A (3:08)

Today, I'm very excited to be joined by Dr. Nikki Sweeney. She is the founder of AI Her Way. Her mission is to drive ethical AI use for a better world. Her membership is AI for Impact Hub, and her mastermind is the AI Growth Machine. Nikki, welcome to the show. How you doing today?

B (3:26)

Good. Thank you so much for having me. I'm excited to join you.

A (3:29)

I'm excited you're here too. Nikki and I are going to explore how to build next level systems to apply AI to businesses of any size. Now, before we get into the nitty gritty and all the exciting stuff, I'd love to hear a little bit about your story. How did you get into AI?

B (3:45)

Yeah, I've been in a data kind of world for a long time. So I did my Ph.D. actually in conservation ecology. And that was, gosh, like 12 years ago. Now. I've been working in universities for the last 17 years. But as part of that PhD, I had to learn how to code, which if you'd gone back and cold little Nikki that she would actually enjoy coding, I think I would have died. But I had to learn how to code for my PhD. So that was an uphill battle. But one day it just kind of clicked and I really, really loved it. I imagine it's much like kind of learning another language. You know, one day, all of a sudden, you actually can understand what the sounds are and what the words on the paper are. So it clicked and I became a consulting data scientist. And I used to do really big projects for wonderful people like the United nations and World bank and AusAID and government. And, yeah, I spent a long time in that space. And alongside that, I was also teaching in higher education. And then of course, generative AI happened, right? And I remember in November 2022, I opened it up and I simultaneously had three thoughts. I thought, wow, this is so cool. I wonder what else I can do with it. 2. My God, why did I bother doing a PhD? This? Probably could have done it in a week. And three, maybe we're all gonna die. And I'm sure lots of people felt the same. But I remember distinctly running into my partner at the time, I literally jumped on him and shook him by the shoulders and I was like, this is what I'm gonna teach people, because this is going to change everything. And for me, it just kind of ticked all those boxes. You know, it was data, it was code, but it was much more human, quote unquote than that. It was really tangible. I knew people were going to have to learn it, and I just thought it was so fascinating and such an important space to be in. And that very quickly morphed into realizing how important it was going to be for women to be involved. So, you know, when I first started talking about AI, I was talking predominantly about how it was going to change education and the future of work. And after a few months of that and being really kind of deep in the space and doing lots of research and really being surrounded by AI stuff 24 7, I realized that women weren't a part of it. At that time, it was mostly just ChatGPT that everyone was talking about. It was about 70% male users, 30% female worldwide. I knew that women weren't in the machine learning workforce. We weren't the ones designing these tools. Yeah, we weren't the ones using it, so we weren't the ones getting the benefit from it. And I just saw that as a huge, huge risk. So that, that was really the driving force between particularly calling it AI her way and, and the mission that I have. But, but that's the backstory, code, algorithms, being in research, loving learning, loving teaching, seeing this big shift happen and just realizing that, yeah, people are going to need help. And I definitely wanted to be part of making sure that it was a force for good.

A (6:42)

I love it. So there's a lot of people listening today, men and women. Many of them are marketers, many of them are entrepreneurial. Why should we marketers and entrepreneurs focus on building ethical and sustainable AI systems? Because I know that's one of the things that you care about, right? Which is, first of all having it be ethical and secondly, having it be something that is sustainable. So talk to us a little bit about that.

B (7:10)

It's so important. And there's lots of different reasons. I mean, ethical and sustainable is often something that's kind of touted as well. It's just good because it kind of feels good for the soul. Right. Which only get so many people so far. And yes, it's good because it's good for the soul and you can feel good about the way that you're showing up in the world and how you're contributing. But you know, as people that have businesses that are in the professional world, it's really a competitive advantage as well. Right. So when you think about ethical, responsible use of AI, that's a key differentiator right now. If you use AI, sure, like, maybe that puts you ahead of the guys next to you because you're using it and they're not. But that's only going to be a competitive advantage for like, a millisecond of history, because very, very quickly, AI will be so embedded in everything we touch and do and interact with, and it'll be so frictionless, we won't have a choice about whether we're interacting with it or not. It'll just be how the world works. So at that point, just using it is not your competitive advantage, but using it in a way that contributes to a greater mission, that serves the greatest amount of good for the most amount of that is, in a way that is scalable and sustainable and will still be around in the next 10, 20, 30 years. That is a key differentiator. So there's real market value in doing an ethical way. And not to mention, if you don't do it an ethical way, you stand a real risk of damaging your brand, your business, your reputation, either as an individual or as a company. There are big companies that have damaged themselves by not doing AI in ethical way. I mean, notorious stories that Amazon were using AI to filter through cvs, right, for recruitment. And after a little while of using this, they quickly realized that if the role was technical, the AI tool was throwing all the women's CVs in the bin because it was based on pattern recognition of the past. And so it was reiterating these stereotypes. And no one had done the ethical, responsible thinking in the construct of this tool. They were just like, yeah, cool, AI can do that, right? Sure. Like, that's something that we hear so often. So taking the time to think through it and make ethical, responsible decisions is good for you, it's good for the business and it's good for the world. So I do think it's a really, really important part. I think it's the whole point of doing AI and being in this age, we are the cohort of adults that get to work out what this means for the world and for entire generations to come. If we don't do it right now, we miss our chance to contribute to that.

