Building Smarter, Not Harder: How AI Agents Accelerate Business and Decision-Making

01I feel like that opening line for every one of them is, this changes everything. All right. Well, we've been able to do at least one episode of functional executives. Hopefully it'll drive some value and some conversation for folks and maybe at least start to get people. Yes, hopefully it was functional. where we landed with that last episode was just asking folks to be curious. Try to lean into being curious. ⁓ And I think it'll be a game changer for folks if they're curious. think if you're starting from a point of view of negativity or the fact that ⁓ fear, we, fear, well, yeah, fear and fear of the future. ⁓ One of the topics that we've discussed in the past, Jim and I have talked about is one of the kind of AI talking heads out there who's really great is Nate B. Jones. ⁓ I kinda have a man crush on that guy. He's kind of great. I'm sorry, but just plug for Nate for a second. He's just so smart. and then how does he get the volume of content and commentary that he does? So anyway, highly recommend him.

02Yeah, right. So he's got a great substack. We'll have it in our show notes. But he talked about, and you kind of pulled out a thread that we've been kind of tugging on, is building in public. Putting something out there, not hiding it. ⁓ I think the best, ⁓ freshest example of that has been ⁓ Peter with OpenClaw. ⁓ yeah. who just got hired by OpenAI yesterday and OpenAI bought OpenClaw and has made it ⁓ a public foundation now. So that was part of his ⁓ commitment to joining them was that they would keep that kind of a forever free and forever independent project. ⁓ Well, we'll see. We'll see how that, mean, like, promises are good, but I remember a lot of promises from OpenAI. I think ⁓ you could go back and fact check me, but I believe that there were supposed to be some other things and, you know, they weren't going to go public. And then now all I hear about is an IPO, which, and let me be very fair.

03⁓ There's probably going to be a need for an IPO because the amount of capital necessary for a lot of what they're trying to accomplish, especially on the infrastructure side is intense. Something we never see. ⁓ this is, you know, medium sized country GDP level dollars. Yeah. It's not even, you know, just like, yeah. I think, you know, we talked about building in public and, you know, and creating something where, you know, it is out there. If you build something, letting people see it, maybe doing something, you know, putting it on GitHub and making it something that folks have access to. If it's something that's not necessarily, ⁓ you know, company IP. ⁓ It did it made me think of ⁓ that concept of right Cortez when he brought you know, like like landed in Mexico and the know, the story is right like like sink the ships burn the ships. Yeah, right. So that there was no there's no turning back, right. And so, you know, that whole story is problematic. Different level. History says perspective, but Turning back. Yeah, yeah Yeah, it really is good.

04That's a whole I'm glad you used this analogy and That's a- That's okay. it's a it's a but right. So right, but the takeaway of like, getting past the fear of moving forward, maybe having a little bit of skin in the game by telling folks like, all right, it's out there, you can go see it and tell me what you think or extend it or, you know, is there in my like this guy did, he put something out there that brought him personal value. And then he was like, other people may get value out of this and the whole world jumped on it for the last couple of weeks. Yeah, they did. And so, you know, I think a big piece of this was that he had a problem, that he just a small problem of just, wanted to, you know, have, you know, be able to talk to his computer and have an agent do something. He wanted to send it a text message or a WhatsApp or a telegram message. And, and he built it. And and, so now it, you know, I think that was December or November of last year's, I think the storyline that I've heard. And now, yeah, it's this great big thing.

05So in the past six weeks, similarly, you've been working on something. ⁓ Well, yeah, guess similarly and dissimilarly, I'm having trouble with that word, but the same but different. Yeah, that's better. That's more my Midwest roots. That's how I'd say it. His background, had a PDF converter company. He had a software business, he built, sold, know, built it. knew how to code. The difference for me is like, ⁓ I don't, mean, I'm a Python VBA. Nobody uses VBA anymore. know, DAX mechanic, right? Cause I've been working with data for a long time. So I understand that. And there's a little bit of overlap into, Hey, this is how you design software and that spills into the finance world because we're dealing with a lot of data. Unfortunately, we don't always have IT folks like yourself that have bandwidth ⁓ to help us build some of the pipelines we need to pull in data to get answers that CEOs need questions on two weeks ago. ⁓ So yeah, that is the big difference is that I'm able to start saying things like, all right, here's my idea.

