Trying to automate what you're bad at isn't always the best first step. Sometimes it's good to enhance what you're really good at and make yourself exceptionally faster at it.
SPEAKER_02Hey, I don't know much about AI, but let me go build an agent doesn't seem too reasonable.
SPEAKER_03Where should they get started? Truly the best place to start is to articulate what you really want. What am I after here? Because it can be an agent, it can be other things, but we have a ability and access to a tool that has infinite abilities to give us a lot of great answers if we know how to ask the right question.
SPEAKER_02What's really possible and what is not yet possible for somebody like me who doesn't have a lot of the background when I get in front of these tools?
SPEAKER_03The foundations of building a good agent can transfer to any platform. It doesn't change, it doesn't matter. If you can do it right in one place, you can do it right in another place. And I think that's the key.
SPEAKER_02Who's coming to you to learn this though?
SPEAKER_03There's a wide range of individuals who want to have this skill.
SPEAKER_02How long does it take for somebody like me to know how to follow that Scouts framework?
SPEAKER_03As far as learning it, again, I've seen it happen in as little as 90 minutes.
SPEAKER_02How can somebody at least take that next step to start to not get left behind?
SPEAKER_03Take a look at your job, the deliverables that you're meant to do, and have a conversation with AI on how it can help you. And I would start with creating a prompt even that you use over and over again. If it's not a full automation, say, you know, what would be a good prompt, a good request I can make to AI that gets me a deliverable that I like a certain way? How would I build that?
SPEAKER_00Eaton helps companies bridge the gap between AI hype and execution, giving leaders practical frameworks to build agents, automate workflows, and turn business bottlenecks into scalable solutions.
SPEAKER_02Welcome to Using AI at work. I'm your host, Chris Daig. Each week we'll be learning how today's business owners, entrepreneurs, and ambitious professionals are getting more done with smart use of tomorrow's tech. Let's get started. Right now, every business leader is asking the same question. What are we going to do about AI? If this is you, ChiefAIOfficer.com has the answer. We give you a simple path forward where we provide executive and team training so your people know exactly how to safely use generative AI in their day-to-day. We also manage the deployment and implementation to make sure tools actually get adopted and deliver results. And we'll also guide company-wide transformation so AI becomes part of your operating system, not just another shiny object. The companies that act now will increase productivity, cut costs, and grow faster than their competitors. Those that wait will get left behind. So if you want to make AI work in your business, visit ChiefaiOfficer.com and see how we're helping companies of all sizes finally get results from AI. Hi, everybody, and welcome back to another episode of Using AI at Work. My name is Chris Daigle. I'm the host, and we're joined today by my guest Aton Polager, who I actually share an office with and one of our locations for our company at uh ChiefAIOfficer.com. Uh Aton is our uh resident genius when it comes to automations, agents, apps, and doing all the things that um kind of we're seeing a lot of non-technical business professionals be able to do now with tools like lovable and cursor and clawed code and now open claw. So um Aton, before we get started, I'm gonna do the the uh the Isenberg move and say by the end of this episode, what do you want people to walk away with?
SPEAKER_03I would want people to walk away with a few things, uh likely. Uh one would be that when we are you know talking about the the space that we're about to, which can get technical, the foundations of what we are going to talk about are quite simple. Um and how they can be used can has infinite possibility, but I want everyone to remember and be grounded in we'll be talking about you know real principles that are grounded in basic things we all know and understand well, while they can become much more complex in execution. But their theory understanding is very simple. So my goal is to demystify perhaps or help makes concrete some of these things that people people are hearing about right now about this space. Uh the evolution is obviously crazy. So having an understanding that is, oh, I can get this and I can stay with this, I understand what's happening and possibly have some new uh things they can use today, you know, and understand how to go to.
SPEAKER_02So listen, if you're uh listening to this episode at any strata of the business, the CEO, frontline employee, it doesn't matter, the access to be able to build some pretty powerful again, automations apps or agents today, just using natural language, is something that's blowing me away. I um I got my first account with Cursor in September of 2024, and it was cool. You could watch it build stuff, but the stuff would break. Now, today, I I, who am not technical, I'm the exact person you were just talking about, I'm able to do some incredible things because the tools have gotten better, but also because of some of the frameworks that I've learned from you. Um so I want to start, we we've got, and I'll I'll give you a perfect example. I was speaking to um commercial property owners in Nashville recently. And you know, one of the things that we kind of focus on when we're working with clients is kind of three things. One, they don't understand the risk, so we want to make sure that they kind of have governance in place. They're eager to get started, but they don't know where to get started, so we help them with pilots. And then the the real kicker is we don't have anybody to help us, right? Um, so most of them were in that boat, yet the questions I was getting were, hey, how do we build agents? It's like, whoa, like slow down, let's let's make sure that you know you guys know how to do the basics. Um so this is a topic that is not reserved for like the the bleeding edge of generative AI application. This is something that everybody's hearing about. So maybe let's start with may uh demystifying what is what's really possible and what is not yet possible for somebody like me who doesn't have a lot of the background when I get in front of these tools.
SPEAKER_03I think uh this is gonna be a question that I probably some people are, you know, it's gonna be maybe uh um polarizing. No, I'm getting it for some people it may be interesting, but I I I believe that anyone, and I know this both from training people and also from you know out in the field, I have yet to see, and this is gonna be, you know, uh maybe controversial, but I've yet to see a lot of the limitations of anyone who is resilient enough to ask the right questions and not find a way to what they're looking for right now. Um I think there is a lot of, you know, again, there there could be lack of clarity on where you're trying to go, and and it could be you know not knowing the right language to use. Uh, but really, I I truly believe this that we are in an age where anyone can pick up any of the tools. Um, now some of them are simplifying that into WYSIWYGs and you drag and drop certain things that people can quite literally connect, um, all the way to, you know, asking a code agent to do it for you and having faith and hoping that it ends up in the right place. And when it doesn't, being able to work through those problems and finding it, you know, I think that's the the actuality is I I think anyone can pick up this stuff. And it really depends on the degree of, you know, uh the fundamentals are being taught the same everywhere. I mean, when I I've, as you know, I've done countless uh certifications for places like chiefaficer.com, uh sort of, you know, from the chief officer program to all the way to uh Google and working through Progressor IBM and the fundamentals and how they're being taught, they're all the same kind of infrastructure, right? If we're talking about how these things work. So uh that's not changing. And I think anyone who can open a few blog posts can see how the main stuff works. Uh, but if the focus isn't on the outcomes or on what am I actually trying to accomplish and understanding the main components that lead to that, um, I think that's where people run into uh their their biggest issues, right? There's it's not about the tools. I think people can figure out the tools. Yeah, it's the foundations that will transfer. Uh here's a good example. Yeah. Uh just one last thing on that, because I think it's so important, right? I mean, the tools say open call comes out. Great. There was there's NA10, there's like you said, lovable, there's this, there's that. Um, and then paperclip comes out. And wait a minute, what about now? Wait, do we change everything? Do we do we not? Do we uh but it's the foundations of building a good agent can transfer to any platform. It doesn't change, it doesn't matter. If you can do it right in one place, you can do it right in another place. Um and I think that's the key.
