going to get this, you know, have hypey kind of word quantum jump, but you're going to get this giant leap of value out of something. And it's not just the fact that these technologies exist. It's going to be, you know, men and women like you and me who have 20, 25 years of industry experience. We can then take these tools and really wield them with just a an experience that I don't think really can be matched. feel like that opening line for every one of them is, is...

How are you, Veer?

⁓ you know, busy, just like you, everybody else that's probably going to listen to us.

Yeah. And we have a lot to get through today on this episode as well. So we want to make sure to get there. so what, but you say you're busy. What are you?

I've got to catch a flight. to get to Philly by this afternoon and helping out a new client up there as a fractional operator for them and doing some coaching for them as they're tracking towards their next transaction in 18, 24 months.

So private equity stuff. ⁓ Yeah. It the world go round at this point. It's everywhere. Yeah, no, I've got, ⁓ got largely the same type of things going on, but you know, I play part-time CFO for my wife and her business. She has a little multi-unit business that's growing pretty rapidly. She just signed, ⁓ she just signed literally yesterday.

The world around, but you know.

the final copy of her, fully executed copy of her new lease. So that's wonderful, but that means I have to go get the money now. So I'm in full on capital raise mode. drove to downtown Denver yesterday and sat with a banker friend and just super nice guy. And I'm trying to structure something. So that's going to eat up a bunch of time. And then later today I've got a, I've got a client meeting as well on software as a service.

All right.

company, or actually it's a membership based company, but there is a little bit of a software as a service component to it. But it's a digital product. Yeah, that's that's that's the better way of positioning it. They kind of have both. So so yeah, full day and then we're doing this podcast in middle of it.

Like a digital product. And we're doing this. one of the things we're talking about today, ⁓ a lot of topics, but one of the things is talking about ⁓ the intersection of the skills and the abilities that we have from the verticals that we've worked in and the new skills and abilities that we can pick up by using some of these tools and why would we do it and what's changed. And you worked on a project, we've talked about where you sat down. spun up a SaaS product all by yourself, no developers, no CIO helping you out.

I don't code is as a matter of fact, I explicitly told the model or you know, the AI overlord that I was, you know, choosing to work with that day. I told it, I don't code. That's not my job. So, ⁓ you know, I've never really done it professionally. The only thing I could say that I've done professionally is I I've, you know, written VBA and not very well. VBA mechanic, not VBA engineer, if that makes sense. So yeah, no, I've never shipped a product. Don't really understand software development, life cycle. mean, all these terms are relatively new to me. I have training as a CFO. I allocate capital, I mitigate risk, I size up risk ⁓ before I mitigate it. And yeah, mean, like all of those things are just like foreign. They were foreign concepts to me, but ⁓ I'm a lifelong learner truly. And I'm very intensely interested in technology.

And so we're talking about like using these tools and the constraint prior was skill, having gone either from an educational side or from an apprenticeship style side to having skill so that you can go build something. And that's not the barrier anymore. And it's not that you have to go get a book.

It's not, but I mean, to be fair, that's not really the barrier in finance. If you really want to look at it that way. mean, if you have enough willingness to dive in for a weekend, there are tools out there that help me ship an Excel model a whole lot quicker than I used to. Of course, I got 25 years of experience that allows me to ask it the right questions, just like a developer would have probably done what I did in... with this software as a service product that I built in two hours versus four hours or something like that.

Well, I mean, but that's the key piece, right? You asked it, and it may not even be that it was the right questions. You asked it questions. You were curious.

Right. Well, I asked it questions. I actually started a whole chat thread that I knew I wouldn't like it. It wouldn't blow up really big because I kept it really constrained just for definitions.

Like technical definitions, like what's an IDE.

Like what is an IDE? How does an API actually work? You know, what does ⁓ CLI mean? You know, all these things. Now I did this several, I mean like over a year ago now. So I basically for more than a year since this AI came out and all the talk around, you know, how it could help you accelerate. I basically taught myself about software. development in depth. now I know I used to know enough to be dangerous. Now I know enough to be a little more dangerous, I guess. I don't know. I don't know. I have a better way of saying that.

No, but like you've taken that knowledge and then these tools have increased inability from being able to just answer your questions and educate you to now being able to take on personas, know, job descriptions and interact with you ⁓ towards a goal. And you had some clarity of something that you were trying to build. You had an idea of a finished widget that was going to pop out the end of the machine.

