PodCrash with Matt Ebert
PodCrash with Matt Ebert
How Business Owners Should Use AI First with AI Advisor Sam Woods
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
So you're you're a fractional um AI officer. Um what for the layman in us, what what's that mean?
SPEAKER_00It basically means that I work with companies mostly in an advisory consulting capacity. Uh my team sometimes ends up doing work with uh a company. And we're just at this point, we're just helping people figure out one, what what to do with AI, and then also specifically what they should do with AI and everything that goes with it agents, uh, image models, language models, machine learning. It covers essentially the whole breadth and width and depth of what a company should do with AI. Because there's a lot you could do, but not everything is worth doing. So often what happens is that it ends up being pretty transformational. Um they do end up changing a lot inside their business because of AI and everything that's possible. So just trying to help as many as possible navigate that and figure out the best way that they should use all the things that are available.
SPEAKER_01That's a really important job right about now. So um I can see that. Very popular.
SPEAKER_00A lot of people need help.
SPEAKER_01They're probably lined up to use you. I can't wait to jump into some of that. But before I do, a little bit about your background. I know you dug into AI like early, like 2016 time frame is what I've heard, right? But kind of background before that, maybe even.
SPEAKER_00For sure. Yeah, fortunate enough. So I had uh used to have a marketing agency where we primarily worked with conversion rate optimization and got converged rates up and all the good all that goes with it. So traffic and conversions and economics. And we were working with that. So I started that business in 2012. And by 2016, we had started working with machine learning, which is like kind of what that's the larger scope that AI exists within is machine learning. Uh, we did it in the capacity of prediction, forecasting, modeling of data, analytics. Because when you're dealing with conversion rate optimization, you gotta have, you gotta have your metrics tuned in, right? You gotta have them accurate reflections so that you know that what you're testing is actually producing anything of value. So we were working with that. And at the time, uh in 2016, one of our clients was Amazon. And just to like preface that, I do not have Jeff Bezos on speed dial. Like I don't know him, he doesn't know me. Okay. I work with one of their hundreds of VPs on marketing at the time, right? So they have a huge company, and I was just one little guy with my team with one of their teams inside the huge organization. And so we were working on conversion rate stuff with them, running experiments and just figuring stuff out. And I had a conversation with a machine learning scientist who worked on their team. And he said, he goes, Sam, basically, the way that these models, machine learning models, predict numbers and do forecasting and all that stuff, he goes, in a few years, you're gonna be able to essentially predict words. So what he was talking about were large language models, which were being worked on at the time, just in like frontier labs, and it wasn't uh just wasn't publicly available, and it was very rudimentary and very basic compared to what we have now. And so he said, if I were you, Sam, I would like dive into machine learning, I would pursue uh exploring large language models and how all that works. Uh, because in your line of work, marketing specifically, it's gonna end up being a pretty big deal, he says to me. And I like I barely understood what he was talking about because I'm not a machine learning scientist. I don't have a PhD in anything except you know the school of hard knocks. And so um, I kind of took him seriously. And so my team and I, we got into that world uh more and more in depth and just trying to figure out how do we use data uh and and data analytics with machine learning to just do better work for our clients. And as we were doing that work, we then started getting into the track on large language models in like 2018, 2019. And in 2019, GPT-2 uh existed and was publicly available if you could figure out how to use it and if you you know knew the right people. And so we started playing with GPT-2 at the time, which was a very basic text model. So you send it a sentence and then you get words back, which now that's just you know, ChatGPT or Claude or any of the other large language models. But at the time it was pretty basic and you couldn't get much back, right? You gave it a sentence, you got a sentence back, and you had to go back and forth a lot. But we got into it then, GPT-2 then, and then in 2020, GPT-3 was available. Uh, and that made a huge difference because now you could get whole paragraphs back as opposed to just sentences. And when that started changing, and we started realizing kind of what was going on, and I just knew people, I was fortunate enough to know the right people at the right time who gave me access to some of these models and tools. And they were all telling me, Sam, like, you know, you're a marketer and you do stuff online with a business, you gotta get into this more. And so the opportunity presented itself, and I just took the chance to get into it. Uh, and here we are. So, like kind of the short version of that.
SPEAKER_01Do you you you feel you were at the forefront of it, or do you feel you were chasing it?
SPEAKER_00At the time I was chasing and trying to keep up with very smart people. Okay. Like I, you know, the the scientists and the PhD level researchers who do like in-depth work with models and compute and GPUs and like all of that. I think compared to them, I'm still chasing, but at this point, um, I think we're mostly on the forefront on uh the crossover between research being turned into real business value. Like that's the hard part is really you have all these capabilities, but how does that translate into something that's actually useful inside a business and uh even better, something that impacts the bottom line in some way?
SPEAKER_01It seems to me like we're right at the time now when the rubber's meeting the road as far as businesses getting value out of that. Is that what you're seeing? Because uh it it seemed like a lot of hype for a minute where you know I'll go to CEO peer groups and stuff and they'll talk about what they're doing. And I'm not so sure that people were seeing it affect performance, bottom line, you name it, but it seems like we're right at the point now where it really is.
SPEAKER_00Yeah, I would agree with that. I think you know, 2020 to 2022 before Chat GPT comes out. So ChatGPT came out in like November of 2022. And right around before that, it was not easy to work with these models, and it took a lot of work to work with them. Like it's not just press a few buttons and then you're good. Like we used, you know, Python notebooks, uh, a whole stack of them, and like thousands of API calls just to get some useful like ad copy back or landing page copy back. And even when ChatGPT comes out and for the for the next like 2023, 24, I think what happened is people uh started to see the potential. And a lot of the researchers and Frontier Labs, they're like, hey, this is transformational. And I think people caught a glimpse of what's possible. But as you know, like this there was and still is a just a ton of hype, a lot of promises, a lot of big picture, a lot of, hey, here's all the things you could do. But the, you know, the difference between a demo and something that exists as a system that works and that's making work easier, life easier, and and have has an impact, huge gap. And I think in the past year, especially, 2025 and into now, we've seen a lot of the hype becoming real. Not all the hype, like there's still too much of it, and there's still a lot of um a lot of people just speculating. But I do believe, like from what I'm seeing on the ground, the people that I work with, other companies that we know in the space, a lot of the capabilities have caught up to all the potential. Okay. Um, still a lot of it, but it's better, and it's easier now to realize the value and to build something that actually works than it was even this time, like May 2025. You started to see, especially with agents, you start to see, okay, we now have consistency in how the models are performing. The models are more capable now, they can handle uh tools, code, and a wider range of use cases. So I think now it's easier to see and easier to also make it valuable and to make it work than it used to be. The hardest thing is just people coming to us and other people too, and like they have a list of things they've seen demos on and the hype, and they go, we just want this and we want it next week. And it's like not gonna happen.