SPEAKER_02Okay, so you mentioned like building quote unquote these things for somebody who's listening and they've they've just heard the term agent, but they're not necessarily clear on well, what it what exactly can an agent do? What are some recent examples of some things that you've thought have been pretty clever that should be either built for clients or helped them build?
SPEAKER_03Sure. You know, and I'll take it one resolution higher, just for those who are listening and are still, you know, wait, what's an agent, you know, as a whole for a moment. You know, the way I look at it is a progression of, you know, when you're interacting with any chat model, LLM, you know, large language model, where it'd be for ChatGPT, through OpenAI or Claude or Entropic, whatever, and you whatever message you sent, that's a prompt, right? Just to make sure everyone's unclear. That's a no matter what it is, even if you're not structuring it or designing it, you have sent a prompt, a request from the model. And the model has a set of instructions, it then replies to you from the knowledge it has, and you have an interaction with AI that is not an agent, right? Where I think we cross it cross that threshold and where most would benchmark it is when you now have a tool that is connected to outside tools. So if I'm talking to ChatGPT and I ask it to do something, it can also run, say, a report for me on my Salesforce account or can write an email for me on my Gmail account. And where you take it to a agency workflow is where there's a certain trigger that makes it do stuff for you automatically. So now say I have Claude that's connected to my email, but then I could say every morning, I want you to check my inbox and create a list of the top priorities I should answer and send me uh, you know, that notification via Slack. So now I have tools, I have my language model, and I have some type of automation of when, you know, a trigger that causes that to happen. Right. Um and that's when we move into a more agency where it's has its own reasoning, it can make decisions, it's it's going through a flow that you've determined for it. Um and it's not just you having to make this interaction with a chatbot that maybe just has knowledge, but no tools or capacity for you know recurrence, um, you know, beyond that, if that makes sense. So just for for that, now yeah, yeah, if you want to, you know.
SPEAKER_02No, I I think the example that you gave is like very simple, but applies to everybody listening to this. If I had something that could give me that morning brief of what did I miss yesterday in my inbox and my Slack and my Teams and all that kind of stuff, and can kind of give me the the TLDR or the 80-20 of what I missed, if I started my day and that's what was waiting for me, so much more productive, like from the get-go. Um, and I think that anybody could benefit from that. So that's the type of thing that you're talking about here.
SPEAKER_03Absolutely. I mean, uh, the best example I have is very recent. I mean, I was out of, I was out for a few weeks. Um, I had to travel, right? And I came back and there's a lot to catch up on. I have emails, I have uh, you know, where where are we on this project? I have notes, but also, you know, so I had an agent go, you know, essentially go out, check all the emails I got from that from particular project I'm working on, all the meeting transcripts, all the notes I have in Notion, combine it all, look through it, see where the status is, see the communications that were happening while it was a way to to give me a progress report. And then on top of that, create me a plan of, okay, so we are here, here's the next move, right? And now I'm back in the flow as if I never left and I'm up to speed. I know where everything's happening with every one of my multiple projects. Uh right. So that for me was an incredibly efficient way of using an agent to just be able to pull all of it together, right? So that's another example, you know, for something simple that we can give more uh business uh centric uh examples in a moment if you'd like.
SPEAKER_02Yeah. And you know, I I failed to mention this at the beginning, but uh, you are not uh a computer science degree, you're not uh any of that type of stuff. I mean, you're a soldier.
SPEAKER_03Right, previous uh previous soldier. Yeah, I mean I spent a decade in marketing and systems and behavior um to get an outcome out of get an outcome out of people, but not a tech, right? Um and that's the interesting part, I think, for me. But it's a um no, I've always been kind of in say I've been in SaaS or from the marketing perspective, working with the tech team, always curious about and wanting to be and kind of picked up some some code I needed for front end. You know, if I'm a marketer, I need to change something on the page so I don't have to wait for a developer to change, like move a banner somewhere. Uh so picking up some some stacks over there, marketing automation stuff. But no, only in the last few years is really when I, you know, started to uh went all in pretty much in my own education on this and you know, working thousands of hours and uh and developing and building with AI, right? Which I think is the greatest feature uh out there.
SPEAKER_02Yeah. You know, but you've got you had the luxury, not to say luxury, I mean it was a sacrifice for sure. You made the choice to dig in and focus on this like uh at an extreme level. Most people listening to this who are running a business, running a team, whatever, they don't have that option and they don't have anybody on their team who they can say, hey, take the next nine months or year and just go figure out this this vibe coding, vibe engineering stuff. So for those people what you you know, you mentioned that you've developed some frameworks and things like that that make it easy or easier. Um where should they get started? I mean, try like, hey, I don't know much about AI, but let me go build an agent doesn't seem too reasonable.
SPEAKER_03Well, I think that the the best, truly the best place to start is to articulate what you really want. I think that's a what am I after here, right? Because it can be an agent, it can be other things, but we have a ability and access to a tool that has infinite abilities to give us a lot of great answers if we know how to ask the right question. So I think having brainstorming and stuff that I did initially with A, I mean, sometimes I would just, you know, I was on a walk with my dog and I'm talking to it using the voice mode when it was originally using it, and just asking, you know, continuously asking questions, trying to figure out how I can use it and say, well, what can I do with this? I think it's hard in general and workflow sometimes to look at ourselves and say, well, what could I use it for? Yeah. And I know what I'm bad at, so I may try to optimize that, but that's not always the best path forward, right? Unless it's a fully automated system. Like I'm not gonna suddenly become great in my email communication unless it's fully automated. Because I'm not very good in my communication. I'm very good in my deep dive in work mode, right? So unless it's fully automated. So trying to automate what you're bad at isn't always the best first step. Sometimes it's good to enhance what you're really good at and make yourself exceptionally faster at it, right? And then that gives you also motivation because you're trying to get good at something you're bad at. You're not gonna get a lot of wins necessarily initially, and you hate be doing it anyway. So even working on that project will be hard for you, um, is what I find.