Yeah, I would say the constraint around all of this isn't, ⁓ do you know how to code? It's not the, the, the skill of coding is not the constraint. It's really more about, and we talked a bit about this yesterday. think product market fit and also what do I, what problem? So like what problem do I, am I trying to solve? Does the market really like the problem? And then can I go about structuring the product in a way that logically makes sense. That's it.

And that means that there's some intellectual curiosity there. You're able to have that like forethought of product market fit.

Yeah, I think that's the thing that I see missing most places too. And I'm not trying to, I don't want to go off on a tangent, but I think the thing that's missing the most is intellectual curiosity. don't know if that's fear. I don't know if that's, you know, just a weariness. Like I'm tired at the end of the day. It's like, because these are, this is like the volume of the things that are changing is, is pretty, I mean, it's significant. In finance language, would say it is material, the changes that we're starting to see. that's, mean, if you work 12, 14 hours a day, because you work in a pressure cooker of a private equity environment or fortune 500, that's a lot. I mean, that's a lot of extra, extra, you you just need to get the job done. But I just try to encourage people all the time that you got to dig in.

Well, people have been listening to us and we should probably tell them who we are and admit, you know, kind of like who are these two guys that are just kind of waxing a little bit philosophical on some projects that they're working on. ⁓ So ⁓ I'm Veer. I'm a CIO. I've been a consultant for a number of years. I work as a ⁓ operating partner for private equity companies, ⁓ helping them align their ⁓ overall investment thesis and strategy against technology. So this tech part is something that I actually do on a regular basis.

Yeah, I'm Jim. And ⁓ I'm a CFO. ⁓ So finance through and through by training by application. I'm a practitioner, not necessarily a long history in consulting, but that is what I do now. I ⁓ bring that 20 years of 25 years of allocating capital managing risk. evaluating projects and just making sure overall that the balance sheet is always protected. bring that now to clients that a lot of times are in like a little bit of a chaotic situation. So private equity companies, pre post acquisition, businesses that are scaling. I have a deep background in growth companies. So I know intimately that scale starts to break a lot of things. so try to put a little bit of, try to play a little chess with them and try to get them to see, you know, four or five moves down the road, what this will be if they really are committed to the growth of the company. And so those are the, those are the type of things that I do.

And so you're going to hear us on functional executives talking about how we're taking new technology, new trends in the market, translating them away from hype into something that drives value in these companies and these clients of ours. And you're going to also hear about how we're using it to drive value for ourselves in both professional development and personal development. We're going to jump into some stuff where we talk about today. What does it mean to be building something or to be in the middle of this ⁓ thing that you keep hearing called vibe coding? ⁓ I don't feel like I don't feel like ⁓ I hope that doesn't stick. I hope it's not coding with vibes.

No, I don't. I don't think it will. As a matter of fact, and I'm not even going to try to remember because I won't, but I think there's already a new term that

agentic or something like that.

Yeah, there's something else that people are kind of bandying about now. So, I, know, we're not, neither one of us are developers, you know, so we're just trying to make sense of, of this stuff. I would say though, like, and you mentioned it earlier, but there is a very, this is new terms that I've learned by the way. ⁓ but we have very specific non goals here. And you know, that is, we do not, we like, We don't want to be a part of the hype. We want to try to figure out, all this stuff's out there. How do we apply it? know, with our cup of coffee in a corner office at whatever XYZ company we're in right now, how do I actually use this to do something meaningful to add value to that company? And so that's really the non-goal is don't. We don't want to, we're not saying everything broke and you know, AI is here to take all our jobs. That's not, I don't even think that's true. I don't know about you.

I think what we're going to show people is that you are going to get this, you know, a hypey kind of word, quantum jump, but you're going to get this giant leap of value out of something. And it's not just the fact that these technologies exist. It's going to be, you know, men and women like you and me who have 20, 25 years of industry experience. can then... take these tools and really wield them with just ⁓ an experience that I don't think really can be matched. And so like in this case, we were talking about that you spent a weekend building, you were coding, but you said you weren't a coder and ⁓ you ⁓ did some testing, you started developing, building, deploying, using these different tools, putting something out there. You were shipping something.

