Leveraging AI

280 | Build a Marketing Machine That Runs Itself with Claude Cowork (Without Losing Control) with Dr Nici Sweaney

• Isar Meitis • Season 1 • Episode 280

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 59:16

What if your entire marketing department could run on one command… without losing your voice, your strategy, or your edge?

The truth is, most businesses are still stuck in a world where growth depends on time, people, and manual execution. But that model is breaking—fast. 

In this episode, you’ll discover how AI agents can transform your marketing from a fragmented, time-consuming process into a streamlined, intelligent system—one that produces consistent, high-performing content while keeping you firmly in control. Dr. Nicci Sweaney is a former academic, data scientist, and business consultant who now helps organizations build AI-powered systems that transform how work gets done. With a PhD in ecology and a background spanning education, analytics, and leadership, she brings a uniquely rigorous and practical approach to AI implementation.

If you’re serious about scaling without burning out your team (or yourself), this is the blueprint you’ve been waiting for.

In this session, you'll discover:

  •  How to build an AI-powered marketing system that creates newsletters, blogs, and social content automatically 
  •  Why “execution” is no longer a competitive advantage—and what replaces it 
  •  The concept of AI agents as team members (not tools) and how to manage them effectively 
  •  What “skills files” are and why they’re the secret to high-quality, non-generic AI output 
  •  How to maintain brand voice, personality, and strategic control with human-in-the-loop checkpoints 
  •  The exact process of turning weekly inputs into 15+ pieces of content with minimal effort 
  •  Why most AI-generated content fails—and how to avoid blending into the noise 
  •  How leaders should rethink roles, teams, and productivity in an AI-driven world 

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

Isar Meitis

Hello, and welcome to another episode of Leveraging ai, the podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business, and advance your career. This Isar Metis, your host, and we've got an incredible episode today. Every business needs clients, otherwise you don't have a business. Now, in order to get clients, you need to know how to do marketing, so potential clients know about your products and services and will hopefully hire you or buy from you. However, doing marketing and more specifically doing marketing well is a tedious. Process with multiple steps and touch points and requires really deep knowledge in multiple topics. Like you need to understand your target audience. You need to know what are their pain points. You need to know what are best practices that actually generate effective, well converting content. You need content ideas, so you know what to create. You need to create multiple versions of it, adapting it or repurposing it. To defend platforms you need to post, you need to comment, you need to learn over time and improve and so on and so forth. It's just a lot of work. Now, back in the day in 2025, uh, you would either hire a marketing team or hire a marketing agency to do all of these things for you, but even if you did have these people on staff and or as a agency. You would still need to be deeply involved because you need to know what's going on in your industry and you need to be connected to your internal processes. What's happening inside your company, what things are new, what big announcements are coming, what tools are you using that you wanna share with your clients? Uh, what changes are happening in the things that you're going to deliver? And it needs to know you in person. It needs to understand your brand. It needs to understand your company's history, the connections that it has. Like there's a lot of knowledge that is beyond just the marketing person or the marketing process. And then you need to bake all of this into blog posts and newsletters and social media and so on. And all of this, everything I just said, can now be done effectively by AI agents and using different types of automations. And in today's episode, we're going to show you exactly how you can build a system that will do. All of that while keeping the humans in the loop and in control in the relevant steps of the process. It's not just, oh, I'm gonna hand it off to AI and hope it's gonna do the work. Well, you'll be able to have detailed, easy to use checkpoints in the tools that you're using today as if it's a part of your human team. And that's where AI becomes the most impactful is, is becoming a team member as part of the rest of your team. Now, the cool thing is what we're gonna teach you today, while the example is gonna be in marketing, can be applied to any aspect in your business. Literally, anything you can do in front of a computer can be automated in a similar way. Now, our guest today, Nikki Sweeney, has one of the most incredible backgrounds of any guest I ever had. And I mean, episode two 70 something. So she has a doctorate in philosophy, a PhD in ecology. She spent most of her career, uh. As the sub dean of the Australian National University, and she spent many, many years as a data analyst for multiple different organizations. So an incredible variety of, of really serious backgrounds. And in the last three years, she's been applying everything she's learned through her entire career, uh, from a very practical, kinda like engineering approach to problems, how to build full departments of AI agents for companies. And she turned that concept into a machine that can be replicated in building new teams for more or less any purpose you can imagine. Now, her unique experience that she brings to the table gives her very deep understanding of how all of this should work, both from an analytical perspective as well as from a process building perspective, from a managing, uh, management perspective. And so. In addition, she's really good at explaining all of that. So it will be very, very clear to you and to me. And so I guarantee you, this is gonna be one of those episodes that you will want to hear again and take notes because it is going to be that good. And since the world has gone crazy in the past few months and understanding it a little better and how to actually build it yourself is extremely valuable. I'm truly excited and honored to welcome Nikki to the show. Nikki, welcome to Leveraging ai.

Nicci Sweaney

Thank you so much for lovely introduction. It's a pleasure to join you. I'm very excited about today.

Isar Meitis

Yeah, me too. I, I, I, you know, you and I had a few, we, we could probably make this a six hour conversation because we can geek about this for so much. We'll try to convince the geeking out to as little as possible and, and focus on, on what's practical for people. Uh, but I really, really appreciate you doing this. The other cool thing is it's, it's tomorrow for you, right? So, Nikki's in Australia and I'm Orlando, and so it's her morning, my afternoon. But it's, it's gonna be absolutely perfect.

Nicci Sweaney

Exactly. Yes. I can tell you that the future is fine.

Isar Meitis

So far, so far so good, right?

Nicci Sweaney

Yeah. Right. Well, all 12 hours of it. Yeah.

Isar Meitis

Uh, so yeah, Nikki, I'll, I'll, I'll leave the state to you, you, uh, share, share your, your, your magic with us, and I will, and I will try to, uh, to add my 2 cents were relevant.