A (9:39)

Wow. First of all, very well said. I love what you're saying. You have come up with what you refer to as 5 levels of how businesses are using AI today. I would love you to kind of just do a flyover on what these five different levels are, and then we can kind of unravel the operating system concept. We're going to be talking about.

B (9:58)

Yeah. So the five levels, you know, it's our AI maturity index that we go through with every company that we consult with and even individuals and in our online AI for Impact Hub, that's what we kind of take people through this journey. So there are five different levels of using AI and I'm going to say most people are at level one and two and there's very little awareness about what kind of comes after that. So your base level of integrating AI and using it well is prompting with large language models. Right? This is using ChatGPT, Claude, Perplexity, PO Co Pilot, Gemini, any of those kind of tools that we put text into when they're giving us information back. That is an awesome skill and it is a skill that everybody should have. Right. Because it teaches you a lot about how these systems work. You can get lots done, you're really productive, you can do thought partnership. You know, there's all sorts of crazy, weird, wonderful, awesome use cases of that. But it is not the end goal. Right? And it's not the end goal because if we just use large language models, you'll only ever become faster at doing the stuff. But you still have to go and do it, right? I've still got to go and open the tool or talk to it or whatever. So that's not the full realization of AI because we want to be able to delegate work to these systems, not just have them do it quickly for us. So the next step are your low grade automations and custom GPT. So they're called custom GPTs and ChatGPT projects include, they're called gems in Gemini. This is basically when we take the power of a large language model and we give it a set of instructions. That's the next step. Because when you're prompting, if you get so good at giving it a task that it's repeatedly giving you a great outcome, then you don't want to waste time always giving instructions. You want those instructions to live inside your own custom bot that does that task for you again and again. So that's your next kind of step up. Now after that is where we get into the land of AI agents and AI assistants. So these things are often large language model based, but they have wide scale access to the rest of our knowledge base. So wherever your stuff lives, in email, in your calendar, SharePoint, OneDrive, you know, in Notion HubSpot, we've seen these tools just released inside of Claude and ChatGPT where large language models can connect these tools. You also have third party Providers that have been able to do this for a very, very long time. But this is where we give general instructions like we would to a human assistant. And then these AI assistants can go and gather all of the relevant information and give us a summary or analysis or an answer based on all the things that they've had access to. Now the next level is where we get into AI workflows. So this is where we have discrete processes in our life or our business that we can clearly describe that may go through a series of steps and also have AI involved. You know, a great example is, and a really simple one, somebody books a meeting on your calendar. If you are prompting a large language model, you might be able to say, hey, I'm, you know, having a meeting with Nikki Sweeney. Can you give me some notes about her? Go and research her on the Internet. I've still got to do that. By the time we get to an automation we can say, all right, let's watch my calendar, let's trigger an event based on when somebody books into my calendar. Then let's scrape their domain name, go and send an AI agent to research them on the Internet and on LinkedIn and gather all of the stuff that you can about this person. Then we're going to send all of that to our AI agent, that is the head of our sales. She's going to write up a discovery call script based on all this information and then we're going to send that to the person who's meeting with Nikki via email so that all the pregnants are done. So that's a workflow.

A (13:35)

Real quick, before we get to the pinnacle, I want to clarify what's the difference between these automations at the. So far what we've talked about is at the base level, number one, we've got these prompting with LLMs. Number two is we've got automations and custom GPT and number three, we have these AI agents and assistants and number four, we have these automated workflows. So what I'm trying to do is discern the difference between the automations at the second level and the automated workflows at the fourth level. Just because we use that word and people might be confused.

B (14:04)

Yeah, so great point. So when we talk about custom GPTs, they're automating singular tasks. So I might train a custom GPT or a project in Claude, right, to write great sales discovery calls, to sort of write these 15 minute scripts when people book in to have a chat with you, that automates that singular task. But it can't necessarily Go and perform multiple steps and take on multiple Personas within that. Right. It's just trained to write discovery call scripts. When we get to number four level automations, this is where we could tie together a series of those kinds of custom GPTs also with automated processes inside of our tech stack to deliver a whole outcome. So in that exact same example, we, we might have a custom GPT that's great at writing discovery call scripts, but what we've also done is take information from point X, give it to point Y. We've brought in another AI agent that's able to do this other thing. We've passed parts of information to our discovery call writer, we've passed parts of information maybe to our notion or our CRM. So we're taking on multiple Personas through that journey. And it's a bit like taking it from hey Mike, can you do this one thing for me? To calling in the team and saying, hey, can you all collaborate together to deliver this outcome?