SPEAKER_02You know, that's interesting because it's contrary to like the strengths finder where like focus on what you're good at, right? And what you're suggesting is that a great place for people to start would be actually building outsourced solutions, agentic or automated solutions for things you're already good at as compared to trying to make your weaknesses better. I think that's uh counterintuitive, but makes a lot of sense.
SPEAKER_03Well, I think the if we look at the frameworks for like the best RY and I've seen, you know, and I've been uh working with different companies of different sizes, you know. Um, but what we see is uh in the field is this desire to build the new and desirable potentials, which is it falls into two categories, like this shiny kind of new opportunity that may be there, and a uh or the pain points of okay, this really sucks. I don't like this. Um, you know, what if I never had to write this report again? Uh but as an operator going into that place, you also notice that it's very hard to get them to actually give you the process of how they're doing it now or how they it's like everyone loses kind of in that dynamic. However, say you're a company who's running ads online and you can increase by a couple percentages your conversion rate, that could be millions, right? Yeah. Depending on what the situation is. Um, I mean, you've seen it before where we suggested a solution that could save hundreds of thousands uh and compress it into a few hundred, not because it was an agentic workflow, by the way, but just implementing that intelligent layer, that AI, into a into a middleware. But the idea is um sometimes it's finding these simple solutions that could be big wins for something that's a pain that's actually proper. And sometimes it's better to have a larger percentage on let's take what's working and make it work way better or faster or find little hinges where small percentages give big wins and have a budget for the uh new, desirable, and pain and the problems that I haven't solved yet in general, that I maybe AI can help me solve. Right. So I think that's uh, you know, at least currently where I sit with it and where I think the bigger wins are.
SPEAKER_02So I know you've done a lot of training of individuals like me, like our listeners, and that sort of thing. What can somebody who wants to learn this stuff, like what does that journey need to look like? Like how long does this need to take before they're able to say, look, Ma, I built this kind of thing, you know?
SPEAKER_03You know, I I think it really uh depends, right? I mean, you know that I, you know, run a run a cohort and I try to do the best that I can within six to eight weeks uh of the program for people to leave with everything they need at the highest level for it. Uh for people to just get in there and get something live, it can be fast. I mean, people are giving templates for all kinds of places to get started. Um, I think where like to know how to do it right, you really have to get your hands in there and you have to see the differences when you make a a small tweak to the even the adjective that you're using in a in a in a in part of this personality or system prompts or that make these agents overall personality, right? I mean, there's the core kind of four things that I think about with agents. Um, and mainly it's gonna be kind of their the character, the capability container and channel. So it's kind of you know, these four C's of sorts, right? It's so it's kind of a, you know, who are they? What are they? How do they think? That's sort of the human mechanism of understanding how to get them to think right and and work in the right process, capability, what are what tools are they gonna use to do this job, right? What are they connected to? How are they connected? What knowledge do they need to perform this task? Right. The container is more so where is it built? And that could change the environment. Again, that could be an open cloud agent, it could be a cloud code edge, it could be an N80 or a whatever you want to be. And within that, the governance and how can I, you know, make sure it's safe. And, you know, what are what am I following there? And then the channel is just where do I interface with this agent, which can be in many places that you can build design, but not just the actual interface, but where how do we, you know, what's the experience of that interaction to make it the best uh for your workflow, right? So I think those kind of core components really travel throughout, you know, all the different uh spaces that it may fill for the agent. Yeah. So if you if you say where do we begin and how long does it take, well, if you come from a background where you understand one of those sections better, say my marketing world and my interests drove me through behavioral economics and behavior shift and understanding models of behavior and influencing steps to create an outcome, really helped me with the top two of those. The, the, the kind of the um the character and understanding kind of, you know, gave me all kinds of creative ideas initially, with you know, replicating different influencers and helping them with their AI journey to capabilities and understanding what tools or what's the process of, you know, and the whole marketing world, at least the one that I've been a part of for a long time, is the funnel, the steps, understanding the process of step by step what needs to happen to get the outcome, right? So, in the same way with AI, that works really well in the capability section. Container and channels, really, where I needed most of my, you know, the technical chops to, okay, so now where do I house this understanding of how to create this, you know, capable personality that can connect to things, but how do we house it and how do people interface with it? So I think if you have one of those, say if you're a developer and you understand a container potentially better, or have the interface chops of connecting a chatbot somewhere, right? That might be, you know, a good place for you to go a bit deeper into and get your kind of roots of like, okay, I can get it set up. Now let me understand how to better design its process and how it should work. If you are more from the marketing, you know, product, you know, potential, maybe you start with more of the capability and and character side of it. And more, you know, if you're more dev backups, maybe the container is your main focus, so to, you know, so really wherever your experiences, I think the agentic, you know, if you look at those kind of Cs, it can help you find a place to go deeper into and how it ties into AI and as a starting point of your strengths that you may tie into better. Uh, you know, I've done work on all across, and I could say that definitely, you know, master the ones that I really love and understand well, and then capture those that you need to to just house it, and vice versa. Or you can partner with people who know those other sections too, also good. But I think most people can understand enough on how to launch this. People can launch, and I've seen it happen where we do a session where after 90 minutes, people have already something up and running. Is it optimized? Is it the best to you know, has it have all the things? No, but they see something that they can connect to tools that works and they can talk to, right? Yeah. Um and I think that's pretty powerful.
SPEAKER_02Yeah. So um, what's the difference between people? Because I'll I'll tell you, I've been messing around with open claw and it's been hellish. Get it built, breaks. Add something, breaks again. Fix it, breaks again. That hasn't been fun for me. What am I doing? And I would imagine that anybody listening, if you've had that experience, you know exactly what I'm talking about. What am I doing wrong that I need to learn from you?