Nicci Sweaney

Amazing. Okay. So as you did a wonderful job of introducing me, I come from a data science academic background, uh, like everybody or like a lot of people listening, you know, I had that moment at the end of 2022 where I opened chat GPT, and, and I thought three things, right? I thought, my gosh, this is pretty amazing. I wonder what else it could do. This is a whole lot of fun. And I fell fully down the rabbit hole. The second thought was. I don't, I don't know if I should have bothered doing my PhD. I think this thing could have written it in a week, and I still kind of think that. And then my third thought was, oh wow, we could be in some trouble here. I think we've just created something which we aren't going to be able to contain, and we don't know what it's going to do to us. So I started talking about ai. Back then, I published an article in Times Higher Education about how it was going to change education in the workforce. That turned into more and more speaking opportunities. But as you said, my kind of background in data science and analytics and business consulting and then in education, it quickly got me to the point where I realized people really needed help implementing this, not just learning about it. And when we implement these systems. There's a lot of things that are incredible. There's a lot of opportunities, but there's also a lot of things that can go wrong. And as I'm sure listeners will understand, you know, the fear is real, but it's justified in a lot of instances because we do have these big issues around data quality ip. Yeah, human in the loop bias stereotypes. So that's always been really important to me from the start. Uh, I talk a lot about gender equity and equity for marginalized groups when it comes to leveraging tech, but I approach it from a really kind of systems thinking lens. I love solving business problems with ai. So the thing that we're gonna talk about today, like you said, I've got AI everywhere. I don't remember what it was to work before this. You know, I, I often get this question as I'm sure you do. I often get this question when I'm talking like, Nikki, you talk, you know, you've got so much AI working in the business, you've got it across all departments. What do you do? Anymore. And I find it a really tricky question to answer because I don't think my concept of doing something is the same as the average person's concept of doing something like, I don't know what, what it is to do something anymore. I don't do anything from scratch. I make decisions and I give feedback and I delegate to my AI systems, but I don't do the doing anymore.

Isar Meitis

Yeah.

Nicci Sweaney

And I think that that's really cool because we wanna be in that strategic space, right? Because when you're not doing the doing, you can actually think about the doing. And that's where magic happens,

Isar Meitis

I

Nicci Sweaney

reckon.

Isar Meitis

Two things. One, uh, Kasi Zykov, who used to be the Chief Strategy Officer at Google, which is a very fancy title, but it means you're really, really smart. Uh, she used to say that AI is gonna separate between the thinking and the thunking. And it's a word you may not. Yeah. But every time you're not thinking it means you're thunking. And the less thunking you do, the more thinking you can do, which makes a lot, a lot of sense. So this is one thing. Yeah. But, but to kinda like compliment what you said, my mindset has been that literally like in the last three months, like December through now, we shifted from a world in which the limitation of a company to grow is resources to three things, good ideas, which goes back to what you said before.

Nicci Sweaney

Mm-hmm.

Isar Meitis

Prioritizing these ideas because you're gonna have very quickly, a lot more ideas than you and your team can actually implement and really, really well-defined requirements. If you have these three things, which goes back to. Decision making strategy, like all the things you talked about as far as the leadership part of it, the AI will figure out the rest, it will build the agents you need to build in order to do the thing you don't want to do. Or even if you want to do, they will do it faster, better, cheaper than you ever can. Uh, and so I think these are now the three key things that people need to learn is, is these kind of things.

Nicci Sweaney

Yeah. And, and you know, it's such a critical point and, and I think we're sugarcoating it to, to sell that everyone will immediately be there, right? Like one of my big passions is how do we foster critical thinking in the masses, because that's not something we've traditionally taught in schools or in workplaces. And I do a lot of work with education institutions as well, because we really have to. Really changed what we're teaching people to do because we don't need so many people that can just execute. Yeah. Execution is done and, and humans are quite, as you just said, humans are quite slow at it when we compare these systems. Right. So, you know, execution is no longer a selling point or it doesn't make you competitive anymore. It's the ability to execute the ideas that actually matter. So, yeah, couldn't agree more on your point. Um, and so yeah, I, I have these working all over the business. I know we're gonna talk about marketing today. Um, as you were just saying, you know, since December, the landscape has really changed. So I know that we talked earlier about, um, I was using workflows and automations and, and NAN to replicate my marketing process, but I've now changed it, of course, all over to, uh, Claude and agents and notion agents. So what I'm gonna show is a slightly different, but it's the same premise. Yeah. Awesome. So I just wanna share my screen and I'll, I'll share and unshare as we kind of go through the different little bits and pieces. But now my, my agents, so I've built them in, um, in call code as, as lots of people will have heard about. And so I have a marketing system. I'm in thought leadership, so my marketing is really around how do I stay ahead so that people turn to me as a thought leader in my industry. And a really big part of that is one long format newsletter per week that rounds up the latest in AI news. I know that you've got a similar one latest in AI news. I do a little bit of a tool spotlight, a use case around something that I'm thinking about or loving. And then we also do, um, some community spotlights and then where to catch me next.'cause I do a lot of public speaking, so it's, you know, come and meet me at these places. So this system is now built out inside of Claude Code. So on a Monday I just type in this one line that says Generate Week, right? And what happens when we generate the week is it then creates on my Notion board. So I use Notion to house everything. It creates the AI News weekly for that given week, and it puts in a template of what I'm going to be talking about. So my personal story notes a tool, spotlight, topics and themes, upcoming events, community wins and testimonials. It grabs from my calendar as well as notion where we already keep all our testimonials and it kind of pops it all there for me to have a look at. And this is one of these important things around the human in the loop, right? So what's really important for me is that I like to add my own personal stories and I like to make sure the newsletters going to look at things that actually have come up for me in interested me during the week. But. I travel heaps. You know, I have a lot on, uh, some weeks that's harder than others. So I, for me, I really needed this system that was going to support the consistency in that space and grow my marketing and my newsletter as well as allow me to actually have some personality in there and not just be kind of a generic, let's scrape the internet and churn through headlines. So why I made the system like this is because I want it to sit there from Monday through to Fri to two to Thursday. This page sits there and I add voice notes to it as I go through the week. So as things come up and they interest me, if, you know, like this week I was doing this last night I spoke at an event this week, yesterday, uh, in Sydney. And so I dumped in all of the feedback that I got from it and the nice things that people were writing about me on LinkedIn. So I collected these kind of ideas throughout the week to add my personal kind of flavor to the newsletter. Okay, so all of this, we're collecting ideas, we're collecting screenshots, we're collecting whatever else is happening in the community. When I get to Friday, or as I got to last night ahead of this, I then just go back.