A (15:18)

Okay, I like that. Now, going back to the agents, isn't it true that agents can do some of this? It sounds to me as if the agents can do this like on a personal basis, but the automated workflows are like lots of agents working together. Is that kind of really what we're talking about here?

B (15:31)

Yeah. So the agents have just released inside of ChatGPT and Claude. They're a new functionality and they do replace some of the stuff that we only could ever do by bringing in third party providers and being in that workflow level. So that's why agents sit at kind of level three. They can do some of this stuff. We can give them directions about things. They can't link to every possible tech stack, at least not right now. And for now, they also can't bring in yet multiple Personas into that conversation. So you can give an agent in ChatGPT description, go and look through my email, go and look through canva, pull this stuff, create something. But when we look at workflows, we can create discrete kind of personalities or task driven agents and then do multiple steps between those agents.

A (16:17)

Got it. So if I were to visualize a pyramid, at the base of the pyramid is what we're all doing, which is prompting chatgpt, Claude. And then the next level up is we're doing some sort of custom GPTs or cloud projects. And the next level up is we're using agentic AI tools that can do some basic stuff where they can like now all of a sudden work semi autonomously or autonomously on our behalf and then the automated workflows are like tying these things together to, to take projects from like execution to completion. Is that kind of how we would describe this?

B (16:46)

Yeah, exactly. The route. Yeah. And this is why it's a stacked pyramid, right? Because you have to get each of those levels right before you can do the next one. Because if you just dive in and build a workflow, if you don't give good directions, if you haven't defined how your agentic AI operates, if you don't know how to prompt a large language model, then you're going to tie together a bunch of your tech tools and you might have a really pretty looking 40 step workflow, but the outcome is going to fall short and you're going to be very underwhelmed and you're not going to know where in that workflow that thing went wrong because you haven't built each individual building block in your skill set and in the pyramid.

A (17:20)

So let's talk about number five because we haven't gotten there yet. So what is the, the tip of the pyramid?

B (17:24)

Haven't even got to the top of it. So the top of the pyramid is AI operating systems. So this is where we have kind of whole divisions of business, where we have an overall objective and then the objective is executed on by a range of AI agents and workflows and GPTs, all working together to achieve that outcome. So again, a great example might be if we talk about, you know, maybe in that meeting space, it's to increase our sales conversion rate by 10%, let's say. Right. And we can work on that objective with an overarching autonomous AI agent that then decides the best way to do that and can delegate the task to all of the workflows and agents that live underneath it. So that workflow I described before about booking in a meeting, writing a great discovery call script, that might be part of the options that are available in that operating system. Because that operating system is all about sales. So one part of it might be about doing great calls, of course. Then you want to take those sales call scripts, actually review them and see what we've done well, what we could have improved on. You want to write a report, you want to give that back to me to say, hey Nikki, next time you take that call, do xyz. And that's all governed by the AI operator operating system. It's driving the delegating of the tasks and the revision of those tasks to achieve said outcome. And it's doing that without human intervention. Of course. We always advise There should be human quality control at the output level. You should always be checking before you're sending stuff out into the world or acting on it. But all of that stuff would happen outside of you. So an AI operating system is like the head of a division in a business. It has staff underneath it, which consists of AI agents, AI workflows, custom GPTs, all that sort of good stuff in the pyramid. And it is the boss of everyone underneath it to execute on your objective.

A (19:13)

Okay, cool. First of all, loving where this is going. So now let's unpack where we begin if we want to build an opera AI operating system. Like, let's just talk the basics at this juncture. Like, where do we start?

B (19:26)

Yeah, So I reckon, you know, when you're looking at AI operating system and when we do this with companies and individuals, it's way less about which tools you're going to use or which program you might use over another. Right. It's much more about getting really crystal clear on, like, those roles and responsibilities and expectations. And, you know, that that maybe sounds obtuse to some people, but when you think about what AI is, it makes sense because generative AI is our best attempt yet to kind of replicate in an algorithm how human brains conceptualize information and retrieve and produce information. So when we think about hiring AI into our business, we need to think of it like hiring people. What is it that we are trying to achieve and what does that perfect person, and now we're going to replace that with AI. What does that perfect AI model look like? What are they responsible for? What are the expectations around how they would function? So we're not looking to replace people, but we are looking to kind of extend those staff roles and design really complementary positions. Right. So we want to start with your business outcomes. We want to think about, you know, do you want to shorten your campaign turnaround time? Are we looking for sales? You know, do we want to cut admin time and then work out which departments they would belong in? So the first step, I guess, to really put it simply, is to actually write out your roles and responsibilities of how you would like these people to operate. And we know, replace people with, with AI, but we want to get really crystal clear on what our expectations are that they would actually achieve for us before we even decide about how we actually technically structure them.