06Let's talk about it a little bit. And like we mentioned before, I think both of us are very thoughtful. I don't just jump into projects anymore, used to do that. And ⁓ now I still have a lot of hubris, I think sometimes when it comes to what I believe I can bite off and maintain, but at least I go into it being really planful. And so I did, I just planned it out. I used chat GPT, started a project. And then not only that, I asked chat GPT, course, you know, you know, Hey, how would you start? You know, you're, you know, the CTO, I'm the CEO of this company and blah, blah, blah, you know, just to give it some guardrails on the conversation and how we should continue to talk. But I also did a lot of research outside of chat GPT on what best practice was ⁓ when building some of these things. And so I drew on a lot of different areas to figure it out. But in a pretty short period of time, and I did use cursor, that was the main tool that I used because I wanted to try it out and see if it was all it was hyped up to be. It turns out, It's pretty darn good. I'm really impressed with it.

07And ⁓ it was probably three days I had a prototype that only my wife and her business use to do some reporting, but to do it all through voice. your ability to learn new products, new tools, right? I mean, you had an idea, you kind of had a ⁓ vision of what end product looked like. You had a customer for your wife's business. And so you've got multiple stores that needed to use this ⁓ product. One of the things though is you had mentioned, ⁓ you know, how did you, rather than it being like a science experiment, ⁓ or something that was just kind of theoretical research, lot of, know, option A, option B, you know, pick one from column A, pick one, you know, pick something from column B to, you know, make your lunch combo. ⁓ How did you get the tool set to move with you quickly towards accomplishing a goal? You had this kind of really specific kind of like, like kind of, you know, set of prompting where you kind of prompted its behavior to work with you. was certainly wasn't my idea, you know, I'm standing on the shoulders of other people.

08But there was some pretty good white papers out there about, you know, how to set up agents. And, you know, they're, they're not fully autonomous, but you know, agent what I've come to understand that a lot of people, there's a lot of definitions for it. But what a lot of people mean by that is just a very specific hierarchy of files. These files just happen. They're not text files. They happen to be called markdown files, but there's just a series of them. And so I developed a two-step process, just kind of learning from other people, watching some of those YouTube hype videos that we all give a heart. There's really some value in those videos. It is. We make fun of them just because I feel like that opening line for every one of them is, this changes everything. Yeah, yeah, it's totally like everything's broken now and it changes everything and all the jobs are gone. That's not true yet. ⁓ And maybe never.

09But so I used a lot of that information, you know, because I'll sit and drink my coffee in the morning and watch two or three of those videos, especially Nate's. And I always come away with like, ⁓ how do I apply this one thing? And I think that would be my biggest advice for anybody who's trying to do this is like, you don't need to like sit and watch the full 30 minute tutorial and watch them. You know, you're only going to get so much out of it because they're doing a very specific thing. They're building in public on YouTube and they're doing a very specific thing, but there's one or two nuggets that you can pull out of what they're doing. And so one of the big nuggets I pulled out was, all right, let's use two. system. Cursor has an agent built into it that works really, really quickly. They just came out with a new like 1.5 version of their, so Cursor has its own LLM. Now you can use multiple LLMs inside of Cursor, but Cursor has its own and it is speedy.

10So what I would do is, again, I started using ChatGPT and I have a long running project and I run it in sprints just like I would if I had a development team. And we go after very, very tightly scoped ⁓ projects. And what I'll do is I'll go, hey, you're at the top of every new chat. I will say, hey, you're the CTO. As a reminder, I'm the CEO. This is in the project instructions too. but I always remind it and I say, we're working on this sprint. And then from the prior chat, it gives me the exact sprint handoff. This is what we just accomplished. This is the new sprint. It has everything. It even has non-goals, right? Because these are all things that I'm learning about now in terms of software development. So it gives me this full thing. And then I become, you know, kind of its eyes and its hands. And what I'll do is, is then I go into cursor and inside of cursor, again, with the help of chat GPT a long time ago, I developed a series of agents, which are just these markdown files. So I have a front end agent. have an architect agent. I have a backend. I have a QA, a verifier.