Isar Meitis

Just a, just a quick question where you're saying you're dropping them in, are you dropping in straight into Notion or are you using like WhatsApp or, or

Nicci Sweaney

Yeah, I usually just drop straight into Notion because I have notion on my phone. Yeah. So

Isar Meitis

it's the same I

Nicci Sweaney

voice record.

Isar Meitis

Yeah.

Nicci Sweaney

Yeah. And I can voice record straight into there. I can open up the page and drop things in there. Um, we can, like, it's connected to Slack as well, but I find it's exactly the same as pressing on one button as it is to the other.

Isar Meitis

Yeah.

Nicci Sweaney

But yeah, dealer's choice in that way. So once I get to this point, I can, then I go back to my console and I just say, can you generate, uh, can you plan the week? Right? So I say plan week, and this is week 11. So plan week 11 for 2026. And what happens at that point is that my agents use all of their skills files around marketing, how I write on LinkedIn, what I write across Instagram, what I have as subject lines in my newsletters, how I structure my newsletters, what language we use, the fact that I use Australian spelling for things, you know, my content, uh, how I write blogs, SEO, the tactics, who I'm speaking to, what my offers are, where my campaigns are sitting. It collects all of those skill sets and uses that to then plan out all of my content for the week. So what it does is it creates a newsletter for the week. It then repurposes that into two separate blogs and it then also writes all of my social media content. And then of course,'cause we're all playing with claw code, I also make a little app that sends all the markdown files of the carousels to an app that then creates all of my on brand carousels as well ready for me to push across to social media. So what happens is I get to Friday and I have my newsletter up and running. It also has hyperlinks to all my events, hyperlinks to the news so people can read more about it. It's done in my brand colors, it looks all pretty. And then I get a lovely little folder of all of my carousels for the week as well. And I can go into these and check them. And then once I've approved everything, so again, it's really important like human in the loop. But once I approve everything, I then say push content, and it pushes to my scheduling services and it pushes to my email marketing platform. So I have three touch points throughout the whole week. On average creates not only the newsletter, two blogs, but around about 15 pieces of content between the reels and the carousels and the stories, insights. And it means that I'm consistent. I get around half a million impressions every month, which, you know, I've only got about 20,000 followers across all my platforms. So I get a huge amount of reach and I add around a thousand people to my email marketing platform every month as well. So this has been wildly successful and I know we're gonna kind of talk about what makes it successful, but I just want to run people through how I've set this up. I use Notion as my brain, right? Claude code for this one is the console where I'm interacting with the agents. And the other really important parts, as people will know is the context. So all of those skills files that define how it operates, and again, I'd love to kind of get into that more, but skills files. They are your differentiator. So

Isar Meitis

let me pause just for one second before we dive into skills files, because there's a lot here to, uh, to talk about That is absolutely brilliant. So first of all, absolutely freaking brilliant. Uh, the second thing is, uh, what you need to understand is two different things. One is that there are several different components here that make it so successful. Hmm. The first component that makes it successful is process, right? There's a very clear process that Nikki has identified, structured, uh, documented, uh, decided upon that can then be executed time and time again in the same. So process is number one. Number two is data. And there's two types of data here. One part of the data is the, I'll call it the constant data, which is. What kind of outputs do I want? What is my brand? What are my brand guidelines? How frequently this needs to happen, what needs to get put? Like there is the data that that is the consistent data in the machine. The other part of data is the ongoing new data that gets fed through the week to the machine by Nikki in any way she wants. This could be a voice message, this could be dumping in a screenshot. This could be a link to an article. This could be whatever she wants, become inputs into the machine and then the new machine knows how to take the existing data that it has plus the existing process that was defined and it knows how to process the new data in order to create all these standardized outputs. Going back to what I said in the beginning, that's a lot of touch points and concepts that needs to happen and the way it happens. And that's the perfect segue to your next thing is skills and skills. Yeah. Are. These mini processes that are, you know what, I will let you explain because you're here then I, I don't Yeah. I brought you, I brought you for a reason. You don't have to. Yeah. But, but this is, this is how it all connects, right? You need all these components for systems as complex as this to work. But then the really magic is the minions behind the scenes that are doing all small steps,

Nicci Sweaney

you know? And I think that's, that's a really important point to drive home because, you know, I know that there'll be people listening that are in various stages of having built out similar things or not. But what I see is there is, there is a flood of this online, right? There is a flood of like, comment this word, take my, um, template, you know, plug and play. This does all of my marketing one click. Those things are only the structure. So what you've just talked about there is when people are sharing these systems, you're only seeing, I look at these three newsletters, I scrape new headlines, and I pop it all into a newsletter, and then I repurpose that into social media content. That is only the process of how you create content in a given week, or how that person is sharing that they create content in a given week. That is not the differentiator. And just plugging in one of those systems and having it do content for you, I guarantee you it's gonna be pretty bland. It's not gonna feel on brand, it's not going to speak to your audience in the way that you would like to speak to your audience. You're gonna spend up so much time editing it that there's no point in automating it in the first place. So the really important work comes down to. It's all the stuff at the start and it's all the really boring stuff, but this is what I sell to my students as well. The really important work is understanding exactly how you would like it done and exactly how you do it now, but also how you could make it even better when you're not limited by the confines of human resources and time and capacity. And then it is the minutia detail to which you describe how to approach and execute on given tasks, which becomes your skills files. And what I think is, you know, my background in science and research gives me. Quite a kind of edge in creating skills files, because I come from a background of let's look at the best evidence. Let's look at what's saying, uh, who is saying, what about this subject? Let's look at what the data suggests around it. Let's look at what's current. Let's critically think through what applies to me and what doesn't. Let's try and find the gaps in that story. What's the evidence that this is the best way to do it? Who exactly am I aiming this for? So my skills files are huge because I put so much research into finding out what is the best information, particularly in places where, you know, this is not like I'm not from a marketing background, but I understand how to find evidence and weigh it up and choose what feels right for me that makes my system work for me and my audience. And doesn't just make it ai, generic slop that kind of just blurs into the background

