Episode 104 - Pat Barry

Erik Martinez:Hello and welcome to this episode of the Digital Velocity Podcast. I'm excited about this topic today, and I have my friend Pat Barry back on the show to discuss building individual capability with AI.

Erik Martinez: Pat, welcome back to the show.

Pat Barry: Thanks for having me buddy. How are you?

Erik Martinez: I'm doing all right, man. Feels like we're starting to do this on a regular basis.

Pat Barry: Yeah.

Erik Martinez: I'm totally excited to have this conversation with you and have some fun while we're doing it. I think when we talked in the lead up to this, you said we were gonna body slam this topic, and I'm [00:01:00] not sure what you mean by that.

Pat Barry: Well,in your comment on LinkedIn, you said you were just gonna tackle it. I said, you know what? Let's take it to another level. We'll body slam this bad boy, as opposed to just tackling it.

Erik Martinez: Body slam this bad boy.

Pat Barry: We got it. The changes in AI are so extreme. We might as well take it to another level. Something new every day.

Erik Martinez: I totally agree man. So we're gonna wrestle this topic, hopefully, give you guys some good tips and ideas to think about.

Erik Martinez: So Pat, just kind of in the run up to this conversation, I've been thinking a lot about individual capability as a way to build team capability. When you're talking about the journey of learning how to use AI every single day, and you go back to the very beginning, when you knew nothing,

Pat Barry: Yeah.

Erik Martinez: How did you think about building your individual skills to get here?

Pat Barry: Or were you born like that?

Pat Barry: I wish I was, no. You know, early career, pre AI it was, watching videos on YouTube, Googling stuff, creating a strong network of [00:02:00] friends that I could rely on for help and ask questions to. I feel like now, one, it's changing every day, but even just how I learn now is I go to an LLM like Gemini or Chat GPT.

Pat Barry: They all have some sort of guided learning, and I just tell it what I wanna learn, and then I tell it, ask me more questions so I can, kind of tailor this to myself. But, just going and doing it. I'd say the going and doing it is a big part of like early days too. Especially as you're, in the tech space, even the data science world, it just changes constantly. Explore new tools every time they come out. Try new things. I think that's part of it. So it's mostly just going and doing and reading up as much as you can in the past, but now kind of asking AI and then going and doing, and executing. I think that's the biggest thing. Best way to learn from me at least.

Erik Martinez: Yeah, I totally agree. I think in order to get better at doing anything right, you have to practice.

Erik Martinez: But I think, one of the biggest challenges for everybody out there is time, right? There's still a bajillion things on your plate. Your work [00:03:00] schedule doesn't get any easier. At least mine doesn't seem to get any easier. So, how do you find the time to squeeze this in with a very busy work schedule and start to practice? Because I think, that is kind of the number one issue for most people is time.

Pat Barry: Yeah, I think it's a challenge for me too, I think just trying to keep my day as structured as possible. I'm up early. Some days it's, laying in bed reading news for an hour. Then, I work from home, obviously I get up and I kind of just sit down and start going. I'm lucky enough to have some clients that are at least trying new things, and so I kind of learn as I'm working. At the same time too, I use time boxing. Like I'll block off just an hour on my calendar. Set an alert for it. So, like this afternoon, for example, I have to work on some training modules for a client. So from 2:00 to 5:00 PM this afternoon, that's all I'll be doing.

Pat Barry: I'll get a little notification on my calendar. Time to start in five minutes. And sticking to that. [00:04:00] So, time boxing, that helps me. And just discipline in general. Like I love the stuff. it doesn't feel like work, which is good. But again, that's first thing in the morning till typically last thing at night. I have wife, kids, family in between. So you carve out time to spend time with them. It's really just discipline and just keeping time on your calendar for this type of stuff.

Erik Martinez: Yeah, like to frame it a little bit like, I used to coach Fast pitch, and I used to kind of tell my players in order for you to get better, like you gotta do more reps, whatever it is, whether it's hitting, fielding, catching, throwing, running the bases, like for you to get better at those specific things you just had to do more of it.

Pat Barry: Yeah.

Erik Martinez: And the coaching is trying to help you refine it, but you had to do the work of, Hey, I've gotta take a hundred ground balls 'cause I'm really struggling with the, with the short hops. Or I'm really struggling with, catching the fly ball when the sun's in my [00:05:00] eyes.

