking about that. Your course. Tell us a little bit about that. I had this thing with a client last week that was really kind of exemplified this problem of all the legacy tools we have and how they're kind of impervious to AI. So I had a customer or a client who has a services company and they have a website and it's got tons of content on it. So like lots of pages, not just a couple. And it's all on WordPress. Right. So then they were worried about SEO, search engine optimization. So there's technical stuff you need to do and there's stuff with meta tags and blah, blah, blah. So like I had. What's that? Sounds complicated already. It is complicated. Yeah. So I had for my website, which I built. Well, actually I built it first in this tool, AI tool called replit and then moved it to this. Like a lot of the vibe coded apps now are out on this server. There's like a web hosting tool called Vercel. So it's on there and I just like when I want to make changes to it, I just go into Claude code, which is hooked up to a GitHub repo repository where all the files for the website are, and that's linked to Vercel. So I tell it what I want it to do, I say, hey, change this. And it changes it. And if it looks good, I push it to GitHub, which means it's actually registered change in GitHub. GitHub is automatically linked to Vercel and it just redeploys on the web. So that's it. It's really easy. Now Claude Coda has this new feature which is amazing, which I'm teaching in my new class which we'll talk about, called Skills. And really what it is is a set of instructions for how to do any function. So like, it could be anything from how do I do a commit to GitHub, how do I implement like for instance, claude code when they originally just created markdown files. Now I can do PDFs and Word files. So like, creating a word file is a skill that it downloads and uses. I didn't know it could do that. Yeah, it can if you give it the skill. Right, so you can use this in the chatbot interface too. If you go under Settings, there's like a place where you, if you go down, you'll see skills and you can add them. But it can also like be. Oh, how we do client reports, like the format of client reports could be a skill. So every time you say client report, it'll invoke this skill and do it the same way. So it's a way to package like a single way of doing something and share it with other people. So I went on, I had Claude code do A search of GitHub has these repos. Some of them are open and people have them. Like Anthropic has a repository repo and you can download stuff from them that they produce and you can use them. It's open source. I said, well, go look at the most popular open source SEO repos. I did a bunch of research, we looked at a few and basically there are experts on SEO who developed this and put it up on GitHub. And now you can download it and basically get an expert SEO report done. And because it can connect to all these things, it uses connectors that cloud code itself calls up through these model context protocol MCP connectors. So I'm just sitting there, it's like, okay, we're going to connect to Google Lighthouse, we're going to do this for that. I'm like, great, okay, so basically it's using all these online tools and it went through my website and gave Me, like a full report. And then I was like, I don't want to read this. Yeah, I was like, I don't want to read this. Just make the changes. It was too long. So, you know, I glanced through a glance. Yeah, it was really long. Too comprehensive. It's like 15 pages or something, so. And it gave me a score. And I was like. It said, oh, you're missing these metadata, blah, blah, blah. You know, I was also trying to get my newsletter that I publish to be on my website. So it gave me the SEO rather than Beehive, who I used to send it. So that was part of it. And it, like, had to put in some canonical link or something like that, and it did it all. And it was like two minutes you were going, you were showing. You were doing this with your client, like, to show them, like, what's possible. So I did it for myself. Then I'm talking to them and they're like, oh, well, we gotta improve the SEO. I guess they gotten some pings on it, but they get a lot of organic traffic. And so I'm like, well, what system are you on? They're like WordPress. And I'm sure maybe there's a plugin for WordPress now that can do this. I don't know, but. But it just got me thinking, like, oh, they can't do what I just did. So I did. I gave them. I had the report. I could run the report, though. So I ran the report. They got a 58 out of 100, so not too good. But they had, like, all these recommendations, right? Metadata, blah, blah, blah. This thing doesn't do this. Right? Well, they're gonna have to go hire a contractor to go one by one through those things and change it. So first of all, weeks of, like, finding someone. Some people have someone around money, and then, like, thousands of dollars to have this person do something that Claude Code did in two seconds or a minute. So it just really showed me, like, if your tools aren't AI enabled, you're just stuck in this molasses of the former world, right? And then if you're free, you're out there with Claude code, you're free, you're flying. It's like Jurassic park, right? Without quad, you're in Jurassic Park. Jurassic Park's kind of cool, but I don't know. But it's old. Yeah, it's old. Yeah, that's right. That's right. So. So it's just like a little example. And that happens everywhere, right? Where people are doing these cumbersome expensive, time consuming things that AI can do in seconds. And it's really just about getting the tools in place eventually. Like we had this with one of the business owner groups we work with and they're like, oh, we're going to hire this person for a few thousand dollars to rebuild the website. I'm like, I could do it in 10 minutes and repl or clock. Now I do clock it and they're like, who's going to maintain it? Who's going to do this? I'm like, I don't know, man. You could probably, you probably train Tanya how to do that 10 minutes. So. But anyway, so people are still doing that. I don't get it. But it was a really good lesson. And I think the only thing I can say kind of leads into my class thing is, you know, I did this, this class in the