Isar Meitis

a hundred percent. I'll say one more sentence and then let's dive into what are scale files and how to build them and how they work together. But the. For me, a very similar approach, and all I'm doing is I'm amplifying what you're going to say. Like, my marketing system is very different than this, but it's identical to this. Meaning it, you know, because I have the podcast that becomes the source of everything for me. But the system knows how to monitor what's happening right now on the different social platforms. It's looking for cues on what's currently working and not working from a style perspective, from a length perspective, from a graphics perspective. Mm-hmm. From a, uh, how long are the sentences, what are the gaps between them for the different platforms? And then it looks for my content, the stuff that I, yes. Speak at stages and I have recordings off my, uh, weekly, uh, AI Friday Hangouts community thing that's happening every single week, the podcast, all these things. It

Nicci Sweaney

mm-hmm.

Isar Meitis

Then looks for content in there. Things that I say that I think that I analyze. And it puts it into the formats that are currently working successfully on specific platforms. Kinda like what you're saying. Yes. This is the science part of it. The science part of it is I don't need to know what's working. I can mimic what's working because these systems are very good at analyzing that, but there still needs to be me in the content. Otherwise, it will be the same vanilla thing that everybody else can create out there, which is not interesting.

Nicci Sweaney

Exactly. And the point that you just said there around, you know, you, you don't need to know what great is, but you can ask these systems to look for great. But that even in and of itself, you still need to remember to ask for the system to look for what is great. Like you need to understand the metrics. And that comes back to process, right? The process by which I decide what is worth talking about isn't just to Google search AI news. It's around, you know, women in ai, it's around tech developments, it's around equity and governance and regulation changes. And then of all those headlines. How do I rank what it's worth talking about?'cause I only usually talk about kind of four or five new headlines in any given newsletter. So how do I make those decisions? And then how do I articulate how I make those decisions so that a system can understand how to go about the same thing, which brings us two skills files. So people may have played with skills files already. Uh, there is certainly, I've gotta say my recent post about how to make Claude skills has gotta be one of my best performing pieces of content. You know, everyone is, is on fire about skill sets, but skills files work regardless of platform. When you get into the land of AI agents, all a skills file is, is a markdown formatted document that uniquely describes how to execute on given tasks, right? So when we create a skills file, what happens is, especially when you're working with agents, I liken it to, um, like a transformer. Right. So a transformer, so an AI agent is going along and it's, it looks like Bumblebee, it's a car and it's like, yep, I'm pretty cool and I'm fast and I can like do stuff and I'm useful. And then it's faced with a situation. And in that given situation, it transforms into the appearance that it needs to take on in order to deal with that scenario. So when we're working with AI agents, it's like the same thing. It's like Swiss army knife, transformer vibe, right? If you've got hundreds of skills files that describe all of the ways that we do particular task in our business and then we task an agent with executing on something, that agent can say, alright, I've gotta transform what is a unique set of skills that I've gotta pick up and push together in order to make me the best person to execute on this job. So you the more skills the merrier, right? But what it does. Before I talk about the technical setting them up, but what it does, it, it pushes us to reevaluate what's possible. Because when we describe roles in the past, and even when we're talking about custom bots and assistance and workflows, when we describe a role, uh, we are confined by how that one role would work. We, we think about it like humans. If I hire a marketing person, you know, I hire someone that's great at emails and copywriting and then I might have to hire somebody else. That's really great at social media'cause I have to have that differentiation because not everybody's great at everything. When we work with agents, we're no longer confined by, I have to have an agent for this and an agent for that. You can have an agent that has access to a huge amount of skills and the agent can pick up whatever the best combination of skills is to execute on that job. So how you create a good skills file, that's

Isar Meitis

where you gotta work. I'll pause it just for one, for one second just to comment on that, to help people understand how this works. So. Many of you listeners, I assume, have built at least one custom GPT or a cloud project or a cha PT project, doesn't matter. One of these things. And really a skill file is kind of the same thing with two. So it has instructions, which is that marked down file. It's just a document. Think about like a Word document and it has access to other pieces of content, which would be like the knowledge base of the tools. And it can use tools that is accessible to whatever, uh, you give it access to. So this is a file, uh, the markdown file. But when you build really sophisticated custom gpt, you would build three or four of them to complete a process or five or six, and then you would move from one to the other to progress through whatever big. Project you're working on or a more complex process. Mm-hmm. This is very similar. Only you don't have to figure it out. It knows how to call the right capabilities and the right scales, the right quote, unquote custom GPTs as it needs them when it needs them, depending on the specific tasks that it's working on. So think about going back to what, what Nikki said. Think about having a toolbox and the more these skills you build, the more things are in the toolbox, which means now when you run a process, your different agents that talk to one another can pull any tools they need in real time. You don't have to tell them which tools they need to use. They figure it out on their own, and that's where the magic happens. Going back to what Nikki said, yeah, I don't need to know how this works. I just need to know that they have the tools to make it work and to make it work well. And then you just see the magic happen.