11I've got like basically a department of developers. And I will drop in the new prompt, the new sprint that we're working on into cursor and have it go through all of those different agents in one shot for a sprint. And there is some back and forth. I mean, it breaks, doesn't always work the way I want it to work. And then I have to go back and evaluate. I'm thinking of some new ways to kind of make it think a little bit harder, but That's how I built this entire thing. And we have an operational app that we now use. get emails every day from my management team. They no longer have to be on their keyboard typing. End of day reporting is really, really time consuming for them. So now they just have to have their numbers up and the mic live. And then we get a, a nice report at the end of the day. That never would have been possible prior to what I'm doing. ⁓ what I'm working with with these tools. Yeah, and now you're extracted, like you took time. mean, but this is, know, y'all's business.

12There's value in running it well and running it with excellence and having, ⁓ you know, I think for your management team at these stores, I think there's also some value in letting them know how you're investing into the business. You're not, you know, creating ways for them to be successful. and to be efficient rather than, I think, just administrative burden for them. Because a lot of stuff has always been like, well, I hate doing this report. Well, that's part of the job. And we all just kind of just do it anyway. And consultants, it was always time sheets and breaking out your time. ⁓ Yeah, but if you could speak that into existence and have it in a structured database, then you could, you know, start to, you know, use a large language model on top of it to pull out themes and ideas. That's where the real power is. ⁓ To me, it's not in the, like, it's really cool that they can voice record it and it comes out in a structured way, but the point of the structure is that I can use it to drive my business forward even further.

13Now there's one thing, sounds like you, it looks like you might have a question, but there's one thing I definitely want to hit on before we move completely past this topic and it's the human side of things. But did you have something else or should I just jump into that? I'm gonna jump into the human piece of this, because you've built out a cool thing. Yeah, I mean, well, it's okay. And it's not perfect. And there's a lot of work left to be done on it. And I'm still trying to add value in other areas of life. And so I'm constantly weighing, okay, am I spending too much time on this developing to niche of product for one customer? I mean, it's possible. mean, like, you know, it's possible. So one of the things though, that I think will start to be a theme. And so this is me kind of like speculating a little bit, playing a little more chess is, you know, founders of businesses like myself, we tend to think had great conversation with one last night. And here's the basic type of conversation. We talk about vacation for 15 or 20 minutes.

14Then we talk about a new idea we have. Then we talk about. our family and our value system, and then we're back to another idea. I mean, it's all over the place. And that's kind of how entrepreneurs and builders and creators think. We think rapid fire. We do not mind. We were very comfortable in an environment where the conversation switches lanes frequently. As a manager, I think one of the things, and again, this is like, okay, what is human? What is AI? AI will just keep going down a path with you forever. But as a human, other humans don't necessarily go down that same, they don't want to go down a path for a while and then decide that they want to stop and then move on to something else and then move on to something else and else and else. And so I think what you can do if you're not self-aware is you can try to take that limitless pill that we talked about. try to accomplish a lot and throw all your ideas against the wall. And that's going to create a lot of burnout, I think, for a lot of teams in the future.

15When you have these people that are always thinking of the next new thing and how do I use this new tool, you got to be on the lookout for that. So my team knows, I have already told them, this is how I think we're going to try this in three months or a month or whatever it is, if it's not good. We're going to fail fast. That's a term everybody knows. We're going to fail fast and just move on. think that's a like important, like we all know it. I mean, there's little sound bites, right? That we've heard over the years that fail fast. This is a place where you really need to take that to heart because like you said, you can shotgun approach and do seven things, throw it against the wall, see what sticks. You can get overextended really fast. ⁓ and you know, in a certain way, I feel like I'm already a little overextended. So I built this thing and I've had a couple of times where people have said, Hey, this doesn't work. And I basically, because they're, you know, it's not mission critical, which is why I feel okay playing with it.