SPEAKER_03So uh without sharing screen, I'm not gonna say that I know all the reasons. What I might speculate, because I've seen other people online having a lot of issues with things like that. And it's, you know, uh, and this falls into a category where I see a lot of um, a lot of the, and this is not uh on you, this is just in general, I see a lot of the uh challenges that show up and and people run into these barriers. It's like, well, you don't say this because you see the technology and you've used it and you do amazing things with it. Um but I think there's a lot of people who see that and say, oh, open cloud's not that all that, or it's not that useful uh because it's stuck on something, right? Um I think there's if we look at the foundational pieces of what needs to work and go step by step, that's typically where we'll find our clouds, right? So how are we connected to, you know, there's the the code that you actually have to use to connect to it, right? And there's options for that. You're either doing it in your local machine or you're doing it on a server, right? And those could affect, you know, certain mechanisms. There's a language model you're using and which one it's better with or worse with. And then what's important, I think, with say something like OpenCloud, but it could be with anything really, is if we look at the structure, or if we don't know how to look at the structure, we have a I analyze the structure and let me know, right? But it can be a uh in a way that I understand, hopefully. And uh, but if we look at it and we say, okay, well, there's these files in here and there's components. Is there something that's clashing? You know, did I by accidentally add something in the soul file that should be in the memory file that's not in there? Did I add a skill that contradicts another skill? Did I get these heartbeats and set up an automation uh trigger that conflicts or issues? Because a lot of times it's these silly little uh say contradiction that happens in there that breaks everything. Or if the model doesn't get picked up because you know, something in a configuration went weird and there's not an update that you didn't do. Right. So I think looking at each piece of this, say we look at the four C's I made, even though it's not four uh, you know, not particulars, but if you look at the four C's, it it would fit where you would say, well, let me test a few things here. Does it, you know, is the con it must be the container of the channel for the most part, because you know, personality things aren't going to be the big. So what in the environment could be not working? Why is this, you know, is it working in one place and not the other? I know that's for say, you know, Aton has one working. What the hell's going on over there? Right. Um, so is the environment different? Because he's running it on a mini Mac or a Mac Mini, where some people are saying don't do and some saying do. I don't know. I just bought one. I'm not, I don't have a strong uh hold on that, guys. Do whatever you want. But that you know, that I think there's an element of um troubleshooting, you know. So to know exactly what you're not doing right, I don't know. But to go through the steps of of each one and trying to understand where is the breakage happening, like a, you know, like a again, and this is where the understanding the who would who would know? Who would know? Well, uh a you know, a developer who is really good at debugging stuff might know why this is not working, where is the break happening? You may suggest adding logs to see where an error happens and it sends you a notification to see clearly where did the issue happened and what is it. Um so again, it's kind of learning, you know. I don't know that this is your situation, but I would say most people need to just add a few more tools or a few more um words to their vocabulary that would help them get, you know, I'm not a like we said, I'm not a programmer, I'm not a CS degree. Yeah, I didn't do those things, but I do understand that there's a process of, you know, QA, there's a process of debugging, there's a process of, you know, people who know how to do this. So I don't need to call those people, but I need to ask AI to become those people to help me figure it out. Um, right. So I had a lot of issues at first as you know, too. It took me a few days, you know, where I just sat down and I was like, okay, I'm gonna figure this out because what the hell is happening? Mainly the issue was the connection and you know, through Claude for me and figuring or figuring out the right model. And it was trying to call the wrong model a lot of times. And every time I introduced a different model, it would just right or no answer. So for me, that was a big issue. But once I got that resolved, everything else started to flow. But the automations, you know, just looking at the components of what's in there and then trying to go one at a time and put your finger on it because I can't really tell what your issue is without looking at it. But I would say that's the approach I would take and suggest also, and that people have seen me do live in our in trainings and things like that is yeah, all right, well, we need to figure this out. We need to research a bit more. Who else has this problem? You know, I'm running into this issue using tools like perplexity, using it, you know, using things like context uh, you know, uh seven that has a bunch of code libraries, uh, you know, and saying, oh mid someone figured out something there. Is there code examples, repositories where people fix this? Not that you have to do it, but you just have to know how to ask it and ask for what you want and see what you get back.
SPEAKER_01So based on what you just told me, I think the issue was my I call the index, it's a high-pack finder, like the high-pack finder plays into I want to do Twitter, I want X, right? And um the algo knows that I'm into AI. I think it's almost central right there. Like you think that everything is way ahead of you in AI when you're in there, because you know, here's my lessons from setting up open cloud and cowork and uh cloud computer users, yeah.
SPEAKER_02And I would save all those, right? And they would say, Oh, just just give this to your agent, right? Let your agent read this. Yeah, and I just crammed so much stuff in there, I think, that it was finally just like, dude, stop.
SPEAKER_03Just just it's not it's not you, honestly. I've seen, you know, so I I've demoed this live with people too, where I I every day pretty much I I look at something online and I demo it for myself and see if I can recreate it. Nine times out of ten, there's issues that they did not articulate in the demo online that you're gonna face. It's just gonna happen.
SPEAKER_01Uh-huh.
SPEAKER_03They probably solved it and then they come back and shoot it or whatever it may be. You know, it's like and it's always gonna be an issue that they did not mention that you're gonna face. And it's like, wait, why? Am I the problem? What's going on? I guarantee you you're not the problem. Well, okay, I'm not gonna guarantee that, but you might be the problem in some cases. But a lot of the times they've done some troubleshooting before, they've done this before, they have it set up, they have their computer set up a certain way. If it's a local thing, there's always stuff that aren't mentioned, and it's kind of like assumed, uh, but it's not really assumed because they know they had to fix it. So it's almost like I don't appreciate when people do that too much because it's kind of like you just did this, this and that, and there's that has never happened. Pretty much never have I just taken a demo. I've always figured it out. It's like, okay, so we actually have to do it this first or move something, or my setup, at least it's like this. Uh, but that's where it comes in. That's why I do it all the time to troubleshoot, because that's really the skill I think that's very helpful. Uh, you know.