Erik Martinez: Right? I mean, those are, those are real things. And in our, I think in our work world, we don't always look at it that way where, you know, you take that time box and say, Hey, you know what? I am going to work on getting AI to help me solve this specific problem. And I'm gonna just ask it questions and have it ask me questions.

Erik Martinez: So I think that's, a critical thing as we move forward.

Pat Barry: Yeah, and that's again, asking it the right questions, I guess detailed wise, I'd, I would do the same thing. I think some of it though too is just building your confidence with it. And just being confident in the outputs and like understanding it. Again, in like a larger organization, how would you go about, building confidence either with yourself, with the tools? I mean, I've had to do that too, but also with people around you.

Erik Martinez: Yeah. You know, for me it was really just starting to do it while I was doing the work. I'd have a particular problem. For example, I had a client that we did [00:06:00] some light web maintenance on their Shopify store.

Erik Martinez: Right.I really hadn't worked in Shopify a whole lot, so, I wasn't really familiar with the structure. I wasn't familiar with where the templates lived. I wasn't familiar with, the back end. And pretty much what I did was say, okay, well I'm trying to do this in Shopify. Where do I start? And then, as I encountered a problem. I ask it the next thing, like, okay, I'm encountering this problem. And what you learn by working on these things, and that's a very tactical scenario, but what you learn is that the AI is pretty smart and has a fair amount of knowledge that can guide you through a problem if you know what questions to ask.

Erik Martinez: And learning what questions to ask is part of the process. But all of us are experts in something. So in the things that we're kind of have expertise, then we can ask deeper questions and have the AI interrogate us.

Pat Barry: That's a good point. I think for me [00:07:00] it's just more trust. And trusting the AI 'cause you know, I'll have it help with tasks that I'm familiar with and expect the outcomes for. I guess more trust for me, but that's a good way to go about it.

Pat Barry: I mean, you have to, build that trust and confidence in these tools around you, especially as it's changing so quickly. I think that's hard too it mixes with judgment a lot. Like, is this the right way to do it? is this the best way? What are those human in the loop skills, that kind of build up to this.

Pat Barry: I know I get challenged with that a lot too.AI handles a pretty good amount of my outputs in terms of that, but, I still, at the end I'm like, all right, I have to put the final stamp on this because client or whomever isn't going to chat GPT or Gemini or Claude going like, Hey, this output sucks.

Pat Barry: Like, what'd you do wrong? They're coming to me and I don't wanna deal with that. So I struggle sometimes with just like the human in the loop. Where do I need to insert myself? How are you measuring that type of stuff? Or where do you insert yourself depending on the project?

Erik Martinez: Yeah. You know,I've been doing a bit more writing and, Chat GPT and Gemini are [00:08:00] my co-writers. I keep hearing I should be spending more time in Claude now too, but, sometimes it's a little overwhelming to tackle all the tools. But where I insert myself, you know, one of the things I've actually learned to do is build, as part of my process, synthetic panels of personas to help me evaluate what I'm doing and they give me feedback on what I'm trying to output. And what that allows me to do is look at that feedback and go, oh yeah, you're right. I do agree with that point, and I will make that change. Or, no, no, no, no. That is totally off base and I'm not gonna do that 'cause it's just dumb in my opinion.

Erik Martinez: Or, or it doesn't align with my values, whatever. So I think you have to insert your specific knowledge where it's most important and impactful. Right? For example, if you're pulling quotes. You're going to use a quote, you need to check the [00:09:00] quote and make sure that the AI didn't make up the quote. That was a real problem a year ago.

Pat Barry: Oh yeah.

Erik Martinez: Today I think it's better, but I still find where it will take a quote and then it'll paraphrase the quote and treat it like a quote. No, it's fine to paraphrase a quote,

Pat Barry: yeah,

Erik Martinez: right? But don't treat it like a quote then.

Pat Barry: Exactly. You gotta treat it as a paraphrase.

Erik Martinez: You gotta treat it as a paraphrase. So, I think it just depends on the stakes of what you're doing as well. Right? If it's pretty low stakes, like this is all only something I'm going to use, probably not as important as it is when you're delivering work for whatever you're doing. If it's a campaign strategy that you're presenting to the CEO. Check your work, right? Check the work and make sure that it's good.

Pat Barry: For me, I guess, being a data science for, so it's a normal part of the process. But yeah, man. How do you measure this stuff?