Nicci Sweaney

Exactly, and so your skills is really where your competitive edge comes into. You know, I talk to lots of businesses around what the future of of work is going to look like in the industry. One, no one can tell you'cause no one really knows, but two, what I always say is that your competitive advantage will just be the humanness in how you execute on, on your AI processes at some point. Like right now, yes, it's very cool. You know, you and I are talking about our marketing systems and there might be people listening that this is newer information to, and they're like, wow, you know, you just do like three lines of in your computer console and suddenly you've got 20 pieces of content every week that are on brand that actually perform well. Yeah, it's really cool, but at some point this will be so accessible and so frictionless that everyone will have the ability to just say, I want a marketing system. And we will have AI tools that go ahead and create the databases and the brain context and those skills files for you. That won't be a competitive edge. Your only competitive edge is how you construct that process and your special humanness that you insert into these AI personas. So when it comes to skills, I really advocate for spending a lot of time on your skills files because that's how we insert any sort of personality into our processes. So for me it comes down to, alright, let's go through a research process of what does the evidence say around. What does well in this particular thing, and that could be marketing, it can be admin, it can be email, triage, you know, hacks that you see on YouTube. It can be things around what makes a good, um, operations person that writes fantastic standard operating procedures. Anything that we do in business, there is evidence and there is research online, easily accessible, that shows you what excellence looks like and that's particularly important when it's not your area of expertise. So it's doing research and obviously leveraging AI to help you do that research as well and help synthesize it. It's deeply knowing what good would look like to you. So if this system performed really well, what are we aiming for here? For me, with marketing, it's driving more engagement to get more people onto my email list. And it's also having kind of clear objectives for every piece of content. It's not content for content's sake, it's content around certain objectives that I have in the business. So you have to really understand your goals. Research what is good, and then it's about combining those things. And again, leveraging AI to help us flesh those documents out, but really combining those things in a easy, in a structured way so that it makes it easy for AI to digest and to understand what's important, which is where the markdown of it all comes into so that we can structure this text in a way that AI can disseminate and actually get to. The truth of what you need it to know, and it's understanding that we really wanna break down all of our tasks into the most kind of nuanced, smallest piece possible. So when we think about that, something like marketing, so for example, when my marketing system repurposes it into social media, it not only has a huge task file on how to repurpose long format content, but it also then has sub reference files on exactly what my LinkedIn voice sounds like. A whole file on how to write hooks that grab attention, an entire file on how to construct good carousels. It has a whole file on our brand manifesto who we're speaking to, and the actions that we wanna create in our audience. So you've got all these kind of sub files that are informing how that bigger skillset works as well. The more flesh that you can put into those, the better. But for people listening, it really does start with good research on what. Excellence looks like in that given piece. And again, leveraging AI to help you find peer reviewed literature. I love using site ai. You know, we want AI that can transcribe fantastic and well performing YouTube videos on a given topic so that we can learn about it. It's scraping news headlines. It's looking across forums about what performs well, what differentiates the top 1% of businesses in this niche? Where could I get to? Because you're not limited by time, money, resources, and human capability. You're not limited by the help that you can afford to hire. You're only limited by your imagination. So if you can choose that top 1%, you can emulate that into your systems. And then also human quality control. So the last thing I'll kind of talk about in skills is it's really important that your skills understand what they are allowed to do and what they're not allowed to do, and where they should look for help or guidance or input from humans and what they're allowed to have access to and not have access to. So I talk a lot about ethics, but that's always a really important part of any kind of structure and instructions to ai. And particularly when we get into land of AI agents, there has to be really clear parameters around what is in your remit and what is not. And that's true of human positions as well. So it's not unlike hiring someone and saying, yeah, this is your job, but this is not your job. That's my job. Don't step into this territory.

Isar Meitis

Two things. One is, if you think about what Nikki said, it's basically having really good SOPs, which most companies don't have, but that's basically what it's, right. Yeah. Uh, and it's SOP at the very granular level. Of every aspect of your business. And I think the interesting thing that is going to happen in this world, uh, which so far has not been the case. And then, and then what I want to do is I'll give you time to maybe prep it. If we can open one of these files, one that you can show and you don't have a problem sharing, and we can go on how it's structured, how it's built so people understand. Uh, but we're, we're going through an era so far, a job was a set of tasks. Mm-hmm. That was the definition of a role. What's your job? I'm a marketing specialist. I'm the head of this department. I'm, okay. So what does this job mean? Oh, I do these 65 things. Yeah. That's going away. The job will be what is it that you need to achieve for the organization, which, what a job actually is.

Nicci Sweaney

Mm-hmm.

Isar Meitis

And the tasks will be done. Right, and so the way we have to start rethinking about work is the role and the tasks are completely separate things. The role is you are in charge of achieving this goal or these goals for our business.

Nicci Sweaney

Yes,

Isar Meitis

you don't have to do the tasks. You have to achieve the goal, and now, yes, you have this endless minions that can help you do it. You still need to define the strategy and you still need to monitor, you still need to create new ones. When there's gaps, you still, you have, there's still the critical thinking, the management, the strategy aspect of it, which like you said, many, many, many, many people don't have those skills, but they will have to develop them because everything else will become irrelevant.

Nicci Sweaney

Yes, exactly. And you know, there's really good data and predictions around what the future of work will look like for humans, given that that, you know, probably humans will stop having a job and they will just have many smaller jobs for lots of different organizations where they kind of manage these AI systems and they provide that oversight because they're injecting that special human quality or that special, um, quality assurance lens or a particular way that they approach things that makes these systems better and that will become our job. Right? Because, you know, lots of more, almost everyone has a mix of thinking and doing, and it's usually heavily weighted towards the doing in their jobs. But as you said, when that goes away, then how do we fill our 40 hours a week? And in my business, I, um. I have chosen to work part-time in my organization, so my entire team, nobody works full-time. Uh, if you work four days a week, then you get a full-time wage. But we've made that choice because we use AI so much that, you know, I'm kind of like, well, I've gotta walk the walk as well. If I'm saying it could save you so much time, I've probably gotta try and take a day off and, and really kind of live by that standard. But, you know, I'm no fool. I know that most businesses are not going to necessarily make that choice. We'll just want more done in the same amount of hours. So the role of the human will become, yeah, this, this manager of AI systems. And you might have to do that for lots of organizations to create the same sort of full-time job, because organizations won't need you to also execute on it. Once you've made the decision and you've delegated and you can provide feedback, then you're onto the next.