SPEAKER_01So interesting you say that because I went to CES this this last year and you know, got to watch some of the robots, for example. And here you hear some talk about like robots are gonna be doing all the work in just a few years, and I could tell you those robots seemed like far from ready from prime time, so to speak. Like they could dance, but ask it to go get a white shirt off of the rack over there compared to the black shirt, and 20 minutes later we're still waiting for a shirt. So like I don't know how it's gonna take over the world in a couple years from what I saw. But is for those that are like it's really polarizing, it seems like some are really afraid of AI, think it could almost then humankind. Um and then what are your thoughts on that? Is it to be feared or or should we embrace it?
SPEAKER_00I I think as with all things, uh moderation and wisdom and understanding, like just because you can do something doesn't necessarily mean that you should do something. I think there are aspects of AI systems and agents that could be dangerous and could be a threat in s in many ways. And the key is all is as with all technology, I think the key is just not letting the technology dictate what you do, but instead just know having a strategic vision, having a purpose, having uh a good head on your shoulders and a good team around you, and then just working with it in a way where your needs and your preferences and your use cases take the driver's seat, and then AI helps enable it as opposed to just going, let's just do everything that AI could do. Uh there are risks, there are threats. I'm not gonna like sugarcoat it and say it's all good, but it doesn't have all the threats don't have to materialize, right? Like we need smart people, good people who uh emphasize and uh really value the human aspect of everything. I think with the right people doing the work, we can see a lot of good potential. There are threats. Um, I don't think the doomers are uh 100% correct in what they you know fear might happen. I think we're at a crossroads. I think how things develop in the next couple years will uh determine what we see, obviously, but it'll determine which way we go. Um I am encouraged that I see a lot of people not rushing into it. There are plenty who do, right? But there are a lot of people are are asking the important questions, which is okay, you know, we have a team of, let's say a marketing team of 10 people, and technically speaking, you could replace eight of them with two people that manage 20 agents, and those two could do the work of 10. But that doesn't mean that you have to do that. You could instead just enable all 10 of your team members to do good quality work and train them and and educate them and make sure they have the right tools available. So just like I said, just because you could fire people because of AI, it doesn't mean you should, and you don't have to fire people because of AI, right? There are ways to make use of the good parts of it and just strengthen people's skill sets, make sure that they're using the right tools for the right job and get more value. I I'd rather see 10 people in a marketing team, for example, use all the available tools and models in a good way, productive way that increases value and and just produces results, as opposed to just firing eight, keeping two, and then you know, hoping for the best.
SPEAKER_01Yeah, I I agree with you. I wonder though, if most companies are really gonna act that way.
SPEAKER_00Yep, that's the question, right? I think it the easy thing to do is to fix your bottom line is always to cut staff, right? To fire people and to let go of payroll and all of that. I do think that if you do so, I think you'll be in the long run, I think you'll be at a at an actual disadvantage if that's your default. But I do I agree. I think I think we'll see a lot of companies uh fire people or just make structural changes in a way that people end up without a job or without purpose or without meaning in their work. Like they feel like all I do every day is just tell Chat GPT things to do, and then I just sit there and do nothing. Those are the risks and those are the threats. And I think we'll see a mixed bag over the next few years.
SPEAKER_01So bringing it back down to just more practical real life today, how should businesses be practical about it? And when if somebody calls you and says, Hey, um, I haven't done anything with AI, I'd love for you to come in and get my business going, what would your process be? What would you look for?
SPEAKER_00First, I would understand just what are their existing processes, how do they do work now in whatever the team might be? Marketing sales, um, off back office stuff, like RD, what whatever the team uh makeup is and what the work is. I would look at how they're doing things now, what processes exist. And uh often, you know, people uh think they have SOPs and they have clean processes and they don't really, but that's okay. I do we just need to start with like how do you do what you do? And then the first thing to do is just okay, in which ways can different AI systems, agents, models assist and enable you to do that work now faster, cheaper, better, even. And there are a lot of just menial tasks that can be offloaded, like filling out spreadsheets, uh, dealing with spreadsheets and just uh documents and all kinds of stuff that people are a lot of the work that happens, I think, for most people in larger businesses, there's a lot of just busy work, things you have to do because there's no other way to do it. Filling out forms, filling out spreadsheets, dealing with documentation, updating things, and doing low-level, low-hanging fruit work. A lot of that can be offloaded to models or agents now. And so we you always start with the low-hanging, the easy stuff. The the things that took five hours can now just take 30 minutes. You start there to get people familiarized with it, to get them to use the available tools and systems, get them to not be afraid of it or even confused by it. Like there's no book you can read. Just you can't study your way to using AI. You really just have to use it with the work that you're doing. I see a lot of people thinking that they have to read books, they have to, you know, understand first how AI works and what's possible, and then they try to fit what they do and connect it with what's possible. Okay. And that's not the best way to do it. The best way is to look at what you do any given day, any given week, and then connect the dots between what you're doing and what models and uh apps and AI tools can do. Start there. So you start from the bottom up, and then while you're doing that and you spend a few weeks just getting people on board and using it, that's when you then have the more bigger picture strategic vision in place, and you start to see, okay, our team of five people are producing marketing campaigns, online marketing campaigns that has, you know, ads, emails, maybe landing pages, whatever it might be. And we're doing it this way now with AI. But what would this work look like if everyone on our marketing team have their own personal AI agent assistant to do work with them? And then what does that enable down the line? Like eventually, you can get to a place, like technically, it's possible to have autonomous agents doing work now, but what they need is access to tools, access to knowledge base, access to data and metrics. And they actually need a human being to keep giving the agents direction, input, feedback, correction, especially. Like you might have an agent perform a task, you know, create 10 landing pages, but the agent is not going to uh execute on that task flawlessly out of the gate, is gonna make 10 landing pages to the best of its current abilities and to the best that it understands your author and everything else. But you need the human being to provide edits, corrections, direction, and giving it continuous feedback. And when you have that foundation, that will enable you to do far more things and to higher quality down the road. But it always starts with you, your fingers on a keyboard, you know, talking to Claude Chatcha Bet. You gotta start there. And before you start going crazy and going, you know what, we're just gonna have 50 agents doing all our marketing and you know, connected to all the things, and just press play and hope for the best. That day will come, uh, but not yet.