Pat Barry: You know, that was good output. I'll thumbs up. I know I get asked that a lot. A lot of the times when [00:10:00] people are coming to me, like, what's the ROI of this? And I'm like, okay,we can figure out how many hours this took you to do before, calculate the new hours, the cost and all that. But, I start to think too, is there like a different measurement of brain peace? Like you're just more mentally at peace now. A lot of my anxiety and fear about anything is gone just 'cause I know, hey, I can go pick up this LLM.Sure, I can marry metrics to what I'm doing, but do you think, are there other like non like math based metrics, like hard numbers that you think about?

Erik Martinez: Well, I don't know if there's a non-math based metric out there. Right. But I think, when you think about measuring this, it depends on who you're talking to, right? What is it you're trying to accomplish? At the end of the day, we need the focus on the outcome. So, if the outcome is, I need a way to reliably generate a draft of a blog article for a client on X, Y, Z topic.

Pat Barry: Yeah.

Erik Martinez: Twelve months a year. [00:11:00] If I need a way to do that. Then, the outcome I'm looking for is, Hey, I need a draft that's 80% good that I can just spend an hour or two editing to get it to the point where it is publishable. And I need it to be able to do that on a reliable basis so that when I feed it the context. I get a consistently good enough result that I can take that work and move it to the final step.

Erik Martinez: So I think, you have to measure these things in terms of what it is, the outcome is. If the outcome is speed.

Pat Barry: Yeah.

Erik Martinez: I'll give you a good example. I was playing around with lovable the other day.

Pat Barry: Yeah. You told me!

Erik Martinez: And you saw kind of what I generated in a couple of hours. I built a landing page and a database and all this stuff. Now, I still have to test it, but the concepting of that might have taken me a day or two. And I just spent a couple of hours. The outcome I'm looking for is a productionable [00:12:00] thing. So, at the end of the day, the AI helped me build the structure fast and in what I would consider pretty good first state.

Erik Martinez: Where the adjustments are pretty minor. I did that fast and I needed to do it fast because I've got an event coming that I need to get this page up for. Right? So I think it just depends on what you are trying to shoot for. If you're only looking for efficiency, we can do that. You can do that all day long.

Erik Martinez: You can find little things in your daily work that you can speed up. Whether it's, responding to, certain types of emails or some of those obvious things, but maybe the non-obvious things are to focus on, can I build something new?

Pat Barry: Yeah.

Erik Martinez: Can I, expand my capability? Can I do things I couldn't do before? I'll tell you a year ago, if you had asked me that this year, I would've written three throw away software programs to do a very specific task. I would've [00:13:00] said, you're nuts. I would've had to hire out a developer, write up all the specifications,

Pat Barry: Yeah.

Erik Martinez: I still have to write all the specifications. But, I can do it, and within an hour or two, I can have a concept. A program that actually does something I need it to do. And I may never, ever, ever use that software again. But the process of creating it, I now know how to do pretty confidently. So I built the capability.

Erik Martinez: Let's pivot off of that. I think one of the things that, I believe, and I'm curious as to what your thoughts are. I think individual capability is kind of the bedrock of team capability. You can't scale your team without another.

Erik Martinez: And I'll give you an example, many, many years ago, early in my career, I worked for a company that was in the process of building their first marketing database, and this was like in the nineties. What they decided is that their existing team was so busy, they hired a set of [00:14:00] consultants to come and build this database for them.

Erik Martinez: All that knowledge of how to build it, to construct it, how it worked, was retained in this team of consultants that never worked for the company. So the company never actually retained the knowledge of how to do this and do it better for themselves in the future.

Pat Barry: So, in terms of individual capability, scaling to team capability, what are your thoughts on that?

Pat Barry: I think you still have to look at what type of team are you trying to build? What skill sets do they have? Like for example, sticking with data science type stuff. There's data scientists, there's data engineers, analysts, QA people, and a lot of the skills overlap in that space.Sometimes, data scientists might do a little engineering. The engineer might have to do a little bit of analysis. It just depends on the company. You have to kind of look at your bedrock team skills first. Like what are they good at? What's their experience in how long have they been in that particular industry? And then try to scale based on that. Because if you're running a data science team, but all of a [00:15:00] sudden, the boss wants you to start writing copy for ads. You're gonna scale totally differently because you have to pivot your skill sets outside of what your team's comfortable with. So to me, it kind of starts with what's the expertise of the team? How can we give them tools that will help make them more efficient with what they're already doing in terms of these day-to-day. Building a dashboard, doing analysis, all those types of things.

Pat Barry: What are they not comfortable with? What don't they do well? And looking for AI to train up that can help kind of, be their crutch for those types of things. So I think, again, to me it starts with those bedrock capabilities. Like what's the human team, like what are they good at? And build around that.