fall for small business owners and teaching them about AI and it was really an interesting process, like learning about interesting, sometimes very frustrating. And I kind of got like a picture of what you can do and what you can't do with most people. And given legacy infrastructure and all these constraints people have. So I've kind of redone the class in two sections and the first section I think is really what all owners need. It's all about just like learning the basics of like what AI can do. So you're an informed consumer when you like a primer for business owners, like if you don't do it, you should know. Right. You need to know enough to like tell your people what to do. Right. Because for some reason people hate AI. I mean it's crazy. Like it's. I've never seen people like, I don't know, I guess people were like that about the Internet a little bit. Are they hating or they are afraid of it? Well, I think they are afraid of it. So they hate it. They. But there's a lot of hate about it, let's put it that way. So I think what I've said is that people aren't. Very few companies are going to have somebody that's going to drive AI adoption. So you're going to have to do it as an owner. Yeah. So I say to them, look, you need to know enough, you need to get your hands dirty enough to know what's going on here. Again, the web was different. The web wasn't. It is technically, but it isn't really a general purpose technology. It's like, hey, how do I publish information in a free way? That's what the Internet was. And I think AI is much broader use cases than that. Now, obviously that led to E commerce and other things like that. But I think that AI really is limited by the ability of humans to think through what to do. And so I don't think the web really was. The web was kind of like a few use cases, right? So I think it's a harder technology to adopt as a result. And I think it requires more from the owner to see where it can be used. Because people are gonna go, try to waste your money, right? They're gonna go like, oh, let's hire someone to do this, let's do that. Cause I don't wanna deal with AI, right? Cause that means I have to go do something. So you've gotta head them off at the pass on that. That's sort of my. That's where I'm coming from. So I really restructured the class to be like. And I wrote about it in the newsletter today. You can go to ownerrx.com newsletter if you'd like to sign up for it. And I was saying that there are like six levels, right? And so it goes. Level one is just like using chatbots better, right? So using your ChatGPT. Yeah. But also understanding, like what Claude is good for, like, it's better for writing encoding. Gemini I like for planning. So we kind of go through some of that and then how to use some of the basic functions, connectors, how to use projects and GPTs and things like that. Then you get to this next level, which is building very basic AI workflows. So that means I can chain together a group of GPTs that each one takes data from the other and does something. And then I can learn a little bit about. That's where we bring in connectors or like what I was talking about earlier, the mcp, the model context protocol that allows two programs or agents to talk to each other, or using something like Zapier in an automation software. And then we get to the level three, to me is like your ability to build basic websites and apps. So that's what we, you know, I was talking about with using cloud code or replit or Lovable or Bolt to build first a basic website, but then also an app. Like you can have an interactive web app with kind of a database if you want. And then that to me is the really basic stuff that owners. So it kind of gives them an introduction to web apps, to agents, to chatbots. That's sort of like the base level stuff. And then we look a little bit about, like, what are the other tools out there, AI enabled tools that you should really be looking at, like AI business development reps. Those are really good AI customer service programs and agents and stuff like that. So that's really like the first level that I'm going to teach in the class and then I think I'm going to do an advanced class where once you master that, we go. Because I tried to do some of this in the first class and it really just got to the point where some people just couldn't keep up with it. So it was like I was teaching two classes at once. But it's really about using coding agents. And so that's when you stop using the chatbot and you start using the API connection, which is really not that hard. So you have to do it in this GitHub codespace thing that you downloaded the web. Yeah, yeah. And then, you know, use or use it in your terminal, which is what I do, which is a little less secure but easier to use. And you load in cloud code or Codex is OpenAI's version or Gemini, you know, you can do any of those. But wait, that's the advanced part. All of what you're saying. So that's the beginning. It's like really just. Yeah, like learning coding agents. So like every day I'm in Claude code, just like telling it what to do. I don't use other programs really. And it can manipulate, it manipulates files on my computer, creates them, deletes them, moves them around, and then I back it all up in GitHub. So you kind of learn that and then we go to the next level of like sort of the agent process of that. So like we were in the earlier course, we're kind of just doing GPTs, which are little custom prompts you save. Right. And then we learned how to chain them together. Now we go to the more sophisticated agents in the second part, which is like that's where the trends are going now, which is, you know, Google's got their Vertex Studio, there's some open source sort of agent, they call them agent development kits. But it's like, hey, you go tell Claude code to build an agent using OpenAI's agent development kit. Or there's a company called LangChain that developed like really good infrastructure for how agents work, how they pass messages to each other. And so it's better than using a GPT, right? Because like if a GPT fails, there's nothing, there's no backup, there's no way to really tell. LangChain has all these monitoring things and it's just better. So you learn about that. And then we talk a little bit about using