SPEAKER_01So interesting, because you know, in my company, what what we kind of looked at first approach is what what about all this administrative stuff that our people hate doing anyway? Yeah. So we kind of took that same approach. And then what about the administrative mistakes that happen from just you know, just fat fingering on a keyboard or or or you name it. Yeah. Um so like how do we prevent mistakes and how do we do stuff that people don't like doing anyway, is like an easy first step for us.
SPEAKER_00Yeah. And and the good thing with these systems, agents and and otherwise, you like by default, there is a log of what they're doing. And you can make sure what you need to make sure of is that you uh have all the agents and all the things you do, just produce a log. Like no human needs to write down the log. The agents can do that work and the AI systems can do the work of producing a log of what the agent did, even down to like how did it, so to speak, quote unquote, reason and how did it it, the agent or the model decide what to do. And because that's traceable, you can have this ability into that full flow from beginning to end, it makes it easier uh to figure out where things went wrong. Often with human beings, like if a person has just knowledge of how to do a task, they're not necessarily self-reflective, they're not necessarily paying attention to how they did what they did. And so a lot of times humans end up doing things like humans hallucinate as well as models, and humans don't necessarily keep track, you know, like we have imagination. We just imagine that we did something, right? And so a lot what that helps, I think, is that in it gives you greater visibility into a task is done. And because of that greater visibility and traceability, it helps the humans on the other end stay in the loop. Like human in the loop is the phrase that people use for this. Okay. It actually, I think it helps people make the changes, make the corrections, and in the long run, it ends up the work itself ends up improving, the quality of the work ends up improving because of it.
SPEAKER_01So, Sam, how quickly are you able to make an impact? Because uh some of the stuff we're working on, I feel like has been long, long, long, long, long time coming. And and from like actually seeing it be helpful, seeing a return on investment, you name it, it seems like it's a big long timeline. How how quickly are you on average able to make an impact?
SPEAKER_00Yeah, depending on on the scope and and so on, but within uh 60 days or less, really, there's a measurable metric tracking improvement or impact of some kind that we can trace. Here's the key. We don't do any, like the first thing we do in the beginning, and which I encourage anyone to do is establish the baseline of what it is that you're doing, the task at hand or the job or the project at hand. Look uh historically, even if it's a short history, like the past 30 days, establish a baseline metrics or KPIs that you know, like, okay, for this project and this task, our KPIs are XYZ. When we've done a similar thing in the past, our results have been LMNOP, whatever that is. And that's the baseline. And then for the next 30, 60 days, you need to be rigorous and focused and only have AI do the work that it can do, that is capable of doing, the low-hanging fruit that are connected to the KPIs and the metrics. The mistake a lot of people make is they go, hey, let's just have these models and tools and agents do all kinds of work, but very few take the time to establish the baseline and then every whatever week, every two weeks, check in on the process and go, hey, the past you know, 14 days, we had three agents do this work. And the work that they did, is it at all affecting our KPIs, our baseline metrics, the foundation that we have? Instead of doing things for you know a few months and then looking back and going, I don't know, did it make a difference? Not sure. Can't really tell. You got to establish that baseline first and understand what you're working from. And it's easier than to, after a couple of weeks, go, our agents have been doing, you know, let's say produce all these landing pages, all these emails. Um, a bunch of them have gone out and we're tracking performance. Okay. You know what? Like these agents. Have done work, but we can't tell after 14, 20 days if it's making a difference. Therefore, you got to course correct. Instead of just running it to the end and after you know four months, three months, hoping that when you look back, you'll see a difference. So regular check-ins, regular measurement of performance, and then course correcting as you go. That's the key.
SPEAKER_01Okay. Well, it it seems to me though, like it's a little bit like just hiring an employee and training them from scratch. So, like, yeah, if you take the time to train them, there's gonna be a period of time where they slow you down. Yes. And then there's gonna be the the the period of time when they become really, really helpful to you and you're able to do way more as two people than you were as one. It's just weathering that in between time where it bogs you down. Is is that true? Like that's my observation. Like it doesn't help you day one because it you got you gotta learn it, you gotta see how it can help, you gotta fumble through training it. Is that right? That's what I see.