SPEAKER_02Well, while you were out of the office, I I said, you know what? If I'm gonna build, I'm gonna work on building these things, I might as well like turn on Zoom and invite other people and just let them know, hey, I'm not teaching you anything, but if you want to watch my experience of kind of wrestling with this stuff and stumbling through it, do that. I think I did maybe I did a Monday and Wednesday call for an hour. And I was building outside of that for sure, but I would jump on and uh, you know, members of the chief AI officer community would come on and just just watch like what I was doing, and I'd kind of explain it. And um, well, I thought it was gonna be a little more uh exciting, but it it turned into uh you know, every session was really like, hey, it was working and now it's not. Okay, let's debug it. Here's how I debug it. And I'm I again, listeners, I'm not some pro at this, so I was doing probably what anybody like probably what you were doing a year and a half ago, just like stumbling through it until you like, ah, something stuck. Um and I finally decided like uh this is not something I want to do alone. Because somebody asked me, they're like, Hey man, you're a CEO of a company, you got other stuff to do than fool around with this stuff. And I was like, you know what? Like, yes, and I think that it's so important to understand this because of the impact that it's gonna have. I mean, the quote that um uh Jensen Huang from NVIDIA at GTC a couple weeks ago, he suggested every company needs to have, I mean, he called it an open claw uh plan, but just in general, like uh every company needs to have a a claw, a plan, right? Like this is a powerful tool. If you can get them billed and introduced and think about how to use them, they can transform your business, right? So this is not something that's you know, with a name like vibe coding or whatever, you think, oh, it's just kind of like something you do casually. No, like this is this can be major impact. So and speaking of, I mean, I know you're working with uh some pretty um impressive companies. Let me ask, are they asking for this solution specifically, or are they just asking for AI and you're coming in and suggesting what the solution should be?
SPEAKER_03Uh well, I I feel like we're right now in a place where people have a at least in the the size of the company of kind of mid-market at the moment and a lot of companies, uh, you know, uh plus private equity, as you know, and all that. Uh but seeing the um there's a lot of ideas and no one knows how to get there, right? So it's open for suggestions on how you go there. Um and I think there's kind of the the the few camps I see is there's the um, you know, I'm already in the process of evaluating these vendors who need, you know, support because it's gonna be big contracts, you know, seven-figure AI agent deals, right? To evaluate and see what that system does versus a building one and hate. Yeah. And how do you make the distinction between when it's time to buy a solution versus build one? Um, and really this the spectrum could be uh it's it's all across that. Like, come, let's let's look at we know our some of our problems. Um, you know, I'm lucky to work with some also really smart companies who've documented a lot, which makes life very, very uh easy for the AI side of it, or not easy, but a giant jump start when you know the process and you have documentation and it's well documented. You're like, wow, okay, we can work with this. Um, I think it's much harder when you don't know. So the the range that I've seen now is uh companies that got to a certain size because they've been able to document and work through certain uh you know, process and they have a good pulse on, you know, they have they have the finger on this pulse of I have an issue here, I feel like this is ripe for this opportunity. And then I come in and suggest and we can build and then we build through there. So what does the solution look like? We may not know, but we know this might be a good opportunity for it. It's kind of where I find people right now.
SPEAKER_02Okay. Um for people who are doing this, maybe somebody's had the experience that I've had, because a few people on the on when I was doing those Zooms, they would be like, Hey man, it's good to see I'm not the only one that's struggling, right? They say they saw me doing this. So what are some of the mistakes that you think people um probably the most common stuff, because I know like you're you're you're taking people like me and you're turning them into people like you who can follow the frameworks and are are thinking about things through uh like an architecture or an engineering kind of uh perspective, but in layman's terms, like what are you seeing like where are people struggling?
SPEAKER_03Sure. You know, I what comes to mind is is uh something that I say at the beginning of uh of of the you know the program around, but that's a it's often this uh idea of uh, you know how there's definitely people who you've seen who are unbelievably successful, and to yourself, possibly you say, How is that person doing like how that's like how are they so good? They're so like they're not smarter than me. Like, let's be honest. A lot of you know, a lot of people I see is like they're not that smart, they're not that, you know, like they're not they're just another person, and they're like, what the hell? How did they do that? Right? Um, and you know, for me, there was this desire to be, I want to be like that that dumb person for all these smart like tech people. It's like, how can I be that guy? Like, how the hell is this guy doing this? Nice, he doesn't have my, you know, he's doesn't have this. How the hell is he doing that? Right. And I think it's try it took me a very, you know, uh uh took me ahead in many ways because I was unaware of what I don't know, but my optimism or my hope of where I can get to is so high that I go probably past where a lot of people were experimenting with it because they said it can't do that, or it won't be able to do that. But my side, I don't know. Can it? You know, can it not? Let's see, let's test and connect that to a problem-solving framework and really just having an expectation of, you know, or or this realization of if some dude figured this out, eating Doritos and Coke, you know, drinking Coke at like two in the morning with his hoodie on, with headphones, and barely sleeping, you know, they're sleep-deprived, and they were Googling and piecing stuff together to try and make something work. And if they figured out code and built apps, it's like, man, I bet I could figure it out when I'm sober, just trying to solve problems and working with my, you know, working with an assistant that has access to all the stuff they had access to and more, and is able to know the benchmarks are getting crazy every day. It's getting better at coding and things like that. If we just talk about that, but the where we are in the field right now is we have this tool that can help us figure it out. We just have to have that determination and, you know, uh going back to kind of full circle to what you were kind of saying, what does it you know take to do that? I think the first step is this acceptance that you don't know, but you have support that will help you get there. Right. It's kind of like I have a tool that can help me get pretty much what I what I want to know and do. And believing that even if you don't fully understand part of the process or what you're getting at initially, like any new skill, right? I may not understand it, but I'll get there if I follow a process. Sure. Right. Um, that's why, you know, the best systems in the world, companies like McDonald's, where you can put any kid on the flinok and flip a burger, knows exactly what to do, when to do it. That's why, you know, that's why people model them for certain businesses, right? I think in this, it's also this idea of I don't know why I'm flipping it exactly every 45 seconds, but I should. And then you start to see the burger come out, you know, the way that it does every single time, like, oh, that's why I do it like this, right? And then maybe if you're an intellectual about it, you'll start to wonder why that is, and you can learn even more. But at the very least, you get the outcome of a burger made well to what you need to, right? And I think that's the the case is learning the framework to work through. And that's a uh a support, the steps that you want to take. So if you can have the mindset of, I'm not gonna understand this yet because I don't know it. But think about this. In what other world we've been able to just immediately have feedback and work with something that can also explain to us what we're learning as we're doing it and executing why why did it break? You know, just having this back and forth that is so, you know, educational and helpful while you're trying to figure it out. There's nothing like it. You know, I think you know, by now it's probably a you know combination of who knows how many hours of learning how to do these things that I've received with this.