A (21:06)

Maybe you could give me an example, maybe through a marketing lens, of how something like this might kind of like. Like the things that maybe you've done this for your business or you've done it for people that are in your community or whatever. But like, let's talk that through a little bit.

B (21:20)

In our own business, for example, we have a marketing and campaign AI operating system, right? And so in our marketing and campaign AI operating system, this is a central space where all of our marketing happens and our head of marketing is given objectives. So as the boss, I say, okay, this is what we're working on, or this is what we've been offered, or this is our overall objective in that space.

A (21:44)

And by the way, when you say head of marketing, you mean AI or you mean human AI. Okay, yeah, good.

B (21:50)

Not a human. I'll clarify if I talk about a human.

A (21:53)

So, so, so far, when we, when we're talking about titles and roles, we're talking about AI roles. Okay, good, Keep going.

B (21:58)

Yeah. So in my particular business, I only have one ongoing human staff member. Sometimes we hire contractors, but everybody else that works for me is not human. Everybody else that works for me is AI.

A (22:09)

Okay, perfect.

B (22:09)

So head of marketing, AI driven. We give kind of overall objectives or we assign the campaign, Something might come in, we send it to them. That head of marketing has a look at that. So that head of marketing is an AI agent that's designed with really great marketing skills, overall objectives about our business, it's trained on the core values, our offers, all that kind of stuff that you would train ahead of marketing on. That AI agent then makes a decision about how best to execute that marketing campaign. And it may decide that we need, for example, social media campaign. It may decide on an email newsletter outreach program. It may decide on some cold call outreach. It may decide that we need to be booked on podcasts. It may decide that media might be best. And then what it does is it assigns those tasks to the AI automations that live underneath it, that execute each of those things. So let's just say, for example, it decides that we need a social media campaign. It assigns how much of that campaign should be driven by social media to the AI automation that lives underneath it, that's responsible for delivering social media content. Right? Then that automation takes all of the information about the campaign. And then it's also trained in our voice, in the way that we show up online, our brand guidelines and our brand language. And it produces all of the draft content towards that campaign. And what happens is that we have another automation that tracks every piece of content that is created so that as a CEO, I can sign into a dashboard and see where is this work at? Have we done 70% of it? Is there 100% done? Of it and I can quickly review the work that it's done and then we can approve the work that it's done and then it goes on to actually execute. So to post it on social media, to send out the email, whatever it is that that may be. And the head of marketing keeps an eye on how each part of those things are executed and does a revision and a wrap up and provides me, the CEO with a summary of, hey, this is what I decided to do. This is the rationale behind that campaign. This is what happened, this is how we executed. This is the wrap up. This is what I suggest for next time.

A (24:13)

I love it. Now you just scared the heck out of some directors of marketing that are listening to this podcast. What do you want to say? Let's just go in aside here for those people that actually already have employees doing some of these roles, how they think about, you know, like, I love that it's just you and one other person and that you've got this incredible staff of AI agents that are working for you. What about the businesses that already have staff? Would they use this to supplement their marketing department, perhaps in areas where they do not have support? Does that make sense?

B (24:43)

As much as this sounds like yeah, you've got all these people working for you that are AI driven. I never advise people to replace people with AI. So I have it because I'm very small, very lean. That is my objective. That's how I started and I've always stayed. I didn't have a big team and replace them with AI. I've always had a small team because we started as an AI business. If you already have people, people are your biggest asset. Right? But how we want to think about it is that we just made everybody the CEO. So what AI does is it really flattens organizational structures. Instead of being in a hierarchy, you're making everyone even because everyone has their own hierarchy of AI agents and systems that sit underneath them. And I think that that is an amazing opportunity because all of a sudden you unlock the capability of everyone to sit in a strategic, creative innovation space rather than a space where they're always doing the doing. Right now that's not always going to suit everyone, but the people that it suits. You are going to quadruple thousandth time what the value is that they give to you and the business because you free them up to actually sit in that place. And that's a really, really cool opportunity. So when we work with companies, we never advise to replace people. That is not the objective. The objective is to Free your people up from the stuff that just needs getting done and give them more space to be amazing at the things that really matter. So if we just think about that example that I gave you, it's not that we're posting everything that AI writes for us, right? It takes real skill to assign an AI agent or a system to do something in a way that is high quality. That takes expertise in that particular field. I'm great at AI, but I always get expert advice to build the systems in the particular skill set that they are replicating inside of my business, because that expertise really matters. And then the other part is, before you're just kind of sending that stuff out into the world, it matters to have someone put that extra 10% glittery fairy dust on top to make it sparkle. Because also remember, at some point everyone's going to use this stuff. So you still need amazing people to do amazing things to make you stand out. That's really important.