Yeah, no, 100%. And I am just taking a note because there is something that I definitely want to make sure I'm going to leave it right there. But I have a question for you after we make it through this that I think is really a critical question, especially for your PE and CIO brethren out there for all of our folks. yeah, no, listen, it's really simple. Every one of these multi-unit companies out there do end of day reports or shift change reports, like every one of them. I mean, like that's the thing. And, there's a lot of, there are a lot of software solutions out there that do this kind of shift handover, you know, end of day reporting, you know, functionality or capabilities like within you know, the, full suite solution. It's typically only available to like enterprise businesses though. I mean, you have to be of a certain size to, to have it. Otherwise you're just, you know, sending a team's message or an email, um, trying to put some structure around it.

Which, so here is a definite opportunity to do something better, to improve operational efficiency from, I'm just going to send an email and you're going to get what you get and maybe I can do it, maybe I can't. ⁓ So much variability in the actual, both the output and the tracking of that output that every multi-unit franchise location group needs.

Yeah. Franchise or, you know, even corporate. mean, but yes, the point is real. It's like there are small, smaller multi-unit operations that just don't have access to some of these big tools. And with my wife's business, which I talked about earlier, we had the same problem. ⁓ So it started off, I was just going to do a teams integration and I was going to make it really kind of like, I don't know, Klugey. You know, just kind of duct tape some things together, power automate this and do a little bit of this in the background. And okay. but Chad GPT keeps telling me that I over engineer things. That's a legit, but I've got like Chad GPT down on like be super candid and tell me what I don't want to hear mode as best I can.

Do you have to lie down on the couch while it's talking to you?

Sometimes it feels like it. Yes. Sometimes it feels like, ⁓ wow, that's insightful. Wonder what part of the model found that.

that you have to go call your mom now and

And we might need to stop so I can do that. So yeah, but it started off as something that was kind of like a, okay, this is a nice to have. It's a little bit of a toy and I, you know, should I be spending time on this? But I over engineer chat. GBT tells me that all the time. So I made it and it just kept getting better. And now it's a real, now it's like a real thing now, like other people could use it.

Right. And you did it on your own using a set of tools. so for this, at least to help people understand, what are these tools? Because sometimes people just hear Chat GPT, and they're like, ⁓ well, it's that and Copilot.

Yeah. So I used a variety of tools. Again, I'm really curious about what's available. I am, you know, I read sub stack Reddit and the wall street journal and YouTube. Honestly, I love the wall street journal, but they all have their different places. Like the real, the really like bleeding edge type of stuff is out there on sub stack and, and then also YouTube, you just have to sift through. It's hard to find adults in the room like, like we're trying to be here that are that are telling you actual applications. You know, it's usually, you know, a little younger demographic that are hyping things up. So I dove into a mix of those things, filtered what I thought, and then I landed on, like to use cursor, you introduced me to mannus, which is actually, I can get to that, but that significantly sped things up. And then I use chat GPT. So in the first round, was just chat GPT and cursor. That's it. And then I set all that up the way that it needed to be set up based on some of these sub stack, ⁓ actually anthropic had some really good, ⁓ blogs or blog posts or white papers. So I read those and that is what gave me this, this kind of workflow. And then I've iterated on it, like, you know, probably five times on like how I'll, you know, continue to improve this with different tools. Cause some of them are definitely easier for people like me that don't know code or really IT in depth.

But you know, and the point of asking that is just the fact that like this is a legitimate piece of ⁓ software, software ⁓ as a service that you've been able to build, deploy, put out there and are actually using. it looks great. you know, and we've had this conversation that there's, I feel like there's a commercial viability to this that, you know, that should be pursued. ⁓

You also told me, mean, like just full disclosure that it could probably be destroyed in a weekend by some behemoth enterprise company.

You know, it could and there's... ⁓ I work in healthcare, a big chunk of my career has been spent in that space. ⁓ it's a very cautionary tale that we're seeing right now in the news where ⁓ these AI-powered summarization and or ⁓ tools that would do some diagnostic work in the healthcare space ⁓ big company sat back, lot of let a lot of entrepreneurs do the product market fit, test a bunch of things out, they, they observed what's was sticking. A couple of people have gotten lucky and got acquired. But almost everybody else just got absolutely just jumped over because they were able to platform. You know, if you run all the software inside of a hospital, they're not going to add another piece of software, they just want another tab inside their software that does whatever it is that you know, was being done. so, yeah, but I do think that we're in a time where there's ⁓ viability for something like this. Maybe you luck out and somebody buys you and maybe you luck out and you have a little, you know, recurring revenue for a while. But here you saw a business problem. You built something to solve it. You're you're deriving value from it today and potentially you could help other people derive value from it. And you did it again in that spare marginal time. ⁓ Maybe instead of reading the next, you know, great American novel or whatever that you could have spent time, ⁓ you know, in the evenings doing you did this.