Isar Meitis

Okay, so let's dive into a skill file or skill files.

Nicci Sweaney

Yeah. Cool. Alright, so let's have a look at a couple. Okay. So what I'm going to open up, okay, so I'm showing you here, this is just my folder with my skill files around, uh, my copywriting. Right. So this is, this is for how I write things like sales pages, landing pages, even how it informs how we structure emails as well. It's not necessarily always around a marketing sales objective, but it's one of these skills that if I'm doing a new offer and a landing page, it would rely heavily on this skill. Whereas even when I'm sending out my weekly AI newsletter, we're still referencing this skillset because it still informs kind of who we're talking to and how we speak to them. So there are five files here. So there is the main skill file that describes this is how I do copywriting and persuasive copywriting. And then to inform that skill, and I'll switch to these skills in a second, but we have four subset reference skills. So we have one entire skill document that is all around my voice and my lens on ethics and how that interacts with the way that I approach sales psychology. The things that I will and won't say in my brand that feel aligned. Then there is one that's specific around format playbook. So this is how we do the difference between a long format, short format kind of online forums, wherever our copywriting is appearing, what does that look like? Then we've got a whole document that shows all the research into bio-psychology for our particular audience. What makes them tick, what are their towards mechanisms, what are their away from mechanisms, and then we have a whole piece on our authority and differentiation. I rely heavily, you know, there's a lot of people in the AI space, so my key point of differentiation is coming from a data science and business consulting and academic background. So there's a really big file on what makes me, me and why I think I'm worth kind of listening to. So if we just open up, I'll say,

Isar Meitis

I'll say one thing before we dive into the actual files themselves. So you now, if you're listening, you're starting to understand two things. One is why. This is different than a lot of other things you see online. Like, oh, I'll get my a hundred prompts and you'll be able to make a hundred thousand dollars. This is very, very different than that because there's a lot of strategy and data and research that goes into that. The other thing that I will say, because that's the scary part, like, oh my God, this sounds really, really scary. It sounds like a lot of work and I don't know if I know how to do this or if I have the time to do this, and I'm sure many of you're thinking of this right now. The reality is yes, it is a lot of work, but yeah, it's a lot of work that if you set it up once, now you have a content machine that grows your newsletter a thousand people a week and gets you a lot of business when you write one sentence, and so the ROI is not 5%, it's not 20%, it's 5000%.

Nicci Sweaney

Yes.

Isar Meitis

What that means though is you need somebody. In your organization. Doesn't have to be you, but you need somebody in your organization who have enough white space on their calendar. And I can tell you for myself, and I'm based on what I'm seeing Nikki doing right now versus what we talked about just a few weeks ago, she did something very similar. I blocked most of my calendar for January, February, March until the foreseeable future to build these things. Yeah. In the short term, I'm working harder than I worked before, but every one of these processes, and I'm usually building three or four in parallel at any given time.

Nicci Sweaney

Mm.

Isar Meitis

Uh, everyone I figure out, I now have a new department, a new group, a new process in my company that now runs 90% on autopilot with 10% supervision. Yeah. By me or other people in my team. Yeah. And yes, it is a lot of work. If you cannot do this because you don't have the bandwidth, you need to hire somebody to do this, or a year from now, two years from now, all your competitors will be able to do. Better work than you at 5% of the effort. And then you'll, you'll be really busy finding a new job. And so, uh, that's kind of like the big picture. It's not, there is no, oh, I, I download this thing and I click a button and it works. There is work that needs to be done. Yeah. You not agree to this quality of output.