A (27:02)

Is your marketing falling behind? In my 15 plus years of tracking marketing trends, I've learned to spot the signals early. And right now, the pace of change in marketing is rapidly accelerating. What used to take years is now happening in months. AI capabilities that seemed impossible last year are already outdated, and I'm concerned that marketers who aren't prepared for what's coming next are going to be in a bit of a bind. The tactics that you're using today, the strategies that you spent years mastering, even the way you think about your audience, it's all evolving a lot faster than everyone realizes. This is exactly why social media marketing world exists. When you attend our 2026 event, you'll be positioned to lead AI strategy conversations in your company and for your clients. You'll be able to master tools and tactics before they become mainstream, and you'll also build authentic connections with future ready marketers that will have a big downstream impact on your business. Don't wait until everyone else figures this out. Join the marketers who want to get ahead and stay ahead. We have a really big sale going on right now. Visit social mediamarketingworld.info and grab your tickets today. Let's navigate this change together. So far, where we're at with this is, hey, if you want to get started building out this operating system, first you've got to come up with some sort of objective. I like the idea that you have this marketing department, if you will, that is essentially a very sophisticated set of AI agents and workflows and assistants and all these things and automations working collaboratively to Accomplish a mission or an objective. Objective. And I can imagine everyone who's listening. If you already have a marketing department, well then you could have a sub department that just specializes in, in writing better persuasive copy. Or you could have another department that specializes in automations or whatever. You could pick a million different things. And I really love the idea of, hey, hire someone who is an expert to take a look at what you're building so that you have good to begin with. Because if you're not starting with good training, for lack of better words. Right. Of these systems, you're going to end up with slop. Right. I mean, and that's where hiring a consultant might make a lot of sense to come in and look at your prompts and your data and all the things that are feeding the outcome so that you can make sure you have a really good system. And in the long run that system could be somewhat autonomous. Did I hear you right on that?

B (29:28)

Yeah, exactly. And that's a perfect summary. And that's the part that's exciting and that's the differentiator. Right. Like right now I know that lots of stuff I'm talking about, you know, it's potentially, it starts to like melt people's brains a little bit, trying to wrap their head around all the bits and pieces. At some point you will think about, hey, I once heard this chick talk about this stuff on a podcast. Turns out now that's how everybody works and no one's even thinking about it. Your differentiator then will be, yes, the expertise, it'll be the people power that goes into those systems. It's adopting that collaborative mindset. We're not replacing one with the other. We're capitalizing on what AI is great at, which is repetitive based, process driven, outcome deliverables, automation, stuff that lives in computers, stuff that's driven by data. Humans are great at creative thinking, strategic thought, thinking about things from other points of view, making decisions, driving direction, being outrageous and outlandish, and tying together odd, weird concepts and ideas. You want to double down on both of those things.

A (30:32)

Okay, let's get into the practical tactical. How do we build this kind of stuff? Because it's one thing to like conceptually talk about this, it's another thing to actually begin to be thinking about how to build this stuff. You've already hinted at some of the things, but I know when we were prepping our notes you started talking about creating departments and maybe we could talk about creating these different AI divisions or departments.

B (30:53)

So you want to start conceptually with top down, but you want to build from bottom up. So what I mean by that is when you are talking about the design, then yes, we think about the department that we may want to design and the roles and responsibilities that would live underneath that and how you expect them to be able to interact. So again, if you already have humans, you probably know all of this stuff, but whether or not you've written it down anywhere is another question. So it's actually documenting all of these expectations and interactions and how they would achieve a said outcome for you. Right? So that's your conceptual kind of idea of that. And you know, a little kind of side note, quick link is that you could put in something like that so you could talk about the roles and responsibilities and how they connect together. Open up Claude, turn on artifacts and say, hey, I'm designing this system. Can you now take all this documentation and draw a systematic diagram for me or draw a mind map or flowchart to show the interactions. Right. And then Claude can help you kind of map out conceptually what that space might look like. That is one thing, once you do that, you're actually going to park that and you're going to go back to building from bottom up. So what I mean by this is you're going to take a single role and responsibility that is inside one of those positions. So we're going to take a single responsibility around yet persuasive copywriting. Right. What I want you to do is I want you to start on Claude, Gemini, ChatGPT, whichever your preferred large language models is, and you're going to design a really great prompt prompt to get this large language model to do that for you pretty consistently to a pretty high quality level. So you need to give it lots of context, you need to give an objective, you need to be very specific about the expectations about how it does that. You need to be willing to test it, iterate it, refine it, right? Once you have that prompt working really well, so you can put in one prompt and it does the task to like 80% of where you want it to be, that's your signal that you're ready to create an AI assistant. So a custom GPT, a CLAUDE project, that sort of thing, with this set of instructions, right? Once we have that and it's working consistently for us, then we need to think about what is an entry level automated workflow that this person could operate in. So if we write persuasive copy, you know, when I get a particular email and I tag it as an idea for Our next newsletter, we want to send that email to our persuasive copywriter custom GPT and get them to produce a first draft of our email newsletter based on the information. I've passed it.