I will say though, family is super important to me. So I always make time for the kids soccer and you know, whatever that is. But yeah, there's a trade off and all those extra things. So I'm not reading whatever the latest business book is. I'm not reading that, you know, I'm not listening to a ton of podcasts that I really like to listen to because I'm spending most of my time with ambient music on you know, sitting in a dark room and going, all right, like, okay, that fits there. And how do I do this? And so one of the things that I did that I thought was interesting is, and one of the things that I have incorporated into my workflow is finding some of these bleeding edge, you know, thought leaders on, on AI and agent take AI and all this, the newer things I've been taking like, their seminal white paper or their, you know, like whatever it was that they've put out there. I've been dropping that and that's what I did on this one. And I dropped that into chat GPT or in a model and I go, I want to do that, but I don't know how to do it. And that's what got me set up down this path of, okay, chat GPT is going to be up on this window and I am going to tell chat GPT to be my chief technology officer, CTO. So there you go. That's your role. I'm also going to tell Chad GPT, I'm the CEO. set the vision product, you know, ⁓ product vision. I don't know. And you follow my lead. I know where the problems are. I know what to build. I've got the money. You work for me. And then using some of this other kind of best in class ways of working with AI and building agents. I had Chad GPT with that white paper build out a series of agents that I then put into cursor. And so what my development process was for a long time was, okay, here's the feature I need to build. Here's what I need to have. I need you to write out a complete cursor prompt in a spec or not, not really a spec, but a sprint. for this particular feature. And then I would drop that into cursor and then each one of those cursor agents. So when the architect was done with the plan, the CTO, I didn't drop that. That's the one touch point. I dropped that back into Chad GPT and Chad GPT said, yeah, they got it right. Or a lot of times Chad GPT would say, no, they don't. They, they missed a key component. You got to put that back in there. Cool. So I put that back in there, the architect ran again, and then once it was signed off on, the front end agent would finish their work, send it to the back end, send it to the QA, send it to the verifier. And then I'd close the loop by saying, here's the output, Chad GPT CTO, did we get it?

love it so much because you went down this road of a like following like a, you know, kind of almost a replacement of a lot of human functions, right in regular corporate America, you know, when we're doing projects or we hire a development ⁓ contractor to come in and do a project for us. They have these roles, they have these, you know, they There's somebody who does architecture and there's a system design person. There's a front end person. There's somebody who specializes in databases sometimes. Sometimes it's the same person, but often in a large organization, we specialize those roles. And you've worked in these large organizations where you've, these people have been line items on your budget.

That's a hundred percent it. It's, and the, the, the one thing that I will tell you about quote onboarding a new agent versus a, you know, a human, I just point them to the, so there, I point them to these files that contain here's what we're doing. And it immediately knows like kind of the the brand identity that we're going for. There was one document. in this particular SaaS product, the email is really, saving a structure, really it's the email though, is like a contract with the customer, right? So it's a very simple workflow, but the thing that has to be nailed is the email. It has to look the same and it has to provide some actionable insight to owners. Otherwise, why buy it? Right. So one of the first things, and I think the, I think Chad GPT suggested this to me and that's just the flavor I was using was why I keep saying that name, but the model suggested to me, you know what? You need a constitution for this email. And I said, ⁓ yeah, that is good. So we created a full constitution markdown file. And then I, My idea was, ⁓ so before the architect ever writes the spec for the sprint, it needs to read that every time, like full stop.