16We could always fall back to just typing these things out, but for, you know, what I'm doing, it's like, it's teaching me some things and you know, it's pushing the business forward and to your point about bias, like is there like a bias that I have? Yeah. Yeah. I did explicitly tell this thing. Like if it doesn't drive revenue, like I don't want to do the next sprint. Like we can go to the next sprint. If you believe that that next sprint is somehow, you know, going to move us closer to revenue in my business. Otherwise don't do it. And that's the idea that you're inviting these models into decision making with you. Cuz there is this example right now of all of these ⁓ MVPs that people are doing and they're posting online and they're showing, I built this, I built this. I told the cursor or Claude code to go build me this and it went and built me this. I feel like the question that doesn't get said out loud is should you have built it? Right? Right. Just because you can doesn't mean that you should. Right.

17And what you're doing at some level here is you are inviting the partner into that space with you. mean, in a way you're kind of giving it an expanded scope of influence. It's not a developer cog and a database engineer cog that you're playing with. You're asking it before it gets started to engage with you in a meaningful conversation about, well, you know my business, you know what I do. I've shared some, maybe some revenue numbers and some things of how this business is driven. Therefore making the, you know, present an argument to me of should we go forward, right? You know, can we assign or attribute value to the next feature? Absolutely. when things like here's the perfect example of this is when things break and it's happened a couple of times, but when things break, it will ask me or I will say, Hey, we got to fix this. And it'll say, ⁓ that's not the core product. If everything is, you know, if that's the only thing that's broken, then your time is best served on refining the output. Like we talked about.

18and just letting that one kind of like simmer, we'll get back to it. Which is something that we have done on the corporate IT side for years. We take high value things, we don't do releases every two minutes. We take the things that are either lower value or lower impact and we put it in a bucket, we save it for a sprint where we have time that we've got an extra eight hours that we can chunk into that sprint and we'll do it then. We don't say no to people's a lot of times, right? We try. mean, I've tried my career, right? Not now. Right. It's not, it's not no, it's no, not now later, maybe. Right. But, but the, the, the interesting thing here, I think is the idea that if you're later in your career and you're used to a very hierarchical model of management where you've you know, ⁓ or, or maybe, you know, more of maybe kind of a command and control model where you've kind of said, go do.

19And, and I think what you're missing at some level, or maybe you're just kind of like, kind of skipping the fact is, is that normally you and I and our colleagues have sat in a room together, CFO, chief marketing, you know, COO we've argued, we argued the relative value of what we need to do next. And then we assign to the CIO, Hey, this is the next project that, we're green lighting. And then we go tell somebody to go build it. And what we're saying now is that there's an opportunity when you're solopreneuring it and is that you kind of have a hive mind to engage with. To have that strategic conversation on top of just operational go build something. ⁓ a hundred percent. And they will push back. But once you've made the decision, they're all happy to get on board. That's a different, that's well, I mean, I'm just pointing it out. mean, like we're making decisions, you know, we, right.

20My little team of agents, we're making decisions together, but we're making them quick with a bias towards revenue and action and try to get six ELT members to align on something in one meeting. There's meetings before the meeting, meetings after the meeting, there's constant, you know, bringing people along type of conversations. So I hadn't even thought about that, but that shortcuts that whole process. Like once you've all like discussed it pretty quickly, you just move on. Yeah, and that phrase, you know, I'm glad you said right, this bias towards action, telling it to have bias towards action. ⁓ certain I think cursor and some of these other tools are kind of inherently implicitly kind of biased towards like, okay, I have my marching orders, I'm going to go build. And ⁓ Anthropics, Claude and ChatGPT and sometimes we'll want to engage in a conversation with you a little bit more, right? So that the bias isn't towards action. It's what about this and what about this and what about this? It goes back to that stickiness, by the way.

21Yeah, it's like burn more credits, stay here. Stay here, stay on my platform. I don't think you're necessarily wrong. mean, you're part of the product. Yes. Well, yeah, I mean, the longer you're there talking to it, the more data they have to train. Yeah, the idea of building this stuff and using these tools and you've engaged in this ⁓ role with some of the things that you've built more as, like you said, you're almost like an orchestra conductor. You have all these and you're keeping them on beat and you're moving them and you're refining what this end product is gonna look like and ⁓ you know, ⁓ which in some ways is different from some of the approaches that ⁓ I've taken recently is going with kind of trying to build requirements and then trying to really one shot it with, ⁓ you know, and either whether it's an agent swarm or whatever that's behind the scenes, but really trying to just say, I want a completed product. I want code written. I want code tested. I want code checked into GitHub. I want.