A (33:28)

Wait, wait. Just so we're clear, are you saying that you got some random email in your inbox that inspired you, you thought was really high quality and you want your custom GPT or cloud project to kind of learn from it and model something? Is that what I heard you say? Or is it something totally different?

B (33:43)

So like maybe you got some information that you also wanted to talk. So for example, like I get lots of newsletters that have AI news in them, like news headlines or tools that you might want to try. So one of our workflow is that we collect lots of different newsletters and news from online and we look through Google News. But I was, you know, simplifying one part of it down and we give them to AI assistants to start collecting ideas from all these different places and then they start to pick out ideas that might resonate with our audience as well.

A (34:11)

Okay, I like that. Okay, keep going.

B (34:13)

So you've done the first two steps in a large language model. The third step where we're starting to build our first workflow. If it's not a tech tool that is currently available in ChatGPT or Claude as one of these connectors that they've just brought out, this is where we start to go into third party platforms. So this is where you start to use make.com or N. They're usually our preferences. Zapier is a bigger market player in the low code automation space. It comes with a higher price point, but it's a little bit more user friendly. So it's a little bit less daunting for people to learn, comes with less customization. So that's why we usually use make.com or nn. So this is where you're starting to build out these first really simple workflows. And again, remember that your workflows are really just what a singular person would be responsible for. So you're emulating a process that someone would follow if they were in that role. Once you have a few of these working, that's when you're ready to start actually constructing your AI operating system. Because your AI operating system system, remember, may tie several of these things together so that when workflow A happens, so maybe we've got those ideas from the newsletter, we're sending it to workflow B that does this kind of stuff with it and we want all of the outcome Reviewed by a head of copywriting.

A (35:27)

Yep. Quick question. We're talking about automations here with make and N8N, which we have extensively talked about on this podcast. For folks that are new to those words, where is the knowledge base living? Because this is a question I think a lot of people are going to have. Right. Because there typically would be a database of some sort. Right. Where this stuff needs to live. And I think we should talk about this a little bit.

B (35:50)

Yeah. So when we do, us personally, AI, operating systems and knowledge bases, our kind of go to is Notion for that kind of stuff. So Notion is a blank slate for people that haven't played with Notion. It is touted as project management tools for neurodivergent people. Was. It's like original kind of advertising because instead of locking you into a structure of a software like Monday or Asana, they have a certain look and feel and you've got to kind of put your work into that look and feel. Notion is literally a blank page. You can make databases, you can make just wiki pages, you can make galleries, whatever comes to mind. It can, just like having an online notebook. So we use Notion a lot for two reasons. One, because of the flexibility, you can do whatever you want. So as a knowledge base, you can literally have an endless database that stores all of your knowledge base and can tag it by what it relates to, which department it belongs to, what role and responsibility it's pertinent to. Second is because it plays really well with automation tools. So Notion, it's very big in the developer space. And if you follow people with like N8N or that are giving away make.com blueprint, you'll notice a lot of time their lead magnets are actually Notion pages that they've published. It's really big in this kind of startup space for them to use because of how well it plays with automation tools. So it just means it talks to these tools in a way that's really easily accessible. Some other tools, you know, you've got to write like hard code or go and get like API permission to make these tools talk to each other. It's not the case with Notion. And the third part is that Notion has integrated AI inside of it. So when we run operating systems and we design them, we like to build in Notion. Because if you think about that knowledge base just as like a side note, AI inside of Notion can look through everything that it has access to. So AI inside of Notion can look through Notion. It can also look through your email and your calendar. It can also look through Slack, a few other tools. So what it means is when I have contractors, for example, come on board, they can ask Notion AI, hey, what are Nikki's brand guidelines?

A (37:58)

We're talking real contractors now, not AIs, right?

B (38:02)

Yes, real contractors, yeah. AI stuff, right. So a real contractor can come in and say, what are Nikki's brand guidelines? The AI inside of Notion can look through all of our documentation, all of the knowledge base, and retrieve that information, rather than people having to find which file it belongs in. But when we also hire AI, it means that everything lives in a central place so that I can actually say, hey, this AI tool, can you go and query Notion for information about the brand guidelines, find that, and then deliver that information to the next step in the workflow to actually get the next AI agent to execute according to X, Y, Z. That's in our knowledge base. So we love Notion for that.

A (38:41)

Okay, I have to ask the question because so many small businesses are on Google workspace. Is there a way to make this work with Google Docs and Sheets?

B (38:48)

Yeah. So Notion will integrate natively with Google Docs and Sheets, which means you can embed Google documents straight into Notion and they're accessible inside of Notion. It also means when you use Notion AI, if you ask a question about a process or you know, a meeting that you had with xyz, if it lives near Google Drive, it also can access and look through Google Drive for anything. That's Love it.