Pat Barry: How do you think about it? I think it's very dependent on the individual and who's gotta scale it up, and for what?

Erik Martinez: I think that, you know, one of the things that's a hot topic right now is adoption, right? Adoption of using AI inside of companies and trying to scale that. I was sitting in the, AI for Agency Summit yesterday and this topic comes up where, There [00:16:00] are businesses that are absolutely accelerating their capabilities with AI.

Erik Martinez: Claude is a great example of it. Andopen AI just came out with a release that did this too. They've taught their system to help them write the next version of the system.

Pat Barry: Oh yeah,

Erik Martinez: That's pretty huge, right? chat, GPT 5.3 Codex came out and they said that it was mostly writtenby 5.2 Codex. So, there's definitely companies that are accelerating and it's frightening to a certain extent how fast that's going, but they didn't get there overnight. It sounds magical, but they've been working on that problem for several years.

Erik Martinez: Several years in order to get to the stage where they can create those capabilities. Well, what did they have to do? It's kind of what you said. You have to evaluate. One, what is it we're trying to accomplish? And two, do we have the skillset to do it and then align your team to do that. But when you're [00:17:00] first starting and everybody's kind of unsure of their skills, the very,first thing in my opinion is you shouldn't throw all the tools at them. You should give them a tool. Copilot, whatever it is, Gemini. They're all good. And then you say, okay, here's a very specific thing that we do on a regular basis. Now, ask the AI questions and see if it can help you make that a better process

Pat Barry: Yeah, let the AI determine the process.

Erik Martinez: Yeah, so in order for the team to start scaling, that individual capability has to be built. They need to learn how to use that within the work.

Erik Martinez: I think the challenge that it creates, in the short term, it creates more work. It's a little bit painful to do because you have to spend a little bit of extra time 'cause it's not gonna get it right the first time. And it might not even get it right the second time. You have to iterate. And this came up in the summit yesterday, if you're just doing the one question and done, you never [00:18:00] get a quality enough output.

Pat Barry: Agreed. It's testing. Just always comes back to testing and what people are comfortable with.

Erik Martinez: You know, It was, really interesting. I was putting a summary together of the summit, and one of the things that struck me was, in order to build trust with your clients or your C-suite or whoever it is. You almost have to show them the ground of all the failures that led up to you getting to the good result.

Erik Martinez: Right, because it's kind of like an audit trail, You look at it and you say, oh man, the first couple times I did it, I got these kinds of results, and that was terrible.

Pat Barry: Yeah.

Erik Martinez: But then I learned how to ask this question. Or, I found this resource online that gave me an idea of how to ask a different question or approach it a different way with the LLM and just have that conversation with it and build that skill.

Erik Martinez: And I agree that that's part confidence. Like, you actually can do this. You [00:19:00] just gotta go past that first level. You gotta get to the 10th or 12th iteration before it gets good.

Erik Martinez: You know, I'm a big fan of Carla Johnson. She's been on the show multiple times and she teaches innovation thinking. And, in her book, it's like over 200 ideas before you start getting to the good ideas.

Pat Barry: So, if you think about that in terms of your repetitions in learning how to use AI, if you only do one or two repetitions, you ain't gonna be very good at it.

Pat Barry: It's constant over time. I think as the models get better too, some of that will change, but

Erik Martinez: Oh, for sure.

Pat Barry: That's a tomorrow problem.

Erik Martinez: That's a tomorrow problem.

Erik Martinez: So Pat, you know, There's this view out there that people should be using AI and build a personal operating system. Not necessarily a skillset of tools. You know, if you're a sales executive and you're using, Salesforce.

Pat Barry: Yeah.

Erik Martinez: Salesforce has some pretty cool capabilities from my [00:20:00] understanding, I don't use it.

Erik Martinez: HubSpot has some pretty cool capabilities, right? That's a very specific subset of all the things you do as a salesperson. So, how have you built your personal operating system to use AI in a way that enhances your capabilities?

Pat Barry: I guess one, it's mostly like my operating system in terms of like my agents I've built, just kind of how I go about my day to day. I kind of break things down into, you know, client work and non-client work. So, client work is all organized into projects, either in Claude or chat GPT, or Gemini. I kind of spread the work out.

Pat Barry: I have a standard operating procedure for, when a new client comes on board. In terms of what documents get put into what systems. Some are more secure than others. Again, kind of start from having an AI, based on what I can share and what the agreement is with the client,have the AI become an expert in the client with everything that I know, plus anything it can find on the web.