SPEAKER_02So you've mentioned frameworks a number of times, and I I I know that that you operate from those in this process, and that's been your ability to transfer that knowledge over to the folks who are in the community who are now doing this stuff. That's the stuff that they reference, is those frameworks in particular. Can can you take a minute and explain maybe some of the ones that um you're you're leaning on the most? And then also how did you arrive at discovering these frameworks?
SPEAKER_03Sure. Well, I mean, there's been a I mean, I've spent a long time, you know, long, I not as long as as some, but I spent, you know, a decade working in businesses, you know, took my I was from a specialist to an executive and then went on to build my own. Um, and in all that journey from you know working within the entities and kind of seeing where things are. Um, I think it's important to note that because um moving on from there, you know, and getting my ass kicked as the initial years, as everyone do, yeah, um, but also examining, you know, when AI came, who can I learn from to understand this particular game best? And as a solo kind of consultant in the world, I want to study who's doing consulting best. So studying and reading and you know, certifications and like we said, and really collecting better, taking all of that and applying right away, which I really believe maybe is something that I do particularly different sometimes than others, where I take something and I immediately try it and see where I end up, right? Or apply it and just push through on it. Um, and I think applying and taking the learnings that I've taken from all of these places, like, you know, IBM, Google, and all that, and combining it into a framework to understand how they execute AI projects and what am I seeing when I've been doing the work I've been doing in my experience and what all these people I've learned from over the years and what I've been able to do on my own. Uh, but taking that and saying, okay, well, the first step typically is to understand and define the problem properly or the opportunity so you can understand what you're trying to work towards, right? So my main framework, you know, and I'll mention it here briefly, but if we talk about the main project framework, the scouts, you know, framework that we talk about, you know, the first stop is that scope the objective. So we say, how do we understand what we're after here, right? And the clarity of the problem. Now, there's plenty of quotes on it, right? Like, you know, defining the problem is 50% of the battle and so on. And uh big consulting firms like McKenzie, they put a lot of emphasis on that's the main, you know, initial part of the work and the research and the data is all to define the objective uh that they're after the problem, right? Taking it, disaggregating it into the components that make it and will make it successful or not, or what do we need to consider? And then moving from there, we move into the C, right, to sort of collect data. So in AI, the data is the one of the most important points that you can have. Does it have the information it needs to execute the job? Does it know the procedure, the process? Does it have, you know, what data do we have? But it's also connected to the research we're going to do. What do we need to look for? Is there people who's done this before? How did they do it? Right. There's uh frameworks that you know that I've they provide with that and prompts and all the jazz, but in the overall concept of what do I need to understand about this? If this is the first time I'm building something here, what would someone who's done this already a hundred times know that I don't? Where would they look? Would it be live, you know, what do I need to collect for AI? I don't have to understand all of it yet. I just need to collect the right information to do the work. So it has what it needs, the clay, so I can mold it, right? But if I collect the data, whether it be internal and institutional or outside or you know, from the world, moving into outlining our plan, right? Of what we, you know, in our O of what does this all look like? What are the steps to accomplish this task? I have the data, I have the objective, what do I need to do with it? Right. And then overall, to the you know, you we unleash and we start to build and we move quickly to get a prototype of what it may look like so that people can test it in RT. So we want to build something quick because once we get it to people, they can test it. Once they test it, and there's you know the steps there, you know, for function, for style, for uh governance is where we test in there. You move to shipping it, how do we wrap it, explain it, create the training materials for it so people can move from this idea or problem all the way to a packaged, you know, SOP of how to use this new tool that they have. But within that, when I talk about the four C's and all that, it could be specific. So when I say frameworks, everything pretty much sits within scouts, and there's a lot with you within each pillar, there's the kind of the best way to do it. But then specificity for you know, if we're talking agents as we are right now, but those things are core and haven't changed in a long time in the big world of products. So they're derived from best practices across and from hands-on, you know, building over hundreds of chatbots by now and you know, uh, many automations, multiple apps, you know, end-to-end with AI embedded in them, you know, solutions ranging in companies all the way from, you know, a billion to millions to whatever it is, right? These concepts work um as a whole, right? They just work. Where we deploy them, how they work you know within a specific tool or process, that's you know, becomes less relevant. But if you grasp those concepts and you apply the right kind of methods to each thing, it doesn't matter what happens with the tools.
SPEAKER_02How long does it take for somebody like me to uh know how to follow that scout's framework?
SPEAKER_03Well, I mean, in essence, if you just if you listen to what I just said and and repeat for a few times, you might get it and just have a sticky note. You'll probably know, okay, and you figure out. Uh, but as far as learning it, again, you know, I think there's I've seen it happen in as little as um you know 90 minutes, and I've seen it happen after weeks of the seeing the same process being executed over different types of projects. Yeah. Um, and then at the one project that's relevant to you, suddenly it clicks. Like, oh, I I get it, I understand. Like this is how I'm doing it. Because it's not necessarily linear. You jump sometimes back to research. You, you know, it's those pieces of the puzzle are always kind of there and may shift, right? You ship something, you're testing it, you may go back to building, right? Um, so it's kind of a um understanding, but it often if you once you get the framework, and when I see this happen, it's honestly one of my favorite things is you start to see people just get it and they can get to outcomes they want to because they understand the oh, this is what this is what this means. This is how I solve this. I might be missing some data, I might be. Missing some you know steps here on the plan and not seeing something, right? On top of that, I think the greatest and this is okay, I'll throw in one last one thing here that's really I think uh I think the greatest one of the greatest things you could do with AI is um ask for the questions you need to be asking.
SPEAKER_01Yes.
SPEAKER_02So who who is who's coming to you to learn this stuff? Like what what who what's the avatar? Is it execs? Is it entrepreneur solo folks? Who is it?