No, that's like that. The idea of ⁓ the model having memory and context and ⁓ having the wholeness of a human being is still something that you have to today with most of these tools, you're having to do for it. But I mean, you're doing it in a little bit unnatural way. But what you're doing is you're rebooting it each time, right? And you're giving it you know, but instead of starting as a baby, it starts as a 30 year old every, you know, every time you hit the button. And so you're giving it the constitution and, and, and rules and rules of engagement and, and, ⁓ and you're giving it access to all the work that's being done before so that it can learn what happened so that it can catch up to the story. Right. ⁓ So, so it gets a last week on kind of an update for itself, and then it can keep moving forward. Those like, I, you know, and so, you know, if I jump in and start telling a story, you know, I took a very different route in building a set of a tool and which was, you know, both of us have said, you know, we love our families. We were involved in our communities. We've got ⁓ the calendar gets full ⁓ and, you know, everybody we're no different than anyone else. Everybody's calendar is full. There's not enough hours in the day. we're all trying to figure out, know, if, if, ⁓ if we've been lucky enough that our, you know, physical needs are being met by our jobs, then, you know, then what we have left to now maximize is our time. And so I felt like I was running and chasing and I wanted more time. I wanted to figure out how to optimize my time and, ⁓ you know, my years of doing, you early in my career having a day runner and Franklin Covey, you know, book with the binder with the way back machine and then, you know, the last probably 10 years using getting things done as a, a, as a, you know, evolution of that and using, you know, electronic lists to, manage my life still, you know, was now hitting a point of like, well, now this is getting like difficult to work with. And so I wanted to,

way way back machine man.

solve this problem. And as I saw these tools coming up, I'm like, ⁓ this would be great. Can I build something that, you know, and so the, you know, what you've heard me talk about, you know, I was kind of joking about like, you know, I want to be Tony Stark and want to Jarvis to do this for me. And I wanted to have this smart, ⁓ you know, thing that, that knew me and knew how I worked and how I thought and knew my ⁓ details and could coach me and tell me what to do next. And ⁓ and could go off and do work for me. so, ⁓ you know, the more I dug into that, ⁓ I realized, ⁓ I'm not Tony Stark. I'm just like everybody else who's in technology. A lot of people are building this and.

I think I sent you a few articles.

Yeah. And I mean, then you go on GitHub or, and you, there are people have open repos of things that they've built. And, ⁓ so one, I'm like, I still want to do this. I, ⁓ I, I feel like I can't call it Jarvis just because it, feels, it's, it's, yeah, I don't know. It just feels very, it feels like very, very basic. And so I'm like, ⁓

I called mine Sam cuz that's Sam. Well, that's the lady from that the her movie Joaquin Phoenix and scar Joe. Yeah Cool enough to say scar Joe. So we should probably not do that again ⁓

⁓ Star Joe's character. So I don't think we're. going. Okay. All right. We'll stop. But ⁓ so I'm calling mine. I'm like, ⁓ mine's mine's Cyrano. So I figured I'd go with something

that name man. I think that's smart.

I just want something whispering in my ear over here and giving me clues. But what I did here was I wanted something to start with that was looking at my to-do list, my future to-do list, my goals, and some of my contacts and networking, some way to stitch this together. And there's many people building many very, very complicated tools around this with lots of layers and lots of technical depth and complexity. I just wanted to go ahead.

right. You know, every time I send you an article, you know how you, you always tell me you just sent me down another rabbit hole. We, should call it Sisyphus. Cause I feel like you're, I feel like, I feel like, feel like you're pushing this thing up the hill every day and then starting over. It's a problem with this. mean, like that's a problem with this whole limitless pill.

It is right. mean, it's almost like a stack with the limitless pill and you know, if you have, you know, any any form of ADHD spectrum in you, you're wanting to do stuff, you want to do lots of stuff. I do I have felt a little bit like I've started this project over several times, but I went down the route of like, well, instead of going with a I'm going to create a bunch of agents and build this. using a tool. I'm using Manus, ⁓ which recently got purchased by Meta. ⁓ they have a agentic platform where it works things through to completion all the way through. It will build and test and code and save your code and do pull requests and do all that stuff in one shot. And the secret to that I think, or being somewhat successful with it is, ⁓ over the course of my career, I've watched paradigms of how we write software and how we manage and project manage teams change radically. ⁓ But early in my career, the idea was still waterfall. You go deep on requirements, you spend a lot of time, you validate, you write corner cases, you think through exceptions, and then you give it to a developer because that was a precious resource and we didn't want to rework things and the idea of iterating was not really like part of that. so, but what that did do early in my career is build in this muscle memory for like, well, good requirements look of kind of a very specific way. And so I feel like that's just a function of, you know, having, you know, been there, done that and have the t-shirt.

that translate? I'm curious, how does that translate to requirements so that same exact process works well with AI or are there tweaks?