Pat Barry: So like a research agent, but [00:21:00] then being able to tell it like what I need to create, how it needs to get done, what I need to check. So it's almost like a set of individual system instructions that I put together based on that client.

Pat Barry: In terms of non-client work, you know, again, some of it is you're an AI consultant. I have to go like, build and try these things. So, I just go and build things that help me with my day-to-day. Like, article writing or, creating slide decks for sales presentations, things like that. And I have a series of agents that I use to help me with that. And they're just, they're very specific. They do one thing. It's typically multiple agents kind of baked into single agent, so agentic workflow.

Pat Barry: For example, my article writer, it's got my tone down. It'll know here's some sources, here's the, tone I want you to take, and what I want you to focus on. Then I want you to inject this opinion into it. That's kind of the easy part. Why I like it the best is 'cause it creates images for the article for me. So when it's done writing the article, it'll create three images based on the article in this really specific,like claymation style that I like. If I don't like the images, I can go and tweak the agent that's [00:22:00] just building the images. The articles come out. I still have to edit 'em, but I've got it down now to about 20 minutes as opposed to 45. But it's something like that, I show off to clients and sayyou also need to produce something similar to this. This is how I do it for myself. Are you interested in something like this so I can get a sense of what they're looking for. So again, my non-client agents end up serving almost like a sales purpose for me as well. They help me with my day-to-day, but I also use them to show people, I know what I'm doing. I've built this for myself. I can go do similar stuff for you. So, I'd say it's kind of the, those two buckets, like client, non-client and the non-client is kind of vast. That's stuff for me, that's experimentation, but the non-client stuff is intended to be samples for clients. That hopefully becomes a client kind of facing product in the future, if that makes sense.

Erik Martinez: What we just talked about is overwhelming To a certain extent, overwhelming.

Pat Barry: But I think in terms of your individual capability, these are things you can start asking AI to help you think about. You don't have to [00:23:00] have all the answers. You still need the expertise of the data scientists, or you need the expertise of a developer to put some of these things into action.

Erik Martinez: You don't have to know all of it, but you can build confidence that you're asking better questions of those people when you prep using the AI.

Pat Barry: Yep.

Erik Martinez:

Erik Martinez: I think,for a busy owner or a busy agency leader or a busy department head, that's the thing that I think is the biggest win in terms of using AI is realizing I don't have domain expertise in these areas. I can use the AI to help me understand that domain expertise.

Erik Martinez: I was giving a, a friend a piece of advice. She's considering a reverse mortgage. I said, okay, let's use the AI to go understand a reverse mortgage and what questions you should ask. Or expect to be asked from the mortgage company that you're gonna be [00:24:00] talking to. And it's just helping her build some confidence that when she goes in there, she can have an intelligent conversation with the individual because she has a list of questions.

Erik Martinez: And then I said, then you can take the answers and feed it back to the AI to get a second opinion.But if you see some discrepancy, you can ask for clarification.

Pat Barry: Yeah, exactly.

Erik Martinez: Before we move to close, on this concept of building individual capability. If you're sitting there and you're that busy agency owner, that busy CEO, you're a busy business owner and you're like, I'm already getting some benefit, but I know I can get more benefit. What is the next step to level up? What would you do?

Pat Barry: Start evaluating tools. Figure out what do you need. What are your business goals? And then kind of start from there. You know, it just depends on what type of business you are. Agency for example, they have multiple different disciplines typically of different people doing different, you know, creative data [00:25:00] strategy, account, project, all that type of stuff. And then you need to look at the size of your business. But I think it's more just, you gotta start experimenting, start trying stuff. If you're a huge company and you just don't,have the time for this. Well, call the proper consultant that has experience with it. You're never gonna start unless you start doing.

Pat Barry: Pat, thanks as always for coming on and having a discussion. I think, we're all kind of on this learning journey together. Some of us are further along than others, but nobody knows where this is going. The capabilities of the LLMs chat GPT, Gemini, Claude, They're expanding at such an incredible rate that what we see today.

Erik Martinez: I know from what I did a year ago to what I'm doing now, wildly different and I expect that wild ride to continue. But that shouldn't be a deterrent to work on and build their skills. So thanks again for coming on my friend. I appreciate you. that's it for today's episode of the Digital [00:26:00] Velocity Podcast.

Erik Martinez: Thanks for listening and have a fantastic day.

Narrator: