From Copywriter to AI-Native Marketer: The Full Playbook with Shaun Thresher

Speaker A: Welcome to APAC's B2B Growth Podcast by X Growth. I'm Sheen Hoda and uh, on this show we're going to be unpacking trends, busting myths, and delivering insights you can put to work today to drive growth in apac. Most marketers using AI are stuck copying and pasting between browser tabs, losing context every time they switch between tools. Today's guest found a way out of that trap entirely. We're speaking to Sean Thresher, a marketing operator who built what he calls a second brain, a centralized system that gives AI full context on his business, his clients, and even his own writing voice. Sean walks us through how he structures agents skills and workflows, why he almost sabotaged his own content by over relying on AI, and how he uses AI to find niche high intent leads that tools like clay couldn't even surface. We also get, uh, into his advice for marketers who feel intimidated by anything that sounds too technical. Let's dive in. Sean, thanks so much for coming on the podcast.

Speaker B: Hey, thanks for inviting me.

Speaker A: I'm very excited for this chat. I've told you before, I think we are both part of a AI, uh, WhatsApp group and I constantly see you posting there and I'm like, I'm constantly saving your, your posts and replies to other people. And then at one point I was like, I just got to reach out to Sean and be like, hey, can we, can we have you on the podcast? So I'm glad we made this happen. I mean, to me, you know, everything that you type and write and communicate feels like you're an absolute marketing AI wizard. First, let's paint a picture for everyone who's listening. What does your stack and marketing operation look like today that you're kind of operating on top of?

Speaker B: Great question. Uh, it's actually, it's like morphing daily, we'll call it. Right. Funny. Right before, you know, on this I was like putting like pushing the button. I was just, I just built a brand new paid media, I don't know how you call it, like intelligence system. Right. And I just onboarded like the first client for it today, just doing some initial testing. So it's, I'm interesting to see how that works. But it's followed over time and I almost think it's like, I call like the, I'm starting to shift to. It's like the plumbing of marketing, what I'm calling it now. Right. It's all like the underground stuff you don't really see happening behind the scenes and I think it's where all the magic happens. I use a system, it's called a second brain system. You, you may have heard that in the past, it's been very popularized. It's been morphed in a few different ways. For me particularly is how I latched onto this was I was in, I was at a point in my career and I was like, I was using AI daily. I was struggling with copying pasting between Claude and they moved that back to ChatGPT and there was like two windows open, you know, I mean you've probably seen it too. There's like, you can put in your instructions in there and you get like an output, whatever and you're in creating an email series, right? And what really codified this for me was like somebody came up with the idea is like if you could host everything in one centralized location, which just happens to be uh, a GitHub in this scenario, and you put it all in one location. When you start chatting with um, an LLM, it has access to everything rather than like pasting back and forth. That kind of opened up the world for me. This is probably six plus months ago when I started using this and it just evolved over time. Like it really accelerated my journey on this uh, path towards AI. Right? Because once you have everything in one centralized location, now you can add in customized agents and then you can add and give those agents customized skills, right? So then they can really fine tune and hone in your business message and your, you know how you go to market in different sectors, right? So you're no longer having these fragmented pieces. So if you have like email results, you can put that in the system and then the next time you write an email batch, you know, you also have those results that you can uh, play off of to improve the results of what you, whatever you're producing at the time, right? That can be social media, landing pages or your opt ins, etc.

Speaker A: What are the three agents that you're using the most regularly?

Speaker B: I would say more than agents is probably like workflows that I would use more. Right? A uh, workflow, just like a uh, series of steps that you've kind of tuned in over time, right? And it's kind of funny you asked me that because in the very beginning I was kind of like a, a hoarder of agents. I was like, oh man, I'm gonna have the best system. I put all these agents in there and you get to a point where it's just overload, right? The system can't read them it's just too much. And you know, you put them in there and like the agent's like why aren't you reading XYZ Agent et cetera. Right. And it's just, it's just utilizing that. So then you have to scale back and you have to figure out how to make, you know, make it work better with less. Right. That's when you figure out like your workflows are ah, really the ticket. Right. And a workflow is basically again, uh, like as I was saying, a series of steps. And you go there and you kind of program that.

Speaker A: Right.

Speaker B: You go through. We just call it your sop. Right. For me I have a ah, content, a very precise content process that I go to. We'll pull down uh, like stats and information from Facebook. We'll push that into a system and then we'll take that same information and we will create a weekly plan for writing our social media posts. And then we just like kind of repeat that over and over again. Right. So it's just pull those stats, create a plan, write the post, review the plan. Basically like a copy editor, right. And then we'll have, we call something, it's a um, we'll call like a pre gate before you publish. So we'll go through and we'll make sure that like there's nothing missing in our JSON, there's nothing weird in the images, all the sizes are correct and all that before we push out to a platform called Xerneo, which does all of our scheduling for us, you know, circle back to what I was talking before. You know, all those are connected through APIs or some sort of MCP server. And that's kind of what I call the plumbing of the uh, of the digital underground. Right. So once you, once you start connecting these different aspects, right. It's just amazing like how much quicker and more efficient you can work.

Speaker A: So, so that was one of your, your kind of social post thing. Workflow is one of the more kind of, you use it more regularly. What is uh, if you want to talk about like two other workflows that you're, you know, you're there every day or pretty regularly. What comes to mind, the next one

Speaker B: would be the social media that I was mentioning, the social media paid media that I work with. So, so it's still the same type of connections where I will connect with Google Ads and Meta Ads, right. And again it's all connected together again same kind of scenario where you pull down and AI is kind of reading your, your ad accounts and it's giving you feedback on those and then it's you human. Hm, as human, in the loop, operating those, telling what to update, et cetera. Right. Hey, create XYZ Creative for these ads. It's really impressive how fast it can work, right? You could have like a, you know, a new ad up live in like you know, just a couple of minutes rather than like trying to like pull down ads and try to figure things out on your own. Works great if you're in a hurry, right? You know, if you gotta do like in depth, uh, evaluation, whatever, you know, you can do like a more higher end ad. But if you're like, hey man, I just need to like, you know, freshen this up or whatever, you know, just get some recent life back in this campaign. You can do that. The third is we, I work with another, uh, with a, uh, partner and we do cold outreach, right. So it's using an AI intelligence system to go out and kind of like scrape the Internet for different contacts we can find. And the interesting part is like I'm finding that like I'm having a lot of success off of platforms like Clay. When we were doing our outreach, we were using Clay and we're spending a lot of money on Clay. And that was like proving to be an issue for us because of budget. Budget concerns, right. So we figured out that like there's a lot of available online databases that we can go and tap into, you know, and one of those was like this FDA database based in the US and we're able to pull down a bunch of contacts of, of companies that had their medical, uh, equipment expired. And we're able to like grab those contacts and like create like a custom campaign, a uh, cold outreach campaign just for that audience, which I was very, very impressed with. Like, and, and I would never figure that out without like an AI powered system behind me helping me find that. Right? Because it just goes out and it's like, hey, we're able to find that. And then even if you go on like a, a website like that and you try doing some search, you'll never find what you're looking for. Whereas with AI, you can just like connect to their API and it will go back and read the database and it tells you exactly what you need on the back end.

Speaker A: That's an interesting one, Sean. I mean, I, uh, wanted to first paint a picture of the level of sophistication that you're operating, but there are more questions come up coming up. As you, as you kind of talking about this, tell me a Little bit about. So you said, you know, you were using clay, costing too much money, using too much credit, and then you leveraged, you know, whatever LLM, um, you're using to kind of find these FDA lists. Can you kind of paint a little bit of picture of like how exactly that, like what, what did exactly you do to then find those lists and what did you do after?

Speaker B: Yeah, so I have a, what's called like a go to market system that I operate. So I have a couple different instances of GitHub. GitHub basically is like Google Workspace, right? It allows you to store files and code, et cetera. Right. And it can get confusing because people that do development, like developers, they store their code bases in there and that's basically what kind of holds like sites like, you know, Instagram and Facebook, they live in these archives we'll call it. Right. And then where that's different for um, somebody like me is you can store all of your articles and all of your information about your business and how you operate in there. So when you give know the AI or the LLM, which in my case would be cloud code, access to that, you know, it can read all of your information. So it's very easy to plug and play that into different uh, platforms. Right. So if you're going to publish a website, you can plug that into something like a Vercel or if you just want to see on your local computer, you just have all this stored somewhere and it can go and access that, ah, very easily. Yeah. And what I was going to say is like how it's different from Google Workspace is like AI has a very difficult time reading like a physical document. So if you give it like a Docx or whatever, it struggles to read that. Whereas if you put it in like a format it likes to read, which would be like a MD format, which is just a markdown file or a basic, very basic text file. It can read that very easily and very rapidly. So that's kind of the advantage of that. The other part is too, it's like store code and archive that in a much more AI friendly fashion.

Speaker A: Your background is copywriting, right? And you've kind of like trained yourself because I think, you know, I've been in this position where, you know, people talk about stuff and I'm like, oh man, that's so, so sophisticated. That's really for like developer world. And over time I've got more and more comfortable with, with some of these, uh, some of the tools or GitHub or whatever. The point that I'm trying to make is. I think it's definitely something that marketers can upskill in.

Speaker B: Yes.

Speaker A: And it gives them superpowers to do so. So for anyone who's listening, I just want to point this out, that if you hear GitHub, please don't switch off because there is so much potential here to explore. I want to go back to how did you go, you know, you were using clay and then you're like, this is costing too much. How did you go to the next stage of finding your list and creating these micro audiences?

Speaker B: Yeah, so it's. I, uh, kind of cheated, to be honest. I found somebody that had a, uh, go to market kind of like system are pre built. Right. And they kind of had their agents and their skills kind of figured out. Right. And they had like a process already kind of baked into it because they're very good at that. And I just basically purchased their. I hate to say this, but it's like purchased their, their system, right. And I just plugged into mine and now it gave, it gave me the ability to do like go to market research, you know, outside of my system, and was very effective. Right. And then just because it knew all of a sudden knew where to search at, it was able to like pull down these additional lists that we hadn't been privy of in the past. Right. And then we pull that into our system. The system automatically knows like all the client details, et cetera. Right. So we can pull, you know, when we, when we send that off to go research, it's like, okay, you know, you have XYZ client, here's all the details, this is what they're trying to sell, et cetera. Right. And then it just goes out and finds it for us.

Speaker A: I see. So once you build that GTM system that you've kind of got from somewhere else, plugged it into your own system, it was really as simple. Hey, this is what we're doing. Can you find sources of information that we can leverage?

Speaker B: Yeah, it's even simpler than that. Right. All you do is basically if how cloud code works. They work on something called like commands. And you see like a forward slash, you'll type something in and you would type in. Let's just do like full audit. And then you would put like the name of the client and then at that point it just goes to spark, right? You just go do something else, come back 30 minutes later, and then you have this full prepared, you know, market report, intelligence report, which is pulling down a competition, et cetera. It's Very, very robust. What comes back when it's done at the end of the day, Right. So it will, you know, spin up, you know, a series of agents on its own because it already knows what to do. You've already kind of programmed it and told it what to do. So basically it's like having a team of, uh, research agents at your fingertips and then go off and do their thing, come back and then give you, like, the finalized report.

Speaker A: Can we share a link of the agent? Is that, is that a doable thing or is that on the public?

Speaker B: Yeah, that's available. Yeah, Yeah, I can share that. I can, I can send it over.

Speaker A: That's awesome. We'll put it in the show notes. A lot of people poke around with AI and it's novel at the beginning and they maybe build something, but it's really kind of getting traction, sometimes a little bit hard. What was a specific project or a problem where the switch flipped for you?

Speaker B: I don't think there's a specific project, to be honest. It's more like I had this client that I'm working with and we had an issue where we're like, trying to get like, new leads in the door for this, uh, this doctor I'm working with. And then, you know, he started posting. You know, we, we, we had, uh, AI and we were able to like, produce more content, you know, without it being such. Writing it manually. And we were producing more and more on our social platforms. And then we had a couple posts, you know, really pick up traction, right? Nothing like we have now, but like, we're like, wow, this, there's, there's some interest here, right? And then it was just like we started leveraging AI more and more. And then we kind of like paint ourselves in the corner where it was just like, it was just. I don't want to call it AI Slop, but it was kind of AI Slop, right? Because we were, even though we were manually editing it over time, you could see, uh, over the span of like 30, 40 days, it was basically saying the same thing just in a different way. We can't keep doing this in this way, right? Because people are going to catch on this very rapidly and they're going to like, discredit the doctor and we had to figure out how to fix this, right? So at that point I was like, I really need to dive into the system and get this thing fixed because we can't have AI Just it's doing what it was told to do. It just. We, we didn't have Any creativity outside of that. Right. And I didn't have the bandwidth to publish as much as we were doing without, you know, having those same issues. So. So that at that point I had to dig in, like, figure out how the system worked and, like, how we could change it. And then, uh, a few things that, that I figured out was, is like, if you train AI to write, and again, it's, it's excellent following patterns. So if you just give it a different pattern to follow, it will have like a better output. And so I just taught it to write in a different way, you know, for example, you know, if you exchange, you change the sentence pattern so it's not writing the same, you know, six to eight word sentence. That was just one small thing. But, like, when you could read the copy, it was, uh, you know, much more bright when you would read it. Right. Uh, it would have, it would be much more.

Speaker A: Explain that to me a little bit more. Because I think that's really relevant for a lot of people where, you know, everyone's LinkedIn post is starting to sound the same. Everyone's, you know, you look at a landing page and you read a little bit. And I'm going beyond the EM M dashes, right? So EM dash. I feel like it's, it's such a classic one, but, you know, the way, for example, Claude writes, it's like these short sentences, sometimes snappy, and you're like, that's totally a Claude sentence. So how did, how did you kind of go about starting to avoid that and evolve beyond that?

Speaker B: Uh, she went back to some writing books and just like, started flicking through some old writing books that I had, you know, back from my copywriting days. And like, huh. I mean, these guys are sharp about, like, writing, uh, engaging sentences, right? So, and I pulled down whatever, like 15 to 20 of their principles and put them into like, my AI writing system. You know, so each time it would produce, uh, a piece of content, it would have to like, go and like, score against these 15, 20 principles to make sure it was on par with, like, you know, what people actually, you know, engage with. Right. And that really, that really took a huge boost forward in that part. And the next part was it was just like training AI to like, baking randomness. Right? So you would. How AI has randomness is you have to give it a series of samples. Let's just say you give it like, you know, 10 social media posts to compare itself against. Right. And then you. Then each one of those would have its own, like, lead and offer et CETERA in there, right? And then you tell. It's like, you know, each time you produce some content, it has to be random and you can't repeat things, right? So then it will, like, score itself. Okay, let's say today you write five social media posts. It will compare those to the past three days and make sure you're not overlapping anything on there. Right. And I would say those are probably the two biggest things that, like, really helped push our content forward.

Speaker A: Ooh, that is cool. I like that. And the scoring method is really cool. Is that, is it just like one time scoring? One of the things that I've seen work really well is like, hey, create a scoring mechanism and rewrite it until you get a score of, you know, 95 and above at 100. Is that how you approach it from a scoring perspective? I'm curious to know how.

Speaker B: Um, no, that's, that's probably grade one. Uh, the scoring mechanism works is you have to, it's a very involved process. So the scoring mechanism works is you have to, like, review all of your past content and you have to score that first. Right? See, so then you give something AI to, like, judge against. Because if you just tell them to, like, write, uh, a 95 out of 100, it's just making it come to its own, uh, conclusions, right? It doesn't know it's good or bad. It just comes to a conclusion, right? And it just pushes itself a little bit better and you get a little bit better content. But at the end of the day, you're still going to get that repeating in there over time. Whereas if you, if you pull down, let's say you pull down your top 10 articles and your top 10 social media posts, you know, in your top 10 podcasts, et cetera, well, now you have a pool of content that resonates with your specific audience, right? And then you can start giving that scoring. Now, once you, once you have this pool of content ready and you produce new content, it will score that new content against the old content. So let's just say you're writing a brand new article, right? And the article has a score of, let's say, eight out of ten, right? Well, when you write the new article, it has to match that 8 out of 10. Now, what gives? An 8 out of 10 is going to be a variety of things inside of there. It's going to be like your sentence length and the words specifically you use in there. It's going to be, you know, how the voice of the content, the tone of the content, et cetera Right. So it's going to score against all of that.

Speaker A: Okay, let me make sure I get this correctly. You define the criteria and then you go and say, hey, based on these criteria, I want you to analyze these past posts that I made. Give me a score for those. And then now based on the criteria and the way that you've evaluated past posts, I want you to evaluate the new posts. Is that, is that fair to say?

Speaker B: Yes and no.

Speaker A: That's your approach.

Speaker B: Yeah, yeah.

Speaker A: The way you're describing it, tell me what I got wrong.

Speaker B: That's the process. But the way you're describing is like V1. So I would say that that would be like somebody that's probably using, you know, just using AI or just hasn't advanced past, you know, just using, you know, Codex or excuse me, Chad, DP or uh, Claude. What you would do is you would have a, an evaluation system already built into your content generation. So when you tell AI to go ahead and make an article for me, you already have a set of rules in the back end. So whenever it starts making content, it has to score against those, those predetermined uh, numbers that you just defined. Right. So, so again you would have like a list of uh, scoring for your articles. And then once you say, you know, here is an idea that I'm working on, please go write article for this, it will come back already pre scored so that you're not baking that into your prompt.

Speaker A: Got it, got it. So that is already set up and uh, and it's just evaluating. Yep, got it.

Speaker B: Yeah, exactly. It just becomes a thing, right? It does it on its own over time. Right. You, like you part of your system is this evaluation process and one of

Speaker A: the things that you, you mentioned was again going back to the flipping of the switch for you was when you went on Claude code. Is that, is that true? Like I remember previously in our, in our chat you mentioned. Tell me a little bit about that because again I feel like marketers kind of like looking at this and they're like, oh, there is, you know, there's, there's cowork and then there's normal Claude. But I don't want to really touch Claude code because that feels like it's a little bit more for developers and I feel like there's a little bit of the same sentiment towards GitHub and Claude code. Tell me a little bit about that experience.

Speaker B: Yeah, so that was, that was a huge learning curve to be honest. I mean there was again, there was just the factor of just not knowing what you don't know, right? But at the end of the day, Claude code works just like chat or excuse me, just like uh, just like normalcloud, right? You have a little chat window, you type in there and instead of typing into the fancy, uh, chat window on the back end, it looks like a little box and it responds in the same type of way, right? Once you kind of figure out how that works, it's very useful and you get, you have much more functionality on the back end using the terminal than you would just on the front end using, you know, standard Claude. And that's where, you know, I think I was like fortunate to be at a time because when I started using all this like Claude work, Claude, cowork and all that didn't exist at the time, right? This was even before openclaw. So I mean this was, you either had to use the, the Claude, the cloud interface which would be on the webpage, or you had to use a terminal. It was like one or the other and that was it. So I was kind of forced trial by fire to figure that out. But, but what cloud terminal allows you to do is like connect, connect how you wouldn't be able to do previously, right? That's changed now with some cloud, cloud work. But uh, still you have to understand how that works. So it gives you the ability to set up your agents and your skills, all that kind of like manually and fine tune those. And then you can also add in your API connection, which is basically just like having two, uh, websites talk, uh, to each other and it allows you to fine tune those. You know, you can tell Claude, hey, I need, I need to connect with xiz, you know, for example, with FDA database we were talking about earlier. It's like if you connect that API, you can tell Claude to go search this thing, whereas you can't do that in like the standard uh, chat interface.

Speaker A: What are the systems that you have in place? Right, You've mentioned a couple of things while we were on the call here. You talked about your Second Brain. You talk about the go to market system. So how many of these are there? How many of these systems do you have in place that you're uh, you're working off on working with?

Speaker B: It's primarily a single system. I would call it like the Second brain is basically where everything lives, right? And then I'm just attaching everything to that because Second Brain contains documents about myself, how I operate as a single operator and my business in general. And then it has all my client files. So I have you know information about the client, who the client is, what they do, et cetera. And then, you know, back to the writing is I have what I call a digital fingerprint for every client. So I will go down and like basically scan like their top, top pieces of content. Typically it's like you know, 20, 30 articles, 30, 50, uh, social media posts that I will scan. And then we create this, this very personalized profile of how this person sounds online. So there's going to be very nuances about like specific words they use, sentence length, tone and structure of all their content. And that flows into what we call the age of your fingerprint. And this is what we publish our content from. So all that lives in the second brain and then we attach these other outside pieces, which would be the go to market, uh, thing I was mentioning earlier, and then we can add in again, I was talking about like the paid media section, et cetera. So I always think like the, the second brain is like the core system and then we just have these modules that attach to it at that point.

Speaker A: And the modules that you have. Are you really kind of talking about the workflows or. It goes beyond the workflows?

Speaker B: It's a combination of workflows and these external. In the dev world, they'll call it repos. So it's basically just like a secondary workspace account, right, that you attach to it, huh. Which has like secondary instructions to do something different. So for example, in the case of like the go to market thing that I was mentioning earlier, it has its own living system inside of there. And if, and if I put that inside my own system, it kind of gets muddied up. Right? It's not as sharp and as clear as it can be if it lives on its own.

Speaker A: So if it lives on its own, it won't be as uh, sharp or if you connect them, it won't be as sharp.

Speaker B: If you connect them, it won't be sharp. So, so when you tell Claude to go in there, Claude doesn't know anything until you give some instructions. Now once you open up that specific archive, it goes in there and it starts reading its own instructions. And then all of a sudden it knows how to do go to market engineering. Now if I put that in my, my second brain, it knows how to do a little bit of everything. It knows the right copy, it knows how to write images, edit videos, et cetera. Right? And when, when that's all combined together, you know, the message is not as clear as it needs to be, whereas if it's just Its sole purpose is to do, go, uh, to market engineering.

Speaker A: So that's okay. So that's an interesting one. So how many of these kind of tools or repos do you have that you know, you work on independently? So it sounds like you got your second brand that you work independently on. You got your, your GTM system. Is there any other kind of repos that you, you work independently and you don't have them part of this one system?

Speaker B: Uh, not repos.

Speaker A: Okay, so these are, these are the two main ones for you?

Speaker B: Yeah. So inside of like the psychobrain system, I have a different workflow set up. For example, I'm able to generate images using a couple of different methods. Um, I'm able to generate images because I can tell Claude to use codecs to generate images, or I can use like a separate system on the back end that uses like HTML and you can just, it just like plots out different points on a graph. You can like animate that way as well, something called remotion. So yes, I install these on the back end and when I have a project, I can go tell it like, hey, we need to create a design for xyz, whatever that may be, if it's a social media image or if it's a webpage. And I can tell it to create designs using either one of these systems. So that's one part of it. And another one is like my, my analytics system. Like how, like on a daily basis I will have reports for each client come in. I just like program in. For example, they call it the Daily Pulse. And then you just like use the client name in their Daily Pulse for, you know, whatever the client name may be. And it will go out and like scan their, scan their social media, how that's performing, their paid media. It will scan their, uh, SEO and anything else I put in there. So like, for example, if you're doing good mark engineering, you can just have it like scan, you know, the past three days, how we're performing on like lead generation. Are we getting enough leads in, are they converting, et cetera?

Speaker A: Sean M. Where have you seen this fall apart right like this. All of this sounds awesome and people want to, want to kind of try it, but there are, there are situations where it either just doesn't live up to the promise or you just don't set it up in the right way for it to work properly. So first of all, I want to touch on what have you tried that you were excited about, but you were like, the platforms are really not there to be able to make this happen. What comes to mind?

Speaker B: So, actually a couple things, and I want to thank one of my friends for that because again, like you were mentioning, my background is not, you know, software development or technical background. I come from a creative background, so I would just like, blindly trust the AI that was going to do something. A good example of that was I installed this memory system back in, you know, a couple of months ago. And I thought it was working properly. It seemed to be. I was getting an okay output. But, you know, we would have a couple weeks where I, uh, would have high quality content, and then a week where it was kind of like, meh, you know, I don't know what would happen. It was AI slot. I had to go back and do some editing, et cetera. Right. I just thought it was like, maybe I wasn't saying the right thing at the right time. Now come to find out that the memory system I had installed wasn't working properly. It just wasn't like reading the files, et cetera. And I had to pull all that out and start over again. So I, uh, would say, like, it's just like, just like over relying on AI without really knowing what you're doing. I think that was one of my biggest bottlenecks.

Speaker A: Tell me more about that. So you, uh, know, how did you. How did you kind of find out that it was not working properly? You said you kind of gave some credit to your. One of your friends who was a developer. Like, how did. How did that work?

Speaker B: Well, I mean, I don't know if he's a developer or whatever, but he's very short mind technically about how to ask like the pointy questions about how things systems work. He calls me up one day, he's like, hey, I heard you were using XYZ system. Yeah. He said, how's that working? I was like, it's working good. He's like, okay, try this little quick little test. And he like, gave me like something, you know, search for xyz, whatever that was, right? And it just fell apart. It didn't work. And I was like, huh, that's strange. And then I got to look a little bit deeper. And I was like, you know, all of a sudden, like, the file system I had on the back end, you know, wasn't set up properly. And what that means is, is like, there's certain files inside of, like, how Claude code operates. There's files called like, Cloud md, which is basically the roadmap of how cloud operates. So it just, it says if you need this answer Go to this agent if you need, you know, if you need this set of skills, go here and it maps all that out in that certain file. Now if you don't have your, your memory system, which is just, uh, a series of files in there set up properly in there, it doesn't really need to go. It doesn't know where to go to search. For example, like your digital fingerprint doesn't know where to go to search and get information about your client. And that's what I was struggling with. So it would go out and it would find, you know, maybe like the latest post I wrote from last week, and that was, uh, that's it. It didn't find like all the rest of the details that I had.

Speaker A: Is there anything else that comes to mind that, you know, you've tried and you're like, oh, uh, it's just, it's not doing what I want it to do.

Speaker B: I guess the other thing would be, is like, once you start using cloud code and you figure out, like, there's this world of repos and open source software out there, you get very excited about like, everything that you find. Like, oh, my goodness, you know, I can add this on there. And this person seems like very, very smart. He's saying the right thing and he has this system about, we'll just call it like a go to market engineering, right? He has all these amazing agents and skills and you're very excited to add this to your system. Whereas I found, like, once you start doing that, you know, and you don't really understand how they operate and you haven't like, sat down, like, figured out how some of these things perform. You just add things on there and they don't, they don't work, right? Because they're written a different way than how you operate, right? That would be like me going in, trying to run your agency, right? It just wouldn't work.

Speaker A: So how do you go about figuring out how it works?

Speaker B: You just start slow, right? If you. Let's just say you and I were gonna work together starting today, and you're like, hey, Sean, I wanna, like, figure out how to use something very similar second brain. And I wanna use agents and skills in my agency. We would just start slow. We would figure out like one problem that you're currently having, right? Let's just say, let's just say it's a customer service thing, right? And you're trying to like, stay in communication with your clients and okay, now, now we have a problem to solve and we would go back in and we say, okay, we would probably need an agent or skill or something to figure out some of these inbound problems our clients are experiencing, maybe sort of reporting, right? So now all of a sudden we can build a workflow around reporting and then, you know, now we solve that problem. Once you build that, you understand how the agents and these skills operate together, right? Because we've solved this problem. You've seen how on the back end, the information you need to give these items or these agents to perform their job, right. It would be no different than like onboarding a brand new, uh, employer.

Speaker A: Got it. So your, your recommendation is start with one specific problem and really start building from there a solution around that particular problem.

Speaker B: If you don't have the time to dedicate on the technical side of it, right. You can get really lost in the weeds and you can use, lose a lot of time, you just give up on it. Which I would like to avoid that.

Speaker A: What is your recommendation of if somebody wants to do this on their own, right. Where should they start? Like what, uh, is there are, you know, kind of thought leaders that they should follow or their courses that you're like, hey, I would start here. Especially someone like yourself, right, who is not super technical or that didn't have a technical background, right? And they're like, yeah, I, I really believe this, I gotta start. But they're like, I just need to learn the basics first and build on that. What is your recommendation? Like, where do they start?

Speaker B: Yeah, I would highly Recommend like an AI community like the one we're both in the JNai circle. Not that I'm trying to promote that or anything, but going somewhere like that, what I found is like the, that people get in there and get involved because all of a sudden you have somewhere to ask questions, right? And you know, I understand some people get in there and they may feel like they're very much a beginner compared to like some technical people and they might be, you know, shy about asking questions, but at least you can see other people asking similar questions and you can kind of get feedback on what you're working on. Whereas I found if you try to do it on yourself, like if you go sign up for Cloud tomorrow and you install cloud Desktop and you try doing it on your own, you know, you're gonna have a very difficult time and you're probably gonna like, you know, you may or may not continue, uh, using that. Right. Which is a shame because it's such a powerful tool you can use. So I would again, I would find out like some Sort of like group that you can get direct feedback from and get your answers. Uh, replied to, you know, where that might be, like the one on one. Yeah. Or some sort of community or whatever. Right. I think people, uh, you could learn the fastest that way.

Speaker A: Was that how you went about it? Like, what was, what was the kind of the starting point for you where you're like this cloud code thing. I got to check it out.

Speaker B: Well, I mean, my journey was a little strange because I was in that community. I was replying and I felt way behind. I saw these amazing things these guys were doing. I'm like, man, there's no way I'm going to catch up to these people. And then there was somebody, uh, that presented this, this second brain system in there. And I just happened like, you know what the biggest selling point for me was if I could just manage my day better and I was just much more efficient at it. You know, I had the project management tools, but if I could just use that and it would, AI would allow me to like, not forget certain details of projects and I was on there, I would be thrilled with that. But it turned out to be so much more than that. It was able to like, you know, everything I've explained so far.

Speaker A: So let's talk about the second brain. Right. We've touched on it a couple of times on this call. Tell me a little bit about. So your second brain is built on CLAUDE code, Right?

Speaker B: Let me reframe that. CLAUDE code is just an LLM. Right. So the second brain allows you to use any of the LLMs, which would be ggbt, you know, whatever, come out now, Kimi, whatever, and Claude to interact with your data, we'll call it. Right. And then the thing about a second brain, it allows you to have a single archive of all of your data. And if you set up that, and you set up in the correct fashion, you give Claude and these other LLMs the ability to read that data much faster and make better decisions about what you're working on. Like I was explaining earlier, let's just say in the case of you're going to generate content, you know, when I go ask it to generate a social media post or a blog article, if it immediately knows my voice, my tone and all that and all the stuff I explained before, it knows like past articles that I've written, uh, I've been successful with. Right. Or in the case if I'm writing content for a client, it knows everything about that client. Right. It can make content much faster and more efficient. Right. With, with less errors we'll call it, right. Compared to. If you remove all that and you, there's actually a good test. If you have a very basic system set up and you use that system and it knows your voice and all that, and then you go try to use just like the, the cloud interface, which is the webpage, right. I mean there's a uh, stark difference between what you get out of them.

Speaker A: How did you decide to structure your, your brain, right? Your or the second brain? Because I think that's always a little bit of a question. It's like folders, do I need, how do they need to be organized? Tell me a little bit about that.

Speaker B: It's uh, going to be based on how you work, right? So it depends if you're a company or if you're like an agency. Now if you're like an agency operator, you're going to have like some, some structure in there about. Let me back up a second. There's a basic structure to it. There's a basic structure to. You need to have your, the basic Claude file set up properly. That would be one I'm not going to get in that. You can look online about that. But there's like a system how you set up your agents and, and your skills and how those communicate with the rest of the system. There's that you want to make sure that's in tip top condition, right. It's almost like going to the gym. That has to be maintained on a regular basis. That's kind of what makes everything operate at such a high, you know, high efficiency. And the next part is like how you, you want to set up your, what's called your client files. So you're going to, you have like an onboarding process. So you have a new client come on board, it's going to digest all their information that they send you. That could be your onboarding documents, it could be their white papers, it could be, you know, just in general, what their unique uh, selling positions are, positions are, et cetera. So you digest all that into your systems and all of a sudden once you power up the system with cloud code, it knows everything about that client that you have on your hand on file, right? And then outside of that you have like, what's called, what's a memory system. Now the memory system, that's where it's going to add details about how you write. Uh, you can, you can put in uh, different, um, information about the client outside of like the basic uh, onboarding. So maybe they have like in my case, I have a client has written several books. I can digest those books and put in there, and I can pull from those as resources when I'm writing different content, et cetera. So that would be the memory system, right. So that could be uh, client conversations where there's specific details about upcoming projects. It could be like results that you've had in the past from these clients, et cetera. Right. That can go in there. So it's easy reference. And then the next system would be, is like we mentioned earlier, would be your workflows. So the workflows is basically like your SOPs. Like we have like a standard process for like doing reporting or how we edit content, et cetera. So that's kind of like the, like the, the bare bones how you set it up.

Speaker A: Okay, so a couple of terms that you've used here that I want to make sure that people understand. You talked about agents, you talked about skills, you talked about workflows. Can you distinguish between these a little bit for someone who might not be familiar with uh, with these terms?

Speaker B: Yes. So an agent is basically like a, like a worker, right? It knows how to do something. For example, if you give it a job to write social media, to write content, right. You can have like an agent that can write content. Now a, uh, skill is like a specific way it does that, its job. So if you have an agent that can write content, you can give it a skill that all of a sudden it tells, it explains in detail how to write a social media post, right. Or you can give it a skill to write an article or a skill to write something else, right? So that's kind of like the difference between an agent and skills now. And again I said we can circle back. So one of the, one of the issues I ran to in the very beginning was like you install all these agents because everybody's talking about them. But what they don't tell you is like how these agents are supposed to work together. Right. What I found is if you have like a higher level type of uh, agent where it's, it's doing what is called like the, the content agent, right. It knows how to work with words, et cetera, and then you can give it skills and it performs much better than having like a, an agent for social media, an agent for article writing, an agent for, you know, press releases, et cetera. Right. I found, I found that way. It works much better.

Speaker A: Oh, I see. So instead of having multiple agents, have one general agent and then provide different skills to it to, to, to basically leverage that umbrella information towards that skill.

Speaker B: Right. Cause what happens is when Cloud reads that, it will get confused about what it needs. Even if you set it up properly, it still does its own thing at some points. Right. Claude's not perfect. So it will go out there and if you have too many agents, it won't know which agent to select and it'll just select one at random or not select one and it's like just do whatever it wants. Whereas if you very specific about these handful of agents that you work with, you know, it does a much better job.

Speaker A: One of the things that I've seen as well, people just use the skills is that something that you've tried as well or kind um, of didn't work very well. What are your thoughts on that?

Speaker B: I don't know about so much, uh, just using skills. But you can. But it. So agents is just like the ability to give it a job. So it's. So you give it something to be able to think on its own. Right. And when you tell it to like work with a, whatever the content agent, right. It knows how to think in words and you know, literature, et cetera. Right. And then, and then you give it the ability to do something else. So I think that's where the skills I'm on. But I'm not sure about like, man, I've seen much better results when you use the agent tied to the skill.

Speaker A: Okay. The other term that we used we've used a couple of times, the workflow, maybe we could kind of bring all this home together.

Speaker B: They all, they all go together, right? So a workflow. Again I keep iterating this like, just like an SOP of a business. It's just a step by step process of what Claude or whatever AI, uh, you're using that needs to go through. Right. So you have a file on the back end, this called, this is like your workflows and you'll, you tell it step by step what to do. Right. For in my example I was mentioning before we had something called a daily pulse. It goes by like step one, you know, pull in the data from our paid media from the last two days, you know, pull in social media, et cetera. Right. So now I have all the data. Step two is like grab the data agent, the data analysis agent, break down those numbers and tell me what it means. Step four might be to actually is to verify all those numbers to make sure they're accurate. And then AI is not inventing things. The next one might be to grab. Step five might be to grab the strategy Agent which is going to like pull all that into something coherent that I can understand and like that I can take action on. Right? And then the next it'll be like maybe like draft report that I can read. Depends on what you need at that point, right? You can either like have like a webpage that's like a report you can have or you can read it straight in the terminal. It kind of depends on, you know, whatever you feel comfortable with.

Speaker A: I think that the SOP component, I kind of struggled explaining that to some of our team members. And something funny happened the other day where we have an agent that does research on accounts to basically say what is happening with that account that might be relevant to us as a company. And we went and said, hey, the same agent, we want you to kind of do this and also summarize it. Where before, what we used to do is we would, you know, the agents would write the full kind of description and it was a fairly lengthy description and we would go and say, just give me the two bullet points that are most important for us as an organization and it will give us two bullet points. It was great. And then we went back and we said, hey, why don't you, the same agent, also do the summarization? And the results completely changed. It became completely not useless, but the quality of it drastically dropped when you didn't have that workflow in place and you tried to do all of it in one go. And uh, and that was, that was kind of like a moment that I was like, guys, this is why workflows are so important because you go through these steps. So I think, I think it's kind of made sense for a lot of our team members.

Speaker B: Yeah, that's a good example where if you have your system fine tuned and it knew exactly what you were looking for as an agency rather than just like randomly like find me two bullet points. Right. Because again, it's just going to come to its own conclusion if you don't have something on the. You don't have a set of instructions guiding it on what to look for. Right? So let's just, for example, let's just say if it knew the key metrics of each client, right. And it can make better decisions based on those from that reporting. Right. Let's just say, hey, this client is very much focused on like dropping uh, cost per lead, right. Or maybe this other one. This is very much focused on volume, right? Those are two different discussions. So if you're. So if your system knew that and it was pulling down this Reporting it would be like, okay, hey, client A, it's, it's going to focus on growth. So we need to like look at those metrics and it would be a very different set of bullet points that you're getting down.

Speaker A: Sean, the other thing I want to ask you, I mean, we've touched on a few things which I think is very valuable. We jumped around in a few different areas, which is, uh, which is great. Is there anything else that you would recommend here? Because I feel like your recommendation for, hey, where do I start with this is pick a problem that you're trying to address and then potentially join a community. Is there anything else, any other advice that you have for, for anyone that is like, what's my first step?

Speaker B: Yep. Here's what I do with first steps. Either pick one or the other, either Claude or Chat gp. The next thing I would recommend is go ahead and pay for the upgraded version. Many people just start out with the 20 version. There's a couple different tiers across the board. They're all pretty much the same. There's like the free one, which is terrible. There's the $20 version, which, you know, many people kind of dabble their toe with. Then there's the $100 and the $200 versions. I would very much recommend investing in the at least the hundred dollar version. What I found is, is when you upgrade from 20 to 100, and I talk about from personal experience is you very much have like a mindset shift because when you're in the lower tier, you're very much concerned about the running out of tokens, you know, finishing your time, et cetera, and it's going to stop working. So you're very, you're limited about what you will ask from the AI system. Right. So you maybe, you know, you can do maybe like certain amount of research and maybe you'll throttle that back. So you're doing like 60% compared like 100% of what you could achieve on that. Whereas if you have like the upgraded package, you know, you can, you can run that without any concerns at all. And right. All of a sudden you become, you know, less concerned about like, you know, running out of tokens and consuming this up. So I would, I would recommend that there's definitely a mindset shift at that level. And then the next thing I would recommend is like, just go ahead and start working with either again. Depends on what you're going to pick. You can go with Codex or Claude Desktop and just start learning how those operate. Right. They're very Good. Now, like, the system I've talked about, you know, is probably a little bit too much for most people. I recommend using that. The user interface is very, very easy to work with. They're upgrading it all the time. You can install your skills and all that stuff. You know, a lot of stuff you would probably use on there on that interface and just get very familiar with one of those, right? So just go ahead and grab. Let's just say you're going to use ChatGPT, uh, which codex is very good right now. Set that up, think about the skills you would need, install, you know, maybe two or three of those, uh, get those working. That could be as simple as just some data analysis, right. Like in your case you just mentioned, right. And get that working smoothly and then at that point you can think about, you know, where to migrate at that point.

Speaker A: I love it. Sean, last question I have for you is what's next for you? Like what's on the radar that you're looking at? And you're like, I'm really excited about this.

Speaker B: Actually a couple things, right. Like at the very beginning I was talking about this digital plumbing thing. So I have a couple of things I'm working on. One is helping practices and businesses, you know, figuring out like their, their analytics strategy, right. And getting that working properly and then using AI to communicate end to end. So let's just say if you're running paid media, getting AI to help you evaluate everything from your ad account all the way to the final sale with the customer, right? So watching that entire journey and that's, you know, it takes quite a bit to get that set up and working properly and testing, that's the one part. And then the next part is like, is tacking on like, we'll call it competitive intelligence. This is probably what I'm most excited about. It's just like this competitive intelligence aspect where there, there's different tools and finely tuned AIs out there that can like basically keep it the pulse on an industry. I'll give you an example. Uh, there was one example that I was, I was hearing about where this, this company sold a. What was it? It was a water pressure valve, right? So that would make sure that you had like the consistent water pressure throughout your entire home. And they were struggling to find new customers. And using this type of, uh, competitor intelligence system, you know, it scanned industry and it figured out basically based on like feedback from reviews and comments. It figured out like on the top floor of these hotels, their guests were complaining about low water pressure, which you know, in hindsight, you know, made perfect sense. But you know, if you don't know what you don't know, you know, you would never think about that. So now all of a sudden, you know, they knew exactly who to target. They could hit these hotels. Hey, I know your guests, I've heard your guests, you know, whatever. Uh, the top three floors are struggling with this. We have a solution. It seems like a no brainer sale. So that's some of the ideas I'm working with, I'm really excited about. Because just think about if you're working with a cut client and they have that in their, their back pocket, you know, you almost crush the competition.

Speaker A: That's really cool. And like that almost sounds kind of market intelligence rather than a competitive intelligence. Like competitive. Would you say it's more kind of like market intelligence or is it like analyzing competition and doing.

Speaker B: It's both. It's both. You can tune it to whatever you're looking for. I would highly recommend tuning it to both. Right. So you would know what your competition is doing and then plus what the pulse is on, uh, the market in general, whoever you're trying to target.

Speaker A: That is very cool. That is very cool. Okay, those are two on your radar, Sean. This has been an awesome conversation. I got some rapid fire questions for you, Sean, before we wrap up. What is one resource that fundamentally changed the way you work or live?

Speaker B: Not a single resource anybody can tap into, but like a, I would say just like working with the right, with the right people. Right. They're very professional. I'll give you an example. I got, we'll call it my dream job and I was able to work at like a very high end publishing company and like just being around a certain level of quality of work that they did like was very much an eye opener. Right. I guess I would assume we'd be going to like for like in the US if you went from like a high school football team to like a NFL, you know, team, you would know huge difference in how they operate and how they train, all that. That was probably like the biggest change for me. Just like just seeing like they operate a higher level.

Speaker A: Yeah, being, being amongst high performers. Absolutely. Okay, question two. If you could give one advice to B2B marketers, what would it be?

Speaker B: Start using AI sooner.

Speaker A: Who are some of the influencers and thought leaders that you follow?

Speaker B: Yeah, there's a couple guys on LinkedIn. There's one guy's name's Ethan Mullock, you've probably heard of him. I uh, list him quite a bit. There's another guy that I follow. His name is Nate B. Jones. And then there is David Arnault, which runs the Genai community, which I probably follow those probably the most. I found out that if you, if you follow too many people, you follow no one.

Speaker A: What is something that excites you today?

Speaker B: What excites me today is like the opportunity with AI. I mean, we, we've kind of, we're kind. I think we're rolling out of the phase of like, AI is going to see everybody's job. And I really think it's very interesting. I've had these conversations with a friend of mine and he's like, like, what's coming up that you're seeing with AI? I just think that people are going to be able to get much more creative about what they're able to do. Right. I'm, um, going to give you an example. Like, there were some people I've heard some lot of complaining about, like, oh, uh, we publish books and our book sales are getting slower or we run email newsletter lists.

Speaker A: Right.

Speaker B: And like, people like, going to like, search on ChatGPT the answers that we publish. And I'm thinking, like, you need to think outside the box. An example I've seen is like, I've seen a page where you can scroll through an article. So if you can imagine, uh, a webpage and like two thirds on the left side is the content and the right third is an, actually an animated graphic. So while you scroll down the page, there's, there's a, an animation happening and it's kind of like playing out the story inside the content. And that would be much more engaging. Reading, like words on a page.

Speaker A: Yeah, that is cool. That is cool. So, yeah, think about like that AI content generation of. Yeah, I mean, that's, that's so true. Right. And, um, we don't, we don't really know. Just scratching the surface.

Speaker B: The question I would ask myself is like, you know, how can you bring this to life? Whatever campaign you're working on like this. Let's just like, like in a B2B outbound reach campaign, if you have like a landing page and somebody's got to go there, how can I bring this to life? The problem that you're trying to express to the client, what sort of like, you know, whatever, uh, graphics or video or whatever, can you, you apply that using AI that you couldn't before, you could like, bring this thing to life. You know, what would make this person like, light up when they see it?

Speaker A: Such a great question to ask because it's so many ways that you could do that now. Sean, this has been an awesome conversation. I just want to say thank you so much for coming on the podcast and sharing all these insights and experiences. I've, uh, definitely taken a lot of notes, and I'm sure a lot of other people listening have done the same thing. So thanks so much for giving us your time.

Speaker B: Yeah, thanks for inviting me. I appreciate it.

Speaker A: We hope you enjoyed this episode. If you like APAC's B2B Growth Podcast, please share it with your B2B friends and subscribe for weekly insights on B2B growth across APAC. Sign up for the XG Weekly Newsletter link is in the description Apex B2B Growth Podcast is produced and edited by Alexander Hipwell, and music is by the mysterious Brickmaster Cylinder. We'll see you next time.