22documentation written, I want CI CD pipeline built that takes continuous integration and development and deployment and takes it all and puts it out on my server and then tells me that you're done. And now I can go look at what you know, you know, what, you built me, ⁓ which I found works if you have the vocabulary to, to, do the upfront, you know, requirements planning and, ⁓ I know I've just been seeing in the last couple of weeks, really kind of leaning into ⁓ Claude code has planning mode and trying to lean into, ⁓ if I plan, if I have a good plan, I think what people are starting to realize is ⁓ your efficiency and efficacy of the system, right? I mean, we talk about burning tokens and credits, ⁓ your efficiency goes up because it's able to kind of either have some decisions already made, have some direction already done so that it can move efficiently through the development process all the way to being done and saying, here you go, here's the nice little thing I built you. Would you like to test it now? Well, you should look at my code.

23I know you've looked at your code because that's what you do. But I mean, like, is, is it a bunch of spaghetti? Is mine probably spaghetti? I think it's a different. I'm sorry, let's step back for a second. said, yeah, what did you say? Because I don't even know the term. CI CD. No, but what is that? So that's continuous integration, continuous deployment. And the concept behind that was the fact that as soon as code is checked in, checked in, checked in, not checked in on your computer, but checked into GitHub, that there's a process that will immediately fire, integrate it, develop and deploy it. So continuously deploy to production. So ⁓ it's not another keystroke. mean, for me, that process is set up where If you commit and check in code on GitHub, I'm using render.com as a platform and it's, ⁓ there's a process that runs on GitHub, it calls up render, it pushes the code over, it builds it on render, the server does a reset, ⁓ the whole thing. And so the idea is carving those manual interventions out.

24Now there's value in sometimes having those interventions, but. ⁓ This was one of those things, CI, CD was one of those, ⁓ this is gonna speed up our ability to deliver because we're always delivering, right? Instead of turning this into like, know, ⁓ we do builds and we deploy code, we take the system down on Saturday night and, you know, once a quarter, ⁓ the idea that we could just always have it going as long as we had strong testing and stuff up front. ⁓ So normally you would do that. into a test environment and then you would have another button that pushes after you've tested it, know, from a test to a stage to a development, to a production environment. And so then the other thing that I mentioned was spaghetti, which basically is just an IT euphemism for it's a big tangled mess. And you can't tell where one noodle starts and the next one begins.

25Yeah, I think the spaghetti code, I mean, just to try to like, like give it a little bit of more color is the fact that over time you're having to add bits and pieces more and more to your product and you write shortcuts or you come up with a different name for something and you write a piece of code that's in the middle of another piece of code. so, you know, it just doesn't, it doesn't read well, right? It doesn't read linearly. And so it is, it's like a pile of spaghetti. I don't think that that's what your code that came out of the system is gonna look like, but because that's not what happened, right? That is not what happened over time. ⁓ What your code is gonna look like is something that a person would not have written, right? It's not ones and zeros, it's not illegible code. it is often very different in how it approaches its sequencing, how it ⁓ builds functions and things like that.

26And so that's one of the reasons why with a couple of the products that I'm building out right now, I'm trying to tell it as part of its deployment process is you have to write documentation. And not only do you have to write documentation, you have to rewrite the documentation. And this is actually where it's like, for me, I think it's excelling beyond what ⁓ I was able to do for years with people is most of the time we would only document what changed, right? But we wouldn't go backwards through everything and go, and this is how it affects everything else. And because that's expensive to defer documentation. And I worked with a CIO early in my career who was ex NASA. And he had worked on the space shuttle program. so, uh, the, the software that he was writing at the time for this startup in Atlanta, he had stuff, he went old school. had three inch, three ring binders on a shelf with, with code written out, numbered and labeled. And, uh, you know, and he, and he had kind of gone down this same route of like, when you make a change to line 32 of this program, there's a.