Nicci Sweaney

Yeah. And it's such a good point. And I think the other thing for people potentially listening is, you know, because there has been so much change in this space, I get people that say, you know, well, if I do all of this, will there just be something that comes and makes it all redundant tomorrow? Right. So I think that that's really important to acknowledge because you and I have just spent all of 2026 rebuilding everything and you know, really kind of fascinated with all these bits and pieces and how do I make that tick and what can I do with this now? But what I tell people is it's still a competitive advantage to be ahead. So the businesses, you know, for example, some of my. Longest running continuous consulting programs are now running over two years, right. Of continuous work with them. What happens in those businesses is because they started so long ago, they're able to really quickly pivot to the next thing because they already understand the fundamentals of having to be very, um, assertive, having to be very discreet in terms of like how much information and how well they put it together of understanding that, like you just said, these need to be detailed SOPs. I always describe them like they need to be recipe equivalent of SOPs. Right. Because quite often people, even if they have SOPs, they're writing SOPs based on what they already know and giving it to somebody else in their team that already works there. So like you go and get the thing from the file and then we save it to the downloads every week. If I say something like that to somebody on the street, they're like, what file? What downloads? What are we looking for? How often do we do this? When do we do it? What qualifies if I don't do it? A recipe is designed so that you at home can make the cake from start to finish and actually get it to, you know, presumably look like the picture and taste okay without a chef or anyone that's ever made that cake before in your kitchen with you. So what I tell people is that when you're thinking about this, that's the equivalent. I need to hand this to someone on the street, like a, a clever person on the street, but a random person on the street. They should be able to sit down at a computer and execute and get to 90% of what I would've done. So that's the sort of detail that we need to have. So in these skills files, um, oh, and, and the other part about, you know, not having time. Yes. Blocked out lots of time. The other thing I do is I do it in the car, so, so often I actually travel. A long way to airports because I do a lot of travel. Um, but every time I'm in the car commuting, I have a list of all of the departments across my business that I'm wanting to automate. And I pick one of those departments. And then in that department, I ask AI to interview me about what are we currently doing? What are all our processes? And then I just pick a process off that list and I just go to town. I will talk for 15, 20, 30 minutes about exactly how I do it, what I think about, where I make decisions. And then I say to these tools, if anything is unclear, if you couldn't exactly replicate this, if I'm missing edge cases, I want you to ask me questions back and gather as much detail as possible and then just get these tools to help you put all of that together. Right? And so that's how I kind of tick through them. So there's exactly an example of that where I was talking to. Claude to have it execute on what I define as persuasive copywriting. Who my audience is, what I like, what I don't like. What are we talking about? So in this persuasive copywriting file, they have an overall description. I always have an overall purpose and objective, and then we go into all of the things that matter. So you can see here straight away, it's saying when you are doing copywriting, I also need you to have a look at all the reference files underneath this before you write anything. But we have sections in here through why I'm different to everybody else. The AI education space is flooded with people who do X, Y, Z. This is what I don't do, or this is what I do differently. This is the context for it. The authority must be real. How I approach, you know, I come from a science background. I find marketing quite uncomfortable because in science you're never allowed to speak in absolutes. You can't say you are the best. You can't say you are the most, you can't say this absolutely will save you X, y, Z hours. You're not allowed to say that in science. So I find that quite hard in marketing. But striking that balance between how do I make people feel confident that this will deliver results while feeling true to myself, that, you know, I can say, this is what I've done with it and this is what other people have done with it. You know, this is up to you. So this is painting the before and after pictures of everything they have. These are the steps of the exact workflow that it will follow. So step one, step two, all the way down. And then this is the core message that you have to keep in mind. So you can see that that's approaching a fairly long document. But not only is that the document, but we also have all of these subset documents that run through why am I the authority? What sets me apart from everybody else who is. Audience, what is their psychology? And again, I leveraged peer reviewed literature around the psychology of my particular audience. What would make them buy? What is keeping them up at night? Let's look at what Google searches they are performing. What questions do they already have? These are my format playbooks for how I write across each one of these platforms and exactly what that looks like. And then this file goes through the essence of my voice and my stance on ethics and the words that I hate and the tactics that I hate, and the things that I really like, and those non-negotiables around how I approach my marketing, given my particular lens on ethics and equity and all of the quality checks that you have to perform before you write. So those five files are involved every time I'm copywriting on top of. The files that say, this is how we organize our funnels on top of the files that say, this is how I write on Instagram, on top of the files that say, this is what our offers are and this is how we launch campaigns. And we're using that transformer approach to gather and execute on a given task with access to all of these skills files. So I would probably approximate at the moment, and I, I'm only kind of fully through marketing and admin and operations. I have not fully gone through every other department in my, in my business yet. I would say I probably have oh, 60 or so skills files across those two sectors.

Isar Meitis

Incredible. I wanna say a few things because we're starting to run out of time and I could probably again talk to you for the next five hours and we can keep on going deeper and deeper, uh, a, a, a few things. Number one, as I said before, this is not a. Magic wand that you wave and it just happens. It requires you thinking about it. Number two. Those of you who are watching the screen, it looks kind of like code. Uh, and that's how markdown file look like. Now you don't need to know how to write markdown. I have no, I don't know if 50 or 60. Yeah. You know, probably 50 or 60 files, uh, kind of like similar ballpark, and I'm adding a few every single day. Uh mm-hmm. I haven't written a single one. The AI actually writes them. So as Nikki said, yes, I just explain to the AI what I want, and it decides in many cases how many subsets there are. Sometimes it makes wrong decisions, so I always ask it and it's built into the system, uh, message that ask me before you decide what skills go where, so I can tell you whether I agree or not. Uh mm-hmm. And, and there's. Probably 50% of the cases where I disagree and I do something different than it suggested in the beginning. Uh, yes. Going back to what Nikki said, that's kinda like how I bring myself into making the process, uh, better. Mm-hmm. But you don't need to know how to write these things. The AI knows how to write them. It knows how to store them. It knows how to then pull them as it needs them. The question that I have to you, which the more technical people are asking and the rest of the people will potentially ask after we explain every one of these tools, has a limit on the context window. A context window, yeah. Is the amount of data that any AI tool you use have a limit of how much data they can run in a single instance. Uh, once you run out of that data, one or two things happen. Previously until a long time ago, about two months ago, they would just get drunk and start forgetting things. Yeah. And now they have this process, they go compacting where they basically for themselves summarize what happened before and then continue with, with more space in their knowledge. But it still means that as it's getting started, you just, everyone wants to write something. You need to read 20 documents that already consumed Yeah. Most of its context window. Do you, how do you, A, work around it and B, what are best practices to maybe limit the issue or, or, or minimize the impact.

Nicci Sweaney

Yeah, so there's a couple of different ways, um, that we kind of go about it. I mean, one part is especially the things that you're going to do on high rotation, it's creating the full skill files and then also creating an a brief version of the skills files with the core information that you need to execute on that given task. So, you know, when I'm doing, for example, the newsletter, the marketing automation that I was showing, we, uh, we are particular around which skills that we are telling the agent to use in order to execute on that. So we have one that is, how do I write my newsletter? And we have one that's repurposing, and then when we're repurposing, we then say, okay, now you're going to have. Access to X, Y, Z skills, right? And so when you can kind of execute on this, because you can create these agent swarms that go off and execute different parts of it, you can kind of leverage various context windows for each subtask as part of that larger task. The other thing that we're also doing is I'm currently playing around with building a dashboard for my agents where they start to log their decisions at key points throughout. Whatever the process is that they're doing. And so what that is what I'm hoping that will enable me to do. I haven't fully fleshed it out, but stay tuned. But what I'm hoping that that will enable me to do is track key decisions to, one, improve the way that the agents work. Two, give them a starting point for when they are coming up with those compacting issues so that I know that they have remembered the right things. Because when we let agents compact, yes, they're gonna make a decision. They normally get it pretty correct, but particularly when we're doing something complex, I want there to be somewhere where I can control what information they're acting on. And through that improvement process, what I'm hoping is I'm also going to build a bit of an agent verification system where every now and then the agents can use that memory in order to decide what can be trimmed from their context files. So I'm kind of like a start off with everything you can possibly think of, take pieces out of it if you need to kind of shorten that context window and then over time also improve and refine that process so you're just really nailing it every single time.