SPEAKER_00Pretty much. But it's faster for an agent or a team of agents to do so as opposed to a human, only because agents don't sleep, they don't need a break, they don't need to eat, they don't need to do anything. They just do the task that you give it. And there's still gonna be uh, whatever you want to call it, onboarding, like uh ramp up period with agents. But I think one of the biggest reasons why a lot of companies have a hard time seeing measurable difference and impact in a shorter time frame is you're dealing with two things at the same time. Okay. People who are new to AI and they don't have a ton of experience with it yet, and they just like there's a skill issue in the sense of like, hey, I'm trying to learn it and do it at the same time. And then these agents are just kind of along for the ride, and you're trying to figure out a way to navigate like what is what does an agent do? What do I do? What do our team do? So you're dealing with two things at the same time, two new things, the person or people doing it, and then the AI agents or systems themselves, which need to be integrated and connected with uh your company knowledge base, company metrics, and all of that stuff. I think as we go along over the next few years, as more and more people are learning how to do it, are figuring out how to do it, that experience will in and of itself shorten the onboarding time. And I think as people are trained and educated on this and they get hired by a company, and if they come with experience working with models and tools and agents already, for them to be ramped up, I think is gonna shorten. I think that time frame is gonna go from months to few months to a few weeks. Eventually, my hypothesis, uh, we'll see if I'm right, but I think when you have a person who knows enough about their work and how agents and AI help with their particular work, they can bring pre-made agents to the company that they're hired by and say, Hey, I have this, like, let's say, simple example. I'm an email marketer. I have these five email agents that I've used over time, the past six months at this other company. The template of those agents and the fundamentals of how those agents work, like they're 50% there. All I need to do is spend this first week with my new company to adjust and modify the agent capabilities and skills and templates that I have connected with my new work role and new access to information, get them all hooked up, connected to each other, make adjustments, and they'll get up to speed in a week or so because they've already bringing not just the experience, but all these, uh, all the prompts that you've written, all the ages that you've dealt with, all the experience you've had with AI tools and systems. It's not like you forget it, hopefully, uh, if you go move on to a new job and so on. So I think we'll see that onboarding collapse. We're still in that period, right, where most people don't have enough experience, just time, right? People just have enough.
SPEAKER_01Time's coming when I'm not only gonna be hiring the person, but the tool, the, the, the workforce they bring with them, I guess. Uh the tools they bring with them. Right.
SPEAKER_00Imagine if if imagine if you're hiring an email marketer and then one candidate goes, hey, like I have, you know, the past year I work with AI and email. I have a team of five email agents that come with me to this new job that you might hire me for. That's a pretty attractive candidate. Because they come with a tool bag and and skill set already that they they need to make adjustments, but they're they're not starting from scratch. Does that make sense?
SPEAKER_01Yeah, does that hurt the employer that they left?
SPEAKER_00Uh probably not, because and here's why agents and AI systems and tools are universal enough that what makes my team of agents productive and useful is to whatever extent the business that I'm in has given my agents access to data metrics, knowledge bases, and everything else. Like it's it's a paradox. Like agents are both, how can I put it? They're not an unfair advantage in and of themselves. Okay. But with them connected to the right context and the right data and knowledge bases and tool set, they become an unfair advantage because they're like they're like multipliers of work. If you have, you know, uh one times zero is zero, but one times ten is ten, right? So agents are multipliers that get plugged into a situation of work and they multiply the uh they can multiply the outcome, the effect, and the quality of the work, but in and of themselves, connected to nothing, they're not gonna, they're not gonna be productive and effective. So it's it's this weird paradox where they're like they're great, but in and of themselves, like if I take my agents, it's not like if I work for a company and I have those five email agents, if I go move on to another company, the company doesn't lose because it's not like I unplug my agents and then they miss out. Like the company will own that, like it's part of the company IP and comp company uh work product that exists, but the templatized version of the agents, the skill set of the code. So you're taking the clones with you.
SPEAKER_01Yeah, you're taking the clones with the case. Yeah, exactly. Something like that. Okay.
SPEAKER_00The other ones are staying behind. And so like no one loses, really. Okay. It's just that the person worth doing the work can now just be more effective faster.
SPEAKER_01Except for it could be looking like at it, like, okay, you worked for for me and you built that stuff on my dime, yeah, to help my business, and now you're gonna go help another business shortcut it. Like it could be looked at like that.
SPEAKER_00It could, but that's that's true of any skill. Like if I'm an accountant, people do it all day, every day, right?
SPEAKER_01Yeah.
SPEAKER_00People do it all day, every day. Like it's just a it's it's the same with any technology. If I learn how to use an email marketing platform at one job, I can't lobotomize my brain and like remove the experience of working. You know, like it's nor would anybody wants you to.
SPEAKER_01It'd be ridiculous to be expectation.
SPEAKER_00Right. But you know, with a Neuralink, right?
SPEAKER_01Like, just wipe it clean, like maybe a couple of things. Like the Men in Black movie, like you're gonna forget all this ever happened, look in the light. Exactly. Oh, that's funny. So look, what I heard then as a user, what you need to be good with it is to know how to prompt it and make sure it's got the right data to feed it.
SPEAKER_00Context, data, connecting it all, and then knowing how to like knowing that let's say you have this log like we were speaking about earlier, like you have this team of five agents, and after a week they've done work, you as a human as well, you need to be able to look at that log and the work they've done and be able to tell, okay, when an agent tried to do this task, like let's say write 10 subject lines, like they got stuck in a loop and they just didn't produce good subject lines. Your skill set as a human is looking at that work, being able to determine whether their work was quality or not, and then figuring out, okay, what do I need to adjust in the configuration of these agents so that they one, don't get stuck again, and then two, that the work that they do next time is of higher quality. So that whole skill set of being able to look at things being done and going, you know, with my experience and my knowledge as a human being, my the what I know about my my work and what I do, how do I adjust the agents to then do better, not get stuck, and then produce better work?
SPEAKER_01So if you hire me as an AI beginner and um you want to teach me how to how to prompt better, what what do you like? Can you give us a quick lesson or anything that might be practical?
SPEAKER_00Yeah, the the the number one thing that I will tell you to do is to practice and study human communication and practice good communication.
SPEAKER_01Which none of us uh well, I don't want to say none of us, all of us like suffer with like needing to learn that better.
SPEAKER_00But that's the key, because the way you adjust and like you're we're at the point now where um if you know code, it's helpful, but you can make agents and have agents do work without touching code, but you need to be able to communicate clearly and have good human communication skills because all of that work is happening through natural language. And like all of these companies and labs, they're all working towards a place where anyone can talk to any AI system or agent and things get better. It used to be that you needed to know code and API calling and all these technical things to make it really produce good quality outcomes and work. That's not helpful now, but it's not going to become the required thing anymore. So for you as a human, like the the paradox is as AI advances and becomes more capable, the thing that'll help you learn to use it and use it better is to just become more of a human being, communicate clearly, understand nuances in language, and understand that like like it's like have having a team of five agents is almost like having a team of five people. They're not humans, obviously. They don't think like humans, they don't reason the way humans do, but it's kind of like that where you you go, okay, these agents can sometimes, even if you instruct them and you have everything configured correctly and set up, they still veer off the the reservation sometimes and they still do crazy things. But the way then to correct it is to have better instructions, maybe changing some of the configuration and what tools they have access to and making use of memory and and other uh other capabilities. But it's really just clear communication, understanding how to communicate something to another being, another human being, because that language that you use to communicate to human beings is the same language you're using to communicate with agents. They're being made to, for better or for worse, they're being made to emulate human behavior. Okay. And so the more human you are in your communication, the better they perform.
SPEAKER_01That's a that's a great tip, I think. Question for you. I I'll go to different CEO like peer groups, and you know, AI is a hot topic now for the past couple years. And so, you know, one peer to another trying to give ideas and suggestions how to how AI can help the business. And in every one of those groups, there's a a number of them that bring in somebody else for AI outside of their say their CIO or their IT team. Now I feel like our team is doing pretty good where I don't I think I and I wonder I leave there wondering if I'm the exception or if I'm just blind to what we really could be accomplishing if I brought in an outside AI like yourself to to do it. Why why do companies feel the need to split it and why are why are like the the normal like CIO group not the ones to to lead the AI push?
SPEAKER_00I think we're still and who knows how long this period lasts, maybe another couple years, but we're still at a place where the experience and the knowledge mostly exists outside and it exists with people who have spent the past at this point five years uh or you know, three to five years working with it. Okay. And like I said earlier, you it's helpful maybe to understand some theory around it. It's helpful maybe to read a couple books around it, but it's not a thing that you can study your way to learn and do at a high level. And if you were to teach this in in college or university level, like after six months, everything you just taught is kind of out of date. So it moves so fast and the capabilities are expanding so quickly that you really need someone who's working with it like every day, basically, to understand, like develop that intuitive sense of how this technology works. And like we've had this period now of uh SaaS and software and tech that most of it has remained almost the same for the past 20 years. Like, sure, like code is maybe cleaner and better now. There are more capabilities uh in terms of just tech outside of AI. But when you look back at the past 10, 15, 20 years, at the end of the day, it's mostly been a change of degree as opposed to a change in kind. Okay. AI is a change of kind, it's a different, it's tech, and a lot of it is similar to what already exists, but it's becoming it's almost its own unique, separate thing that it's hard to know what's possible, know how to make it valuable to the bottom line, unless you're immersed in it right now. Most people don't have the time. They have their jobs, they have their tasks and they have the work that they're doing. They don't have uh another full-time job available to themselves where all they do all day every day is just figure out AI and work with it. The people that do, like people like me who spent a few years doing this, we just have that advantage of having enough experience and knowledge. But I think give it another couple of years as people learn to do this, like that gap will go away. And so I think right now we're just in that period where you need probably you need someone from the outside to come in and help you come alongside you and help you understand and help you figure this all out. Okay. Um, but that becomes institutional knowledge, like as teams work with it, as the CIOs work with it. So it's a part mentality question, part just experience level question, part capability understanding question too. And my guess is that uh people like me, like there won't be a need for people like me uh at a large scale in maybe two years.
SPEAKER_01Oh, that quickly.
SPEAKER_00At that point, that quickly, yeah, 100%. But in two years, if I you know my assumption is I'll be doing this, right? So unless get hit by a bus tomorrow, right? But in two years, there'll be other capabilities that are at the forefront that just other people don't have the time to keep up with, right? It's beyond a full-time job to try to keep up with AI as it is. I would I don't even try anymore. I just have agents do it for me.
SPEAKER_01So that's that's great. Yeah, I I would say like it it almost seems to me like even if you haven't done anything with AI, like you know, the phrase pioneering doesn't pay, like there's a lot of companies that probably spent a lot of money and time trying to be at the forefront and really not gotten ahead of people that much is is my bet.
SPEAKER_00It's a speed thing. So large organizations move slower than smaller organizations, just by the virtue of like enough people being involved and maybe you have more decision makers and stakeholders being involved. And so this is the AI is still at the place where speed rewards more than anything else, right? Like the fat the the more immersed you can be and the faster you can be at adopting it, the more at a the more advantage you're at. At the same time, that advantage is only useful if it is applied to other business contexts, right? And other situations and other other work that needs to get done. So it's this weird mixture. And I like the best way I can, the best way that I've uh understood it is like I said, maybe I have another two-year advantage at the same time. Uh, within those two years, other capabilities will reveal themselves and become available that the the frontier I think will keep uh I don't know what the word is, like it just keeps getting pushed into the horizon. Okay, and unless unless you have, like, unless you go, hey team, you all need to become frontier AI experts. You know, it's just there's a there's a practicality to it as well. Like, I don't think everyone needs to be at the forefront. I don't think everyone needs to keep up. It's not necessary. Uh, because like the hype and the horizon and the forefront will always be uh, you know, 10 miles ahead, but your job happens now and what you're trying to get done happens now, and you'll only ever be able to use the capabilities that exist now as opposed to the hypotheticals that might come five years from now. It's this weird like dynamic, I think, between timing, speed, availability of expertise and knowledge, and where that line keeps moving. So it's uh it's a fun time to be alive. It's also like uh exhausting at the same time.
SPEAKER_01So I think so, right? It's a big mix of optimism and sc and fear and all of it all at once.
SPEAKER_00Um thrill and terror.
SPEAKER_01Yeah, all of it. What about do you really think like I think part of the thing that we have to solve when when trying to use it is proprietary information staying proprietary? Do you really think that that's possible? Like it seems like if you're feeding somebody else's software or AI or whatever you want to call it, all your data, like is it really safe?