SPEAKER_03I would say that the the there's a wide range of individuals that seem to be uh, you know, who want to have this skill, but it's you know, a lot of people who aren't necessarily technical, um, I've noticed a good amount, and the good amount that are incredibly technical. I think that really we are right now in the age of uh agents. I mean, that's why the market, every time Claude has a uh update that, you know, their stock is going down because they're worried about mainly seats being taken. I don't, you know, won't go into the market, but but the overall, you know, I think we're in a time where everyone needs to at least be aware of how this affects them and how to do it. But we've seen people, you know, who have never even touched a line of code or seen an automation tool, or castration tool, or or you know, move all the way. A lot of people who are certified GI officers, of course, uh, you know, from the strategic and landscape want to get the hold on that skill as well, uh, which makes sense. I mean, if you understand the landscape of that and you want to possibly either be the closed loop on it or know how to effectively find people who know what they're talking about, then it's typically the people I see, you know, in different environments. Very few people who are on the on the, you know, a lot of people who are high-level uh, you know, operators out there who want to add a skill set that's, you know, the the the rise of where we are right now. I think that's the that's the people I see mainly.
SPEAKER_02So I saw a tweet from Cody Sanchez based in Austin. Cody does a lot of talk about entrepreneurship and that sort of thing. And I've seen a few posts from her saying that what really transformed her business recently was she found some builders, some people that know this stuff, right? And I I don't know that she's necessarily somebody who's building things. And I don't know that she has an army of them, but the impact that she said of having a couple of these like claud coders or whatever come into the business to where the operator can identify hey, there's a constraint, there's friction, there's whatever, and they can turn that over to the person who's like, okay, let me let me go over there with Claude Code and see what we can fix or whatever. Yeah, she said that's been like a huge lift for her business, and that any business that is looking to uh like adopt AI, yeah, get your people trained. Yeah, teach them how to use it safely, all those sorts of things. And make sure that you're not leaving out that builder, you know, cadre. You've got a couple of those people or one of them who can come in there and what's the problem? Okay, great, give me a couple days. And I'll tell you, you know, a perfect example, like my big aha with this was we were working with a construction company in Orange County, and they had a like, like everybody, you know, every company's got them, they had this big monster spreadsheet that was like just as it was the only way they knew how to manage all of these elements related to uh insuring, making sure that their subcontractors had appropriate insurance and it wasn't expiring and all that. And it was taking three people, off and on, you know, they all kind of had their hands in it, but with replying to emails and following up with stuff, about 20 hours a week between the three people. And it just was it painful? It was just part of what they did. They didn't really think about it that way, right? And I brought in a guy and I was like, Hey, can you build an automation? thinking like N8N was gonna be the solution. And he's like, uh sure, whatever. 24 hours later, he's like, uh, hey, I fixed that. And it wasn't an automation that he did, it was actually like he clawed coded a solution. And in 24 hours, he had created something that took it from a thousand hours a year of maintenance to a hundred hours a year of maintenance, right? And that was one task and you know, one department in this company. And if a company doesn't have those people who can come in and well, we just wiped 900 hours of you know, uh, like uh inefficient bandwidth off the table. If you don't have those people, and I do, eventually it will start to show you will not be able to keep up with that that company that's got the builders or the company that's led by a builder or something like that, right? So um all that to say for the for the listeners, this is a role that you need to be sourcing like immediately. It's probably much easier for you as a listener who's like exploring AI to say, Oh, I don't know about a AI transformation, I don't know what that looks like, but boy, we've got one of those spreadsheets. It'd be great to not have to do that stuff anymore, right? Um and I can tell you that as a as a listener, if you are that person, great if you can find them, hard to find. There's there's not a lot of people out there. There's people who can build you like a little cutesy app or whatever, but they're not thinking about business through the lens of an operator like Aton is, right? Um but even better is if you, as the executive or as the team lead, you have the skill and you can say, Oh, you know what, I'm just gonna fix that tonight, right? Done, gone. Oh, and by the way, tomorrow I'm gonna show my team how they can do this thing too. Like, then you don't have to worry about going to source those people out there because they uh they're hard to find. They're one of the hottest commodities that are out there. Hey, just a quick break. I want to tell you about something that's worth paying attention to, particularly if you're enjoying the topic on this episode. If you've been listening to our show for a while, you know that I talk a lot about AI that's actually working, the kind that saves real hours is cutting real costs in companies, and is showing up on your CFO spreadsheet. But I get asked all the time, how do I actually build these solutions? And that's what I want to talk to you about now. That's exactly what the AI agents and automation builder certification was designed for. It's a six-week live, hands-on program taught by A Ton Pollinger, right here in this episode. And it's built for people with zero technical background who want to come out the other side, able to architect, build, and deploy real AI systems for real businesses. You'll work through the same professional frameworks that we've been discussing on this episode. And by the time you're done, you will have a fully built AI mini app, a portfolio of production-ready builds, and a client-ready offer that you can take to market. So our next cohort is kicking off soon, April 21st. Live sessions will be held every Tuesday and Thursday, noon to 1.30 p.m. Central. So if you're ready to stop experimenting and start building, go check it out. The link is www.caio.cx forward slash agent. With all of our cohorts, the attendance is limited. So if this is something you're interested in, I would encourage you to go there, check it out today, and enroll. The classes will be starting in just a few weeks. Anyways, now back to the episode. So we're getting we're getting to the end of the episode, but I want to make sure I'll be doing everybody a disservice. If we didn't talk about this cohort that you've got going coming on, and the reason that I did this episode now is because we're a couple weeks out and they fill up, but this is something I just I see the impact. I see people that are graduating from this certification and like like they're blowing me away with this kind of stuff. So can we talk a little bit about what that what that experience is going to be like for those who are early? First, let's talk about what it's what is it called?