A (39:11)

Okay, cool. So we've never talked about Notion on the show before and I'm familiar with it, but I don't use it personally. But I know people that love it and I thought it was worthy of going down this path. And I know some people are thinking me right now, whether on the run or whatever. Okay, so we've been stuck for a little while by design on this automation. We talked about make or N8N. And do you have a preference as to which one you tend to use? I know one is way more technical than the other.

B (39:35)

Not really. It kind of depends on the use case. I mean, we're probably not super the right people to ask about that because we build so much stuff for other people as well. So there's always a bit of kind of client preference. Our big thing is that we go into businesses and we teach them how to build this stuff. So it's also sometimes about the appetite, who we're talking to, who we're teaching, what their tech stack is. My general advice is to have a little go at at least two tools. See what feels maybe most natural. And then also look at the tech stack that you currently work in. So all of the software programs that you currently and choose the tool that's actually going to talk to all those programs or as many of them as on the list as possible. It's no good if you and it turns out that 50% of your tech stack is not going to speak to N8N. So do a little bit of research.

A (40:24)

So what's the next steps beyond this?

B (40:27)

We've got all our workflows, we've got our space in notion. This is where we particular tie it all together. So this is where our AI operating systems live. So this is a central space where every decision is made, all the deliverables are tracked, all the summaries are put, all the information is held, all of these workflows kind of get triggered by. So this is where everything kind of comes together. So for example, with our marketing space, it lives as a central space in notion, it's literally just called our marketing and campaign operating system. In that system we have automations where we can send directives to the head of that department. Right. Those things live inside of ours in particular lives inside of make.com for this one. So we can send a directive to the head of that agent. The agent makes the decisions about how to execute it and then assigns tasks on the notion table to the various AI automations, workflows. AI assistants that live inside of the marketing department, they then execute their tasks and they track all of their tasks and deliverables back onto this notion space. So for me, once this is all set up, I only have to look at that one page to understand what's being worked on. How many campaigns are we running, how many pieces of collateral are out there, how have they performed? What are their recommendations for the next time we do this? You know, was it worthwhile? What was the return on investment? All those sorts of things that you would expect when you kind of sit down with the head of marketing to run you through how are things going? It all gets housed inside of notion.

A (41:56)

Okay, love it. So what about these agents though? I understand the workflows and stuff, but some of these agentic tools that are out there, like for example, ChatGPT just announced something I think called agent. Like how does that tap into all this stuff?

B (42:11)

Yeah, the agents are a little bit different to just an automated workflow. So the key kind of difference is that agents usually have broader access to more tools and they have more autonomous decision making power about how to access those tools. And what to do with it. So if I kind of break this down into like a practical example, an automation might be, hey, when I tag an email as relevant news in AI, we want a automation that scrapes that email and then we're going to send that to a custom GPT equivalent to have a look at anything that might resonate with our audience and then go and research those bits and pieces. Yeah, if that was an AI agent. An AI agent may be given a directive of constantly looking for and pulling out relevant ideas for AI her way audience. And then you grant access to the agent to your email inbox. And instead of it being a discrete workflow of when I do this, then that happens. An AI agent is able to say, hey, I'm just going to like keep looking through her inbox and I'm going to decide if I think it's relevant and then I'll scrape the email and then I'll do stuff with it based on that decision. So they're the kind of key differences for people that are out there listening and kind of wondering when do I do what? Right. Least amount of excess first. So what I mean by this is until you feel quite confident about building these things, I would always go with a discrete workflow before I built an agent. Because when you build agents you are giving broad scale access. You aren't just saying, hey, when I tag this one email I want you to do X, Y, Z with it. You're saying you're allowed to look through my whole inbox, which is a whole nother level of security and it comes with a little bit more risk. And you have to be very good at quality control and making sure that you have a regular maintenance system for making sure that no data has leaked, that everything is safe, that it's done the right actions again. If we think about becoming a leader, this is the difference and this is why we design everything in this pyramid system. Prompting with a large language model is like having an intern sit right next to you and you talk to that intern all day and every piece of work it has, you're like, oh yeah, like that part. Not so keen on that one. Actually this is how we do this one. We build a custom GPT that's like, hey intern, can you do that report for me? And you kind of trust them to do a report because you've done that like 20 times while sitting next to them, so that's okay. But you're not really going to ask them to do everything by themselves. When we get into automations, that's like you go and put the intern on their own desk and you're like, I think maybe the intern can work for the day and that's going to be cool. And like, I probably want to check everything before it goes out into the world. But like, you know, I trust them. I don't need to baby them all day. An AI agent is where we start to step into. I trust them and they have delivered consistently so much that I know that I don't have to baby them. I don't have to review every single step that they're making in this kind of decision process. The AI operating system is like having a whole division that kind of sits in a whole nother office. So that's the kind of way to conceptualize it. You want to baby it. And I know lots of people and this is something we come across. They, they know what's possible and they're like, I want that no. That's just a no. Because you're going to land in danger if you build up here at the top of the pyramid before. You've kind of earned your stripes along the way. So it's really important to take it slowly. You will get there and your end result is going to be way better than if you just jump in at the top end straight away.