SPEAKER_00Yes and no. Like mostly no. And so um this but here's the thing. What we think of as proprietary knowledge and data and everything else actually probably isn't. I think a lot of companies think they have proprietary data or context and information. That's how I feel.
SPEAKER_01I feel like we don't have the form, I mean, unless the formula for Coca-Cola is an AI somewhere. Like we don't have it. Like there's there's nothing that secret. And if there if it is, it's secret for like 10 minutes. Like that's how I feel.
SPEAKER_00But exactly.
SPEAKER_01But I don't know if that's reckless or not, like to think that way.
SPEAKER_00I don't think it is reckless. I think in in in some industries and for some businesses, like Coca-Cola formula, you do have something that is unique and proprietary. Most businesses don't. And here's like here's the I'm gonna try to explain this uh because I haven't even been able to formulate it clearly to myself, so I'll try to ramble for a minute and keep it short. But the secret and the advantage is no longer in knowledge. For a long time, it's been if you know something that the rest of the market doesn't know or your competitors don't know, then you've had an advantage.
SPEAKER_01Okay, this is getting good.
SPEAKER_00That's no longer that's no longer true. There's no longer uh an unfair advantage in information that you may or may not have. Okay. Unless, like, we're talking a patent or something that is very uniquely right.
SPEAKER_01Or an insider stock tip or something, maybe.
SPEAKER_00Insider stock tip, right? Like that still has value, especially in in prediction markets. But anyway, so but most people think that okay, our business knows how to do the work that we do, whether it's a product or a service. And so uh they think that the knowledge somehow is the unique thing. Okay. Um maybe it was for a time, but that time is long gone. Like, I'll give an example. Uh, some of the work that we do is for other companies is helping them build competitor monitoring systems and intelligence systems that collects information about open source, stuff that their competitors are doing. And when you give a model enough context about a competitor, which also includes you, but for a competitor, the models and the systems can reverse engineer most of what they're doing to the point where even if they had some secret knowledge, a sufficiently uh configured agent team or AI model system can reverse engineer whatever it is that they might be doing behind closed doors. Right. And so that being the case, I think the companies that'll uh I think capture an outsized market share in their in their uh market are the ones that go, you know what? We probably don't have enough proprietary secret knowledge. We probably never had it. But the key now is how do we utilize AI systems and models to do the one thing that'll matter the most? There are two things left distribution and capital. You need to focus your energies and task and work that you do on ideally both, but distribution especially, which is just another word for market share.
SPEAKER_02Okay.
SPEAKER_00So I recommend all the companies I work with, like, don't worry, like unless you really have something super secret, which very few do, but don't worry too much about sharing your data and your context with these models. They could they could reverse engineer it if they wanted to. So you're not doing anyone like the the only one you're hurting necessarily is yourself, probably, because you're slowing down on your execution speed. And instead focus on how much distribution can we capture, how much market share can we capture as fast as possible, and what's the best use of our capital deployment and how we invest in what we're doing? Because the two things that are gonna matter the most for the at least the next five years is distribution and capital, as opposed to knowledge. As opposed to secrets that you may actually never had. Does that make any sense?
SPEAKER_01It does, yeah. I I I appreciate how you explained it, because that basically is saying a lot of times we worry about the wrong thing. Imagine imagine that. Yes. Yeah. Yeah.
SPEAKER_00We hallucinating, right?
SPEAKER_01Curious on your thoughts on where should kids be on this? Like in school? Like where should they should they be using it? Should they not? Like I just don't know. Like I have young kids. My kids are my daughter's about to turn seven, my son is nine. I I've seen where there's schools now that at their age start to push, like not the normal public school system, but there's there's a handful of private schools out there that are like specializing in this stuff. Is that what kids should be doing now?
SPEAKER_00I I think if you use the right way, it can be an excellent educator, tutoring, uh assistant in in so many subjects. Like, man, um I'll tell you a secret that uh now becomes public. I'm awful at math, like terrible at math. I all throughout high school, school, like I struggle with math just beyond what was uh I love math, I can't spell, but got it. Right? So like math to me was like my nemesis. Like I just could, for the life of me, no matter how much time I spent, and people try to help maybe awful at math. But I started using uh Chat GPT and Claw to help me relearn math, basically, the past few years. Really? And now, now I'm like, I wish I could go back in time and do math with an AI tutor because I would have probably aced it. I'm not trying to like brag or anything, but now like I finally understand math. Really? I can finally do math. I can find like it's so it didn't do it for you, it helped you learn how to do it. It helped really exactly. I use it as a tutor to help explain things and to quiz me back. Like instead of me prompting it, I made it prompt me instead, right? And like so using it as a tutor and an educator to help explain concepts, to help me understand, to give me quizzes, to give me things like um examples to look at, to give me tasks and to give me homework. Uh like finally, I understand that at a decent enough level. I think that's a key. Um, I do want to stress, I think it's like this another paradox. Like the more in the world physical things you do, I think as a as a kid and growing up, even for adults, but as uh with the question about kids, get them involved in the real world hardware, hard touch, like get them to plant a garden, like if you have that uh available to you to do. Show them how to fix a um a flat tire, get them involved in the physical world and doing things in quote unquote reality as much as possible while using AI in a tutoring kind of a way. I think that's the key. It's um the danger is when all of our attention and all the things that we care about only exist on a screen on the internet. Like that's there's good things about the internet and there are bad things about the internet. I think the key is like don't let it consume you. Like, don't let your 10 hours a day watching Netflix or YouTube give yourself limits and like spend as much time outside and with other people, talking to people, doing things with your hands and trying things, take dance lessons. Like I can't dance worth anything, but I like I want to do the bodily things, you know, being physically engaged in the world. I think that's key. And and people uh that I speak with, and there are, you know, in Texas, like there's the alpha school that have uh every kid has basically an AI tutor that's alongside them.
SPEAKER_01That's one of the ones I was talking about. Yes.