like antagonist agents. So a really good. That's how I think of the antagonist are the work checkers. Yes, exactly. And so they challenge the main agent and usually make them better. Right. And that's where you're going to get loops where like an antagonist agent, you know, looks at the work another agent did and then sends a message back to the agent to redo it. So there's a little loop there. So that gets more sophisticated. Right. So you need these like LangChain kind of things or you can like. I also have like user advocacy agents. So like whatever your ideal customer is, you kind of build a profile of them and then make an agent that reacts to content based on their profile. Right. So I do this with all my LinkedIn posts and stuff. I have ideas, I write out the basic idea, then I put it through an agent that's been trained on all my LinkedIn posts and what works and what doesn't. Then it has an AI ism checker. It has a user advocacy agent checker to small business owners care about this. And then it gives me my LinkedIn post based on my idea. And early I write out part of it. Then every couple of weeks I'll download all the metrics from what worked and didn't on the LinkedIn post, feed it back to it to a analytics agent, and it'll look at it and be like, we, I think that based on this, you should do more of this and less of that. And then if I approve that, then it'll update the agent. So the agent's like continuously improving. That's gonna be so helpful. Like, what a game changer, right? Yeah, yeah, it's amazing. So we teach you a little bit about that. And then like finally we go into this thing where I was talking earlier about building tools that build agents and apps. So that's where I think the future of work is. Humans building tools that do the work, or building tools that monitor the tools that do the work. You know, it's like this very meta place. So it's like, it's like going from working in the business, working on work and working, you know, working in the work and then working on the work. Right. And that's where I think we're going. So the last part is kind of like learning about this Claude skills and plugins and how those can really accelerate. Like we had that SEO example. Like now I have an SEO specialist that I can run. And so that's the final part of the class. So that's much more advanced. Right. I mean I think most people aren't going to be able to do that. So I've actually been contemplating whether level four through six, that advanced part should be like for the AI advocate on the owner's staff, if they have one. I mean there's some owners that are super into it. Yeah, of course. Right, so we'll do. You know, I talked through some of those concepts in the intro course. So you know, you've got a smattering of them, but we don't. We're not actually going to build like Jane in the first course. So. But that's sort of been my evolution, like how I went. I went through those six levels and I think that now you could probably go through those six levels in a month or two versus the six, eight months that I went to because most of this stuff didn't even, I'm like Google Vertex didn't even exist and like Claude code was terrible. Right. So it was like you couldn't do it. Right. Like just six months ago, eight months ago. Oh yeah, I would say even. I mean Claude Skills came out last month or a few weeks ago. Like three weeks ago. Right. So that significantly changed the game. Like that blew my mind actually. Because you think about it now, you could go take the. Like A programmer has 30 years of experience. Or actually someone said this in a blog I read, I could take good practices in programming book, feed it into Gemini has really big context windows. So I can take a whole book or I could have another one work chapter by chapter and turn every best practice into a skill and then I can download them all and then I can program like that book says because it'll do all the stuff the book says or, or a 30 year programmer. It's like this is how you do this and so you do that. You know, you're like just create a whole repository of programming skills and then bam, I can adopt them. So that's cool. That, that's cool. Insane. That's insane. Like, I just can't believe that it's possible. Yeah, and it's like three weeks old and I'm already like doing the SEO thing and I'm you know, like piece by piece I'm like, oh, let me go research skills rather than like I had this thing where it's like planning out because I'm going to be taking over development of the OwnerX app from the developers and as I go through looking at adding new features and stuff, I Downloaded a. Like, what are the best practices for how you plan out a feature using cloud code? Right. And now I've got somebody who's a lot smarter and more experienced than I, has like built how to do it. I don't have to figure it out. So it's, It's. That's like, to me, this is mind blowing. I mean, yeah, this is like the knowledge. Yeah, it's. It's really unbelievable. So I think that if you're listening to this and you're thinking about it, go to. We'll have the advertisement for the class up on the website this week, I think, and then we're just building out the back end database for it. So you can register. Register now. Yeah. Or you can email me alanrx.com and yeah, we'll be starting that in January. But I do think, like, whatever, take my class, take somebody else's. I think mine's good because it's designed for business owners. So it's supposed to be, I'm a business owner, so I'll show you how I use it. But you gotta go do something. Cause this is just like too crazy. Like, I. I find something every day. You should take yours. What's up? You take mine. Just take yours. I'll say it for you. There you go. But do something. Cause it's insanity. Like, what you can do now is just beyond. Beyond what I even imagined even a couple months ago. So. Yeah, so I think that's it for this week. All right, well, thank you. Thank you. Yeah, thanks. We'll talk soon. You've been listening to Owner RX with Alan Pence. Want to apply what you just Learned? Try our AI business advisor@ownerrx.com. it has all the insights and lessons from this podcast, plus hundreds of playbooks ready to solve your specific business challenges. Owner rx, stop being the bottleneck. Start being the owner.