It works great with AI and the idea with it is...

there's not any tweaks like you don't you can't be more vague or

You can, but I think for me, getting one-shotting, trying to one-shot some of this work ⁓ where I give it all the same things. So it's not just requirements, right? There were functional requirements, right? But then there were technical requirements. And then again, having the background meant I could define those upfront instead of letting the AI come up with architecture and solutions on its own. I was able to kind of give it guardrails, right?

your requirements building is still like the crux of the work. It still takes up the most time probably is what I'm hearing. but what is the, mean, are you able to iterate quicker?

You are because the build cycle is crazy fast. And so it took me for one project, ⁓ which is kind of like a CRM for a company that's doing some care management. It took me an hour to really sit down, think about what tools I wanted it to use, what kind of database I wanted it to use, think through things that are pretty natural for me to think about, about what are these objects, what are these entities that represent a customer and a lead and ⁓ the demographics around that customer and what was the workflow I needed it to do, what the end user have to do. Took an hour, then 20 minutes later, working software deployed.

You spend, it's requirements are an hour given the assumption that you have 20 plus years of experience in a discipline, it took you an hour. What does that take? Not 20 years of experience guy, person.

Well, but not 20 years experience guy who's curious, who understands that part of it is just knowing the right question to ask. Right. I mean, for you, you did very similar. You said to your language model, you're my CTO. And you gave it that role. You gave it the role that I have in what I'm doing. And so you made it 80 % of the way there just by you know, asking it to do the things that I would do.

Yeah, I still think though there were several points in the process. I'm curious how you fixed this, but there's several points in the process that, you know, I would spend for an hour, an hour and a half to fix something and I don't think you would.

I, there were times I did, those are learning things. I think you and I have talked about with these tools. I mean, they've evolved even in the last six months as we've used them that, you know, asking it to ⁓ fix something is ⁓ not always the right thing to ask it and being able to feed it exact error messages and, and, you being able to grab things out of the console and the stack and give it to it or, ⁓ you know, we're getting like down in the weeds with it, but we've all, we are, One of the more important things that I think I've talked to several people about where they're just like, oh, AI is great until it's not great. people have told me it craps out on you. And I'm like, no, you are right. It does. Because it's like you have somebody working all day long. And by the end of the day, you're a little tired. Your brain's a little mushy. You're a little spent. In a way, there's a humanness to what AI and these large language models are set up to do where it gets a little mushy at the end of the day. And so you have a couple of options.

That is maybe the best analogy I've heard for that.

Keep going. you get to the end and you're like, ⁓ now I'm not getting the same answers. The quality is not good. It's not answering my questions the way I want it to. It's messing up. And so there's two kind of things you could do. You could try to start over. But if you start over fresh, the context gets lost. ⁓ We're seeing two different models here for trying to preserve what we call

Smart.

Context rot where the AI has been treating the earliest part of the conversation with the most weight Statistical weight and by the end it's treating the next things with less statistical weight so we can either ask it to do a Summarization or a compaction of this information and move it to a new fresh thread and get it to kind of start over But you have to have this handoff document that says what's everything we did? you know, what's the narrative? What are the system events that happened? What did we learn? What were bugs that we fixed? What was coaching that, you you have to give it enough onboarding material for the next person who's going to work on your project. But you give that, you extract that out of the thread. You start a new thread. You're going to get, you know, fresh, you know, it's like us running with fresh legs underneath you. ⁓ on a new race. And so I had learned that to make it do that when I saw my project start to get mushy and...

And you usually know that happens because everything like you'll type in something and you'll prompt it and then like 30 minutes later, it'll come back with a prompt. Yeah.

I had the AI ⁓ come back to me because it was a bug. I was trying to get it to fix and it wasn't fixing it and it was a deployment bug. So it had to do with getting things put out onto Microsoft Azure. So the AI also had to know enough about Azure to coach me through the deployment process and to manage the deployment process. It tried, try this now, try this now, try this now. We went multiple loops and It was one of the more shocking things that I've seen come up in the thing. It came up and said, I'm sorry. I have failed you. I'm clearly not able to figure out how to deploy this piece of software. It's what it said in the chat thread to me. I'm like, that, like, it is self-aware enough to both apologize to me and to say that it is exhausted whatever statistical ability it thinks it had at winning.