SPEAKER_00Yeah. And the kids who do that in that way, like they're excelling, like they're learning faster. They're um some of them are jumping uh grade levels because they're like they're they're getting, they're just they're learning in a way that works for them, while they're also have they also have a huge focus on the physical in-person stuff as well. So they're not sacrificing that, like they're emphasizing it.
SPEAKER_01Can I price Sam? Do you have kids? I do not have kids. If you did then, hypothetical. It's a big hypothetical. Would you tell them to use AI? Or what if the school says no AI? Would you tell the kids I want you to use AI, just don't get caught? Or would you or or would you tell them to not use it? What would you do?
SPEAKER_00I would tell them to use it and I would make sure that they're using it in a way that it acts as a tutor and not as an outs outsource way of having the AI think for them, for them. And I would probably like have limits to how long the like screen time. I'm sure every parent struggles with it. Like my siblings who have kids, like that's their that's a struggle for them. It's monitoring screen time and everything else. So my older brother, he has uh four kids. Uh the first one is a teenager now. Uh-huh. And so he's he's using Chat GPT, but he's doing it in a tutoring way. And he's only using it for a couple hours a day, and he's not using it to like do my homework for me. I'm sure maybe on his own, he you he's using it to cheat on maybe homework. But at the same time, like, I don't think the answer is for someone who does not have kids, or take this with a grain of salt, right? I don't think the answer is don't let them touch it and just keep it away from them. I don't think that's the key. I think better use, healthy use of any technology, I think is a better way to steer your kids as opposed to going, don't ever touch it, like don't never use it for anything. I think my theory as someone without kids is maybe it's better to teach them good habits, healthy habits around technology than it is to just prohibit them completely. I don't know. What do you think?
SPEAKER_01Yeah, I I I think you gotta be able to, it's it's crazy not to learn how to use tools that are available. Yeah. But I I don't want it to hinder their learning ability. I want it to help I love the way you described it, to use it as a tutor. I want to accelerate their learning ability, not not make it to where we all get dumb because all we gotta do is have the computer do it all for us. Like I that you because you want to learn thinking ability and pattern recognition and things like that, like that uh that'll help you in life.
SPEAKER_00Yes. And like I don't like, you know, on the one side of this debate, yeah, people going, you know, AI is gonna replace all the jobs and you know, maybe not plumbers yet, because they deal with the physical world. Right. And maybe but maybe eventually humanoids will do it. Like, I don't believe that we're gonna go to that extreme because at the end of the day, humans are humans and we work at human speed. Like, even though the AI capabilities are far above and beyond what most people realize, still people are go to work today and they are in an office or wherever they are, and they still do work, right? And so the way that humans adapt and adopt to technology is always slower than the speed of technology. And it like we live in a world of humans, and unless this is the matrix and we're all in a in a pond somewhere, like the reality is that the world runs as a human system, and the only way to remove humans from the system is to just eliminate all humans. And I don't think we'll get there. I don't think that'll happen. I think we'll find uh it'll be a mixed bag. There'll be a lot of bad stuff. There'll also be a lot of good stuff, and it's just gonna be a mixed bag of experiences as we figure out how to live with this and how to work with it.
SPEAKER_01Do you think that I can exist and and not touch it? Like what like uh is there a group where at my age I don't I don't I got this far without it, I'll be just fine. Do you think that that's a dangerous mindset to have?
SPEAKER_00I don't. Okay. I think that I think that's interesting. I think you can't. Yeah, I yeah, I think sure, like can you exist in the world without a cell phone? You could, but it's hard, right? And I think with AI as well, there may be situations where it's impossible to ignore it. And you know, there will always be, I'm sure, an AI somewhere. Like Chick-fil-A, like they have an AI taking order, right? So if you want to eat Chick-fil-A, you're gonna have to talk to an AI agent.
SPEAKER_01So it's around you all the time, but yeah.
SPEAKER_00It's around you, but I do I do think that it's possible to go, you know what? And I like there are some CEOs I work with who are you know in the late 50s and they'll they're eyeing up retirement and other VPs and so on. They're like, you know what, I'm I'm 60, I'm getting close to like my retirement and so on. I just don't want to do it, and I don't want to have to learn something new again. And I said, and I always tell them that's a perfectly viable position to take. And you don't have to. You just gotta make sure that the people who are younger than you and are on your team like actually work with it, right? And and don't don't cut it don't cut it off from everyone else. But I don't think that AI is gonna be so required that you can't live life or have things to do or get paid to do anything if you don't want to. Like, I I I hear people say that, and people are going, it's gonna be so ubiquitous and it's gonna be so uh uh demanded and required that like you're tough luck, like you're not gonna be able to function without it. I don't think that's true. I I think that there'll be jobs and things to do that won't require you to use it. And uh peep like when it comes down to it, people want to interact with other people. People want to have human connection and human relations and human interaction. That's a fundamental part of our biology. Like, you can't just delete it just because AI came out to the scene. It'll look different, it'll be a bit different maybe than the past 50, 100 years have been. But I I I want to stress like life does not have to be dictated by AI. The thing about AI that blows my mind, and it's the reason why I'm excited about it. Finally, for the first time in I don't know, a hundred years, 150 years, AI can help us do technology on our own terms.
SPEAKER_01Okay, interesting.
SPEAKER_00It's always been that we had to sit at a keyboard, look at a screen, and do things according to the requirements of technology. Like technology has dictated to us for the longest time how we do what we do. But I'm like right now, you can, with some elbow grease and work, you can have agents do things for you on your behalf on the internet that enables you to log off and just not touch it. And you check in with your agents once a day to see if they've done what you task them to do. It's gonna be possible, it's already possible in a small scale, but as capabilities grow, it's going to be possible that if you don't want to be on the internet, you don't have to be.
SPEAKER_01Interesting.
SPEAKER_00Your agents can send you an email every day with an update and you reply back and say yes, no, and change this. And agents can work on your behalf online. Like the internet was not made for human beings. It's not a human environment.