Isar Meitis

Awesome. Again, my 2 cents, uh, something very similar to what you described, I have two files, uh, three actually now, so it keeps on changing. Mm-hmm. But as of this morning, there are three different files that go across all my projects and there are different kinds of lessons learned like organizational.

Nicci Sweaney

Yeah.

Isar Meitis

Knowledge and, and they're, they're on different topics because they started getting too big. And I'm like, okay, this is ridiculous. They don't all need to know all. Uh, yeah. But there's one file that defines what's in the other files and the agents before they load. So in the system message of all of this, it tells it to go and read that one file and then it decides based on that, which are the other files. So there's ways that you can build into your quote unquote agent infrastructure that, that reduces the amount of thing they have. Like every project I'm running has a project progress file. That project progress file and gets above a certain length. The older stuff gets archived because otherwise, as soon as it reads the document, it's out of context before you do it. Yeah. And that's kind help anymore. So the reason I ask you the question is because people need to keep that in mind because otherwise you're, you will. Literally running an agent that can run because you give it so much information that it doesn't have enough memory or capacity to do Yeah. The things that it needs to do. And it's just something to take into account.

Nicci Sweaney

Yeah. And it's also, um, you know, if we think about this, like I, I'm always telling people to think about this. Like, we think about interacting with, with people, human brains are almost limitless. Right. But whenever I walk into a workplace, I have so much context and experiences and life experiences and different things that I, that I know about. I don't use all of that to execute on a given task.

Isar Meitis

Yeah.

Nicci Sweaney

So when I think about what makes great skillsets and how you kind of build out your infrastructure. I, I'm like, document as much as you possibly can. Get it all out of your head. We want everything. Having a system, a process, you know, data behind it, evidence behind it, a persona, a personality, what they can do, what they can't do. But then as you execute on tasks, and this is a little bit of trial and error, and this is why, you know, people like you and I find it so exciting, but it's very understandable. There's also lots of people that are like, this is very overwhelming and I can't, I don't wanna engage. Because that's where the, again, that work comes into it, where I've got everything. Now, what do I actually need in this given scenario? And again, a little bit like a recipe, you know, we don't dump in all of the salt and pepper straight away and then find, oh wow, it's inedible and you can't undo that. It's kind of like a slowly stacking of skills. So again, while I talked about making all these skills files, I also give people the advice of like, start off with the basics first. Do as little as you think will be relevant. If that creates the outcome you are looking for, you don't have to add more for the sake of looking clever. You can do it just like that. If you find actually I'm having to edit it or give it corrections or it's not quite always taking the right action, that is your sign that you need more context in there. And then there are different ways of managing how we can structure that context as well.

Isar Meitis

Awesome. Quick recap and then I'll let people know how they can work with you, hire you, find you learn from you. Uh, so the new universe is very, very different than what we had in 2025 in the sense that literally anything that can be done in front of a computer, you can build AI agents and skills to do instead of you and or your team. Mm-hmm. This is not future technology, this is the current present. To do this well, you need to understand. The tasks that you're trying to perform, or the outputs and the outcomes that you're trying to achieve, and what great will look like in each and every one of these things. And then from there, you can reverse engineer the different steps. And AI can help you in each and every one of these steps. And the last thing that is very critical that, that I do and Nikki does, is you still have to have the human touchpoints and oversight. So Nikki does it in notion, I do it in Clickup. I have some things running through Slack. I have some things running through WhatsApp, but it doesn't matter. Like I have ways to interact, impact, change, manipulate, verify, uh, the AI. Outputs because otherwise you have slop, vanilla that everybody else has. And the fact that you created it is not gonna make you different. Uh, just gonna make you be like everybody else and then nobody's gonna hire you and so on. And you, yes, you can get away with this in the next two to three months, maybe six, but then it will be over. Uh, Nikki, this was absolutely brilliant, brilliant, brilliant. If people wanna learn from you, take your courses, hire you to build this stuff for them because they're terrified, they won't be able to have time to do this on their own, uh, what are the best ways to do that?

Nicci Sweaney

Yeah, absolutely. So you can reach out to us. Uh, you can send us an email at hello@aiherway.com au or find us on the web@www.ai her way.com au. Or you can also look me up on Instagram ai her way or on LinkedIn as Dr. Nikki Sweeney. Always willing to chat. I always share a lot. We've got a lot of free resources as well. You can come in and you can try our courses for free. We've got a whole bunch of systems that you can take and have a go at. Um, you know, we also have these systems that interview you about your work and then give you skills files back. So if you are struggling with that overwhelm, we've got a lot that can really help. And for overseas listeners, I, I'm obviously based in Australia. I'll be over in the US in April and May this year, and then in South America. So look, lots of places to catch me. Would love to have a chat.

Isar Meitis

Awesome. Thank you. Thank you so much. This was absolutely fantastic, uh, as I hyped it a lot in the beginning, and yet I Thank you. Over delivered. So I, I really, I really, really appreciate that. I really appreciate you. You're gonna meet in person in April, so I'm, I'm really excited about that as well. I know. Very

Nicci Sweaney

excited.

Isar Meitis

Yes. Have a great rest of your day.

Nicci Sweaney

Thank you so much for having me on the show.

Head Shot

Isar Meitis

Host