It's wild.

and said, you know, like, like, I'm calling this game. So

You know, I just thought of something that I'm going to experience. So the way I, first of all, the way I handled what you're talking about is that's why I use sprints. I had, as soon as we were done with the sprint, there was a sprint handoff and then it needed to go back through and basically wait all of the foundational files again in the new chat. And so It always knew what we had just done, what we were trying to accomplish. And here's where the whole full project is from start to finish. But at the beginning of every chat, that's kind of that sprint methodology, which we're all probably pretty familiar with at this point. That is what that's how I got around it. But one thing I thought about was, I wonder if at the end to get a better handoff, And I don't know if either of us have tried this or if somebody's probably tried it, but I wonder if you could say, I don't care. Take a little bit longer, but I want you to wait everything in the project equally and give me the handoff. Or I want you to, you know, wait the, you know, the, last third of our conversation more. I wonder if the model would do that.

I don't know. mean, think what we're seeing as people are building things is that this idea of handing off knowledge, memory, context is definitely something people are trying six, seven different ways at handling and to different depths. It is the prescribed solution though, I think. And what you were able to do, and I've talked to other colleagues who able to build things, who built things that because you took an iterative

think I've as good as mine though.

There's there's was not it was not as good. I remember the look on your face when I said I I had lunch the other day with another CFO friend of mine here in town.

I don't want to hear about your other colleagues. I'm the only colleague. And I was like, I'm the only CFO you have lunch with. I know I'm 1500 or 1600 miles away, but come

Hahaha Come on, man. will come there for a burrito. But I think what we're getting at here is that there is some very specific things. We talked about this podcast being something where we give some practical advice. We've gone pretty deep into kind of practical things. And if I kind of lifted up a layer here, I think it is that if there are folks who are having trouble, folks who are kind of going in that direction of like, oh, it's good until it's not. I'm like, well, part of it is that... you're not using it right. Or you haven't learned that there are solutions to when you feel like it's drifting in its accuracy and its ability to ⁓ drive value for you. But at the end of the day, what you and I have both seen is that we are capable of building things, not just because I do technology, but I can build several things. You've built several things. The ⁓ economy of what we're doing here ⁓ is, you know, the parallel that occurred to me the other day was I was with ⁓ my oldest girl and she's a Girl Scout and we had Girl Scout World Thinking Day where girls learn about other countries around the world and the Girl Scouts troops all present information to each other. ⁓ my daughter and her troop had Bangladesh and ⁓ that is my heritage. And so it was easy to be able to like pull a bunch of stuff together from our house and from grandma and grandpa's house. ⁓ You know, one of the things that Bangladesh is known for ⁓ is there's a Nobel winner, ⁓ Mohammed Yunus, who built the concept of microfinance, right? And these microfinance loans that were given to ⁓ women, especially in ⁓ villages, enabling them to build a business and to sustain their families. But the macro concept there is a small targeted investment has value. when we work in private equity and stuff, we talk about projects, set, these are big projects, right? We're talking, we're normally working with mid-market. We're talking top lines of 500 million or more. The numbers are not small that we discuss on a day-to-day basis.

I mean an ERP implementation is probably three, maybe north of $3 million at that size. Easy, easy. Depending on how messy it is.

And so this macro concept of small investment having great value, in a way, this where we've landed with large language models and this agentic coding and vibe coding is kind of ⁓ equivalent to that. We are able to invest in small increments into things that have great value and long-term value where before we might have written off making small investments saying small investments equal small outcomes. And now small investments could have great outcomes. a little bit of it is a function of right place, right time. But a little bit of it is just understanding what we started with. Did we pick something that has value?

Yeah. So I mean, then that's near and dear to my heart because I'm in finance and that's all I'm ever looking for, especially, well, I mean, in any environment where the company is going to, you know, or is contemplating exit. That's what I'm thinking about, you know, is like, how does enterprise value go up based on whatever project it is? we know we can, you know, like to your point, we know we can build all this stuff. how do we sift through it and find the stuff that we should build? And I think personally, I mean, we don't have time to go into it today, but I think that might be a topic that you and I should explore more deeply because this is impacting finance, right? Like if everybody can build something, it's like, well, I mean, like, why am I spending my resources on that when if I build so, Why am I worried about this back office solution? And as a CFO, I always want to worry about back office solutions, right? mean, that's my bent, right? But I have to have an enterprise view. So this marketing thing over here, that's actually the real driver. How do I do that? And so defining value and figuring out where you want to allocate these scarce resources. And then for the IT side of the house, the question that I wrote down that I wanted to ask you. right. And you're not going to be able to get to it today. So this is just like, ⁓ yeah, we accidentally are teasing, I think a future future conversation, but what does this do to, know, we, we talk about it in terms of shadow IT, but like forget shadow IT, let's just call it something that everybody will recognize, which is

I'll do a hot take on it maybe.

building stuff and the IT doesn't know about it. Like insecurity, especially in these private equity companies, because a big part of the value is do you have your house in order? Especially when it comes to cybersecurity, nobody wants to talk about it. It's the least sexy thing on the IT list. And it's arguably one of the top two most important things.