SPEAKER_01I can say it was made for technical. So you can go and ask it to tell you, give you an update on the things you know, and then you don't go down the bunny trail of bunny holes because of something that popped up that you want to chase, huh? Save a lot of screen time that way. Tell me what I need to know without without me chasing something I don't need to know.
SPEAKER_00Yeah, exactly.
SPEAKER_01Like there's or the clickbait, get past the clickbait.
SPEAKER_00Right. And like FOMO and our like we have this. Um people think that I just if I know more and have more knowledge, then things are better. And that's not true. Like what we know about how models are trained and how agents work. The more narrow you keep an agent, the better it performs. If you give it access to everything there is and all the knowledge and all the context and all the things that it that it could do, it goes off the rails. You need to keep it focused. You need to focus its attention on just a narrow band of things to do. Same thing for us. Like, I the more I use AI and agents, my like my ultimate uh thing that I'm working towards, which I'm getting closer to, is like I want to be able to on a Monday morning tell my agents to do things for the next two weeks. I close my laptops, I don't touch the internet for two weeks, and they just do stuff for me, and it works. Huh.
SPEAKER_01You have to keep me posted on that.
SPEAKER_00I will, I will. The longest I've had them do work now is like four days without running into issues. So technically, I can have agents do things for me for three, four days. I could close my laptop and just say, I'm out, and I don't know, go sit, go talk to my friends, go hang out with my wife, go do things, and then come back after four days, and they've done stuff. Sometimes they run into issues and like they get stuck or whatever. But that's now in May of 2026. This thing is not slowing down, right? Right? Like capabilities are getting more robust by the week. And we're gonna get to the point where they can work on their own on a specific set of tasks for days on end. And if you don't want to be involved, you don't have to be involved.
SPEAKER_01But you made me feel a lot less scared of it and a lot better about it in this short conversation. Do you it's all choice?
SPEAKER_00We we have the choice to make early on. We talked about which way is it gonna go. If we make the decisions for ourselves, and like I can't control the world, I can't influence what Athropic does or OpenAI, but I can't influence and make choices about how I use it, how my family uses it, how my friends and my employees use it. And I'm focused on that little world because I can something there I can make, I can do right. And I'm hoping to stare at that direction.
SPEAKER_01Just like any tool, though, people will use it for bad. And that that does worry me because I can the good outrun the bad is what that seems to be a game of, probably.
SPEAKER_00Yeah. And that's an open question. And I'm I'm hoping to influence people to make the right choices, you know.
SPEAKER_01So kind of out of time here, but quick quick quick couple questions. Um what AI tools do you like the best for what nowadays? Because it seems like that changes too, as to who's in the lead and who's better. But maybe just some quick tips um as to which AI to use for what.
SPEAKER_00Yeah, so um I'm laser focused on agents, and for agents, there are a number of agent frameworks that you can use. Two of my favorites are uh Paperclip and OpenClock. And these are like frameworks, agent frameworks you can find on GitHub, so they're free. Um I use paperclip the most. It's uh it's like um uh it's an agent orchestrator system that enables you to create as many agents as you need for whatever task there is and you can have them do things. Um in terms of models, I become a big fan of Claude models, so Opus, at this point it's 4.7, I think is the latest one. Opus and Sonnet, because like their anthropic and Claude have been focused on the business side of things, whereas ChatGPT for a long time were focused on the consumer side of API, right? So they kind of focus them that way. ChatGPT models are still very good. Codex, uh the model codex as opposed to GPT models, but the codex model uh we use all the time in my business. And that's more for like uh developing coding uh analysis and dealing with complex information and getting stuff done with that. So codex models and opus models, and like the way we work is um most AI platforms that you see out there, they all use the same models underneath the hood, right? They all use Opus or Sonnet or Codex or GPT or any of the other image models. Like most of them are just layers on top of the core models. So at the end of the day, if you use a particular software or app, that's really a preference question because the models underneath are the same, and all these other platforms use the same models, they just give you a different interface to do it. But I would like Opus for me is an excellent thinking partner if you use it as one, like to help you reason and to use it as a collaborative engagement. Um, Opus is great, Codex is great, and for agents, paperclip and open claw is what we mostly focus on. But there are so many, like, if you don't want to deal with code, you don't have to. There are things like Gumloop, Lindy, and a bunch of other platforms that give you the agent capabilities that all you do is click buttons and set it up. You don't have to touch code. So I, you know, a lot of our clients they use something like Gumloop or N8N to orchestrate and deal with agents without touching code.
SPEAKER_01All right. Thanks for those, thanks for those suggestions. How can people get a hold of you? Where can they find you?
SPEAKER_00LinkedIn. If you search Sam Woods AI, then I'm probably one of the first to show up. And I'll provide a link for that. But also my company is Daringrobot, daringrobot.com. We do our own projects and it's like a boutique AI, applied AI studio where we do a lot of research, exploration into agents, mostly for our own stuff. But every now and then we work with clients and we do things like training, education, uh, consulting advisory, and also if if the if there's a need for it, we also build out agent systems as well.
SPEAKER_01Love it. Thank you so much. So I got a closing question we always ask everybody. We're all about championing people here. And so if before we let you go, who's who's someone in your life, given the platform like you are now, that you would like to champion and why?
SPEAKER_00My wife. She deals with the pipelines of data that make all the other AI systems work. And so without people like Hurst, the data engineers and so on, without them, there ain't no model that's going to do the work for you. So the infrastructure, the deep underground pipelines of data is not sexy, is not exciting, but it is the work that is required and that actually makes AI do anything useful inside a business. So my wife and people like her who do that plumbing.
SPEAKER_01That's awesome. Thanks so much for spending the time, Sam. I enjoyed this very much.
SPEAKER_00Thanks for having me, man. I appreciate it.
SPEAKER_01I hope you enjoyed this conversation. If you want more podcast in your life, which of course you do, make sure to like, follow, and subscribe to us on YouTube, Apple Podcasts, Spotify, or wherever you get your podcast. Catch you next time.