SPEAKER_03So right now we are working with the, you know, so we're calling it the AI agents uh in automation certification, you know, strictly to what it's meant to do, which is, you know, we're moving in this economy of agents and you know, all it's only going more and deeper into it. The goal is to take people from uh understanding how to approach a solution build uh all the way to the strategic, you know, plan on how to execute it, to have the tools, the templates, their own AI chatbots and you know, agents that have been trained on this, on the on these things. But take them from wherever they are now all the way to being able to create, execute, launch, deploy, you know, agents that would be helpful for their business or life or whatever it is they're wanting to do with them. You know, we go all the way from truly, you know, the design and the thought to the elements of what makes a top 1% agent. And this is something that this is really the differentiator. I think there's a lot of people who are showing how to do the basics of what these things can do. Uh the results I've seen and that we see in the field are so instrumentally, you know, 300% performance increases and just applying one of the changes we do to how we upload a document, even to a chat bot, that increase drastically the results of how good it is, right? So we go through the fundamentals that won't change and we make and we go through understanding how to execute them and give the tools to do them. Uh, and then in each step from automations to fully connecting different tools to that agent conversation through chat or through some type of trigger, but really the A to Z, you know, how do we go from uh the you know an idea and even evaluating a good idea to going all the way to the solution design and build, deploy, ship, and present. Now, for those who are in you know, offering it as a service or within a company, I've had people from both uh in a program. Uh, we've just had a lot of success with it. Uh, but the main goal is not just the having that the education, but having tools to help you jump in faster and are able to see the the outcome from probing the problem to articulating the problem. There's pretty much a by, I believe I gave like 30-something custom chatbots in the last cohort. Okay. Most of them are just for fun or, you know, or people can explore with them if they want to. But all the tools are there. Templates to fill out, you know, really uh a lot of as much as we could do to deliver the highest level of expertise in a kind of packaged way so people can walk through it. And uh, the more you do it, the more you'll be, you know, confident in it and have the skill and the mindset of deploying these things well. So I'm very excited for it. This is the third cohort we're doing. It's been incredibly successful. And, you know, if this is someone's listening and they want to, you know, have the operator mentality, you know, there's a lot of operate, you know, a lot of prototypers out there, like Chris was saying, you know, people who can do something nice on YouTube, but to be an operator, there's a few other elements that we require to take it to the finish line and to make it a business outcome. That is great. So I'd be excited to have anyone who wants to join us there.
SPEAKER_02So so for the listener, two paths here. One, if you're uh you know the business owner or the leader, I would say, just like with what what Cody Sanchez was suggesting, get you a couple of these people. Um finding them is tough, especially like, okay, great, but what do you really know? And are you are you the listener? Are you able to judge that? Um, I'd rather send somebody who already knows about my business, who knows our industry, like got get somebody on your team, send them through this. Secondly, um, if you're so inclined, best case scenario, you know these skills. Whether you're the leader, whether you're uh, you know, senior level management, whether you're somebody who's early in their career, if you have these skills, you are playing with you know a jetpack on your back and everybody else is lacing on their running shoes. You really are able to, oh, here's a problem fixed, solved, agentic solution, automation, whatever that is. So there's no there's no scenario that I can see for a business to maintain its economic viability where they don't have access to this type of skill set, whether it's going to be you, somebody from your team, or again, if you if you can find them, if you can hire them, but I mean, you're gonna get the guy off YouTube. Uh yeah, I mean they're building cool stuff, but if they've ever been inside of a business before, a lot of them, the answer is no, right? So I don't want them necessarily saying, hey, the finance team needs your help. Can you go over there and you know make sure that the the month end closing is happening faster because of this stuff? So um, so what we're gonna do is uh we're gonna put a link in the show notes uh for this. And if you're interested, I don't know, uh there is a limit on the seats. I don't know that we've hit that limit by any stretch, uh, but it's a certification, it's recognized by the International Association of Chief AI officers, the whole deal. So it's a it's a real thing with a business intention uh taught by people who are doing this in like businesses. If you heard some of the names of these businesses, you would you go, oh, okay, if they're using this, I want to use it, right? Um so Etan, outside of that, if this time's not right for whatever anybody, how can somebody um at least take that next step to start to not get left behind?
SPEAKER_03I would say the first step would be if you're working within somewhere, truly take a look at your job, the deliverables that you're meant to do, and have a conversation with AI on how it can help you and how you can build something to help you with it. And I would start with what's a you know, creating a prompt even that you use over and over again. If it's not a full automation, say, you know, what would be a good prompt, a good request I can make to AI that gets me a deliverable that I like a certain way? How would I build that? Um, and understand based off your role. Now, if you're the owner and you're trying to figure out a bottleneck that you have, I would do the same, but more so from the vision of your, you know, how do I see AI playing out here? What are the places and having you know a sounding board of working with AI to find and articulate, have it interview you. Say, ask me all the questions that you need to about my business and tell me what are the you know few opportunities that are currently proven to work with AI, right? And it's okay, I know the visionaries of the proven word can go either way. But start with proven, then you can go towards radical or whatever you want to play with there. But but start with the, you know, what is proven because things are happening in your industry right now, 100%. People are developing stuff, people are working there somewhere out there. There's someone who got ambitious in one of these companies and they are starting to develop stuff because they want that, you know, they want the seed, they want to develop it. Um, and it's gonna be, you know, it it is happening in a pace now more than ever, and it is the time for it. So, you know, it's uh start with that though. Analyze if you're in a role, analyze your role, see what's the problem you can use. If you're a leader, have it analyze, you know, ask you questions, interview you uh on your business and see where the opportunities lie and what's been shown. There's already data out there, I'm I'm pretty sure in most industries. So uh check it out, see what works, start with what's working, and then move from there.
SPEAKER_02Awesome. Well, man, thank you so much for taking some time out of the laboratory. I know you've got uh a limitless amount of projects that you're working on for all those clients. So again, I appreciate you, and I appreciate you taking the time to help upskill people like me, the non-technical business leader who um is certainly intrigued by this stuff, but maybe a little frustrated if they try to do it themselves. So um, and for those of you listening, uh, if this is something, maybe it's not for you, but it's something that you know somebody who would benefit from this, uh, please share this episode with them. And any episode, obviously, you know, any reviews we can get from you, any um spreading the love, if this show has been helpful or just in general, the episodes are uh entertaining or helping you with your uh your learning curve when it comes to AI, um, the biggest thing that you could do for us would just be share it with somebody else. So uh thank you so much. And as I always say, you know, go out there and use AI, and I hope to see some of you in that cohort. I'll be joining it and um we'll have a chance to work together on some of these agents, automations, and apps. And thanks again, Atan.
SPEAKER_03All right, thank you so much. Great to be here. Thanks, everybody. Bye-bye.
SPEAKER_02Thanks for tuning in to Using AI at work. Don't forget to subscribe for more conversations about how to use AI at work. And a special thank you to our sponsor, Chief AI Officer, for empowering businesses with AI education and training. Visit their website for a free AI readiness assessment and AI strategy guide to help you get started using AI at work. That's www.chiefaiofficer.com. Follow us on Twitter at the handle UsingAI at work and visit www.usingai at work.com for free resources to help you harness AI in your role.