No, that one is a good conversation to have. when I get into the conversation about security and cybersecurity and with folks and it is, there are certain pieces of it that are just table stakes. We all know it and there's conversation to be had about that. ⁓ But we also want to always like try to remember that the weakest link in any of that is still the person. So we can get a little bit wrapped around the axle about people building things that IT doesn't know about. ⁓ People are printing things that finance and HR doesn't know about. There's pieces of paper. I've worked in ISO organizations that had very strict desk lockdown rules, no loose paper. Anything that was printed had to say uncontrolled copy on it. ⁓ But ⁓ rarely. do you see organizations having that level of ⁓ controls in place in an office, even, know, setting? So I think, I know I've talked to many of my colleagues on the IT side who are afraid of what this conversation that we've been having will cause. And I think we can talk about it more. But the fact is, is putting your head in the sand, you know, and just like saying like, it's not there or I have to ignore it and using of the excuse of security as this like nebulous thing that's going to get you is, ⁓ I think, a disservice ⁓ in the bigger picture with the finance side and as we try to figure out how do we drive value for our companies.

Yeah. I just think me personally, we've seen this in technology several times, like every time like email came out, same exact issue, know, cloud computing came out, same exact, exact issue. mean, so I think, yeah, I think you've got to embrace it. I do think there's an argument to be made that you need to attack it with a little more urgency than some of these other things, because it does feel like the capability shift is much, much greater than a lot of these other things. And so now people have these tools that we just talked about for an hour. it's a conversation. It is definitely a conversation worthy of having. Yeah, at the company at the most senior level of the company and they need to have it pretty quickly

Yeah, the genie is out of the bottle at this point or, know, right. You know, Pandora's box has been opened. It's out there to pretend like it's not. And to then, you know, ⁓ you know, ask folks to ride around with an arm tied behind their back. I mean, it's only going to be, you know, viable for so long.

Yeah. So just to, just to kind of close us out, like we covered a lot of ground, but I mean, we stayed pretty tight, I think on, you know, the topic. So basically to sum up, we, we can build a lot of things now. I built, I mean, we have proof points. gave real life examples of things. I think we glossed over the fact that you said, ⁓ yeah, I built a CRM. So anyway, for free, we, gotta, we gotta come back.

to that kind of side.

Yeah, because it's like, ⁓ yeah, I built a point of sale this weekend. It's like, holy what? And that's going to happen.

That's 100 % going to happen. And I think we need to be ready for it and ready to not necessarily just sign off on all of it, but you use the word often, right? Just like approach it with curiosity.

Yeah. Yeah. A hundred percent. we covered, you know, Hey, we can build things and then we've got to be really, I think that now it's like, is it worth building needs to be, you know, you guys can frame that up as any different way you want product market fit. ⁓ Is it the wisest use of internal resources? You know, if you're looking at it internally at just whatever it is, I think there's a lot more time that needs to be spent there, but we also have a couple of other really, really good topics that I think we can explore much more deeply. Hey man, well, I hope you have a good flight.

All right, well. I'm gonna try. You have, know, hopefully all the banking and finance stuff works out.

Yeah, yeah, I hope so too. I mean, it kind of has to. So, you know, I'm

I think, you know, we'll just, you know, will that kind of stuff into existence and wrestle it to the ground.

I've got my vision board up here. don't, I don't. It's all Lamborghini's. That's all my vision board is. I don't have a vision board. I just, you know, it was a bad joke that kept getting worse. Somebody has a vision board and hates me now.

Hahaha Well, you know. Probably. And yeah, I'm sure we have managed to offend at least one person along the way. ⁓ sorry, not sorry, but we hope that we're giving some value, cutting through some of the noise that we've seen in the marketplace, real life examples, real world, real corporate examples of ways that we can use this technology. yeah, next time we'll figure out another... you know, interesting angle to dive into.

Yeah, so if you made it to this point, thank you, thank you, thank you for listening, and we'll see you back on the next episode.