27separate, you know, kind of reference guide that tells you everything that talks to line 32 of this code, right? Because you can't have a bug happen in orbit with space shuttle. And I was like, it was just, that overhead was important and, and justifiable for that, right? We can't write. mean, it would get, it would get crazy and trying to do that in today's, you know, just building web apps. Yeah. So you build all this stuff now. It's, you know, it looks different, but it works. And then how do you sell it? I mean, do you sell, mean, like, you know, how much of this is going to be sold? How do you price it? And then, yeah, I mean, is it worth it? Or do you think it's going to be a lot of like these little like mine, which is I've got one customer. It's me. I think there's going to be the, the, the, the entrepreneurs and the folks who need point solutions just for their self or for their five stores or, and, and it, it'll be good enough and, and, and it'll be better than good enough. And then there's the, Hey, do you want to take this and commercialize it?

28⁓ I think everything that you build. You could, you now you could take it and run it through one of the other. products are back through cursor and say now commercialize this, right? Analyze this product that is now achieved a certain level of maturity after having used it for six months, we've enhanced it for six months. Now let's do version two and, and, and version two, might even change like, like what it's written in because you might have this, this desire to, right? We've talked about like, could you take this and sell it to other, entrepreneurs? Sure. ⁓ Then you take on the, should I make some changes to what it looks like intellectual property? And ⁓ is it something maintainable that I could sell ⁓ or sell away, right? And just sell the platform, not just sell licenses to people. Yeah. So I don't know, for me, I think the biggest takeaway here is that ⁓ even non-technical people, now I have run, when I say run, I have over, I've had the privilege of overseeing a few different IT shops.

29So, and largely have kind of gotten that opportunity because I know just enough to be dangerous and you know, can talk to the IT guys in a little bit deeper way than some other folks at the company could. So that's been great. But now I, who I'm not a coder, right? So, but now I can do a lot of this stuff. So that's a big takeaway for me. And I think the other big takeaway, just in terms of being practical about how we do these things. I think you need to be very thoughtful upfront about what value or what the major drivers are for you. So for me, was, I want to bias towards action and that action should create more revenue. is it. This is it. So mean, you're lanes. Yeah. Yeah. Is it valuable? Is it going to drive revenue? And I think for me, that's the big takeaway of what we just discussed is like, hey, everybody can do this, but you gotta be thoughtful and you gotta plan it out. You gotta plan it out, put some guardrails in there, don't over engineer, right?

30I think that's one of these areas where we've talked about like, do we use some of these tools because we want to use them or is it like asking the systems to, especially for the folks who doing like solopreneur stuff, like is this the right tool? Does it make sense to build this out in ⁓ one of these like ⁓ graphical make? or N8n type platforms where you're having to like, you know, really work inside their ecosystem and you have to learn their system in order to build things in their system, which has its own kind of like, ⁓ you know, conversation that we would probably have. But, or do you have a system like you're using with Cursor, I'm using with Manus, where it's writing code and, you know, we've chosen, you know, like the front end tools, whether we're using React or we're using you know, Python for the middle or whether we're using, you know, different, you I think you've been using super base for the database side. I've been using a Postgres SQL. ⁓ And so there's these choices that we make and that's super getting super in the weeds and nerdy about it.

31But ⁓ it plays out with the very last thing you said about like, you know, how do you sell this? And well, mean, if that's the goal, then we do need to think a little bit about it ⁓ because At the end, we're going to have to maybe harden it and make it more enterprise commercial grade. But for the middle tier right now, don't have to. You've got several stores. You're never going to throw a ex Twitter data feed at your system. You'll never achieve that level of... of inbound data flow. And so you don't have to build a system to handle that. You don't have to necessarily even follow the rules that apply to those types of solutions versus just building something that works with the volume that you're going to work with today. Yeah, I think if anything, you know, I think we fulfilled our goal a little bit here. So just so thanks again for spending time with me today. Yeah. And ⁓ I would say folks walking away from this should feel a little bit more like, OK, software as a service and things like that aren't completely broken. It's still I mean, I think it's still hard to do.

32It's not as mystical though as it once was. And we talked through exactly how you could use it today with the tools available in any size business, but especially like high growth and midsize companies. think they have an opportunity to lean into some of this. I agree. Well, thank you. Alright.