A (45:46)

I love these pitfalls that you're, you're focusing on. Let's talk about overwhelm, because this sounds like this could get overwhelming pretty quickly. So how do you not get overwhelmed? Right. Because honestly, this is amazing but also complicated. Right?

B (45:59)

Yeah. It lead into the overwhelm. No, like, okay one. I mean, in all honesty, it is okay to be overwhelmed. Everybody is. This is my full time 24, 7 gig plus. I've had 10 years in data plus 17 years in research and statistics. It is overwhelming. That's the rate of it. Humans brains are not biologically adapted to deal with this amount of change. So it's okay that you feel like that first of all. Second thing is don't try and do everything all at once. Right? It is all possible. And I've got to say this is one of the biggest pitfalls, even with our consulting clients that we work with, is they want everything right now. It is all possible. It is just not all possible right now. You have to do the thing first before the next thing. So my biggest advice is to just pick one thing that is repetitive that you feel like you could explain really well. And that's very process driven. So what I mean by this is there's not a lot of nuance. And there's not a lot of, like, there's 40 different ways this thing could go. It's got to be a process that when this happens, then that happens and that happens, and then we get this outcome. And businesses have a lot of this stuff. So it's repetitive, process driven. You could explain it really well and it does not fill you with joy. Right? We want to pick something that ticks all four of those boxes. And I often say to people, just write down, like, think through your typical week. Write down all the things that you're doing, and then just get out a highlighter and, you know, highlight the things that, you know, you could explain really well. Put a little asterisk next to the things that are pretty process driven. And then, you know, put a smiley face or a frowny face next to the things that really bug you that you'll be very happy never doing again. And just pick one thing from that list, right? And just work with a large language model to write a really good prompt to get the large language model to do that thing the first time around. And again, my biggest tip here is to actually just talk to these models because if you're anything like me, you can talk to the high heavens about tasks that you do. You don't necessarily want to write like two pages of instructions. So just talk to these tools. Like, you've had an intern show up at your front door that is really desperate to work for you. Very, very happy to stay back late, very eager to please. All right? Doesn't take a lunch break, is never sick. It actually has no idea who you are or what you do. And you're going to look at this little baby face, fresh intern, and you're going to really clearly describe one job for that intern to do, one job for that intern to prove that they get to stay another day. Right? That's where you're going to start. Don't worry about trying to put everything together all at once. Start with something repetitive, process driven that bugs you. And then we're going to go up the pyramid from there.

A (48:49)

Wow. Dr. Nikki Sweeney, this has been spectacular. If people want to connect with you on the socials, do you have a preferred channel? And if they want to check out your company and all the products and services you have, where do you want to send them?

B (49:04)

Yeah, absolutely. So we are predominately on Instagram and LinkedIn. On Instagram, at AI Herway and LinkedIn, it's mostly underneath my name, Dr. Nikki Sweeney. So we're pretty prevalent on both those platforms. Our website is www.aihaway.com, but we're also going to put up a bunch of extra documentation around this. So we've got a video that actually walks through our marketing operating space and a couple of the automation blueprints that we use to operate that space. We're going to make that available for everyone that listens to the show. So it's going to be at pages.aiherway.com and then/SME.

A (49:41)

Okay, so pages.aiherway au SME. Is that correct?

B (49:48)

Yeah. Perfect. So we're going to go over everything we kind of discussed in a bit more detail, especially for visual people that kind of of want to see it all mapped out. And we're going to give you some of the automation blueprints that we actually use. So there's an awesome one that, yeah, when we press a button, it sends all the campaign material to a bunch of our agents that write up some content and then bring it back to the notion space. We're going to give you those blueprints.

A (50:09)

Sweet. We'll have that link in the show. Notes. Dr. Nikki Sweeney, thank you so much for sharing your insights with us today.

B (50:15)

Thank you so much for having me. It was a real pleasure.

A (50:17)

Hey, if you missed anything and we did talk about of a lot lot, we took all the notes for you over@socialmediaexaminer.com a70 and be sure to follow this show on whatever app you're listening to. And if you've been a listener for a little while, we would love a review. Also, let your friends know about this show and we've got some other shows if you have room in your listening time. We've got the Social Media Marketing Podcast hosted by me and the Social Media Marketing talk show. This brings us to the end of the AI Exploration Podcast. I am your host, Michael Stelzner. I'll be back with you next week. I hope you make the best out of your day and may AI help you become more successful.

B (50:54)

The AI Explored podcast is a production of Social Media Examiner.

A (51:02)

This is the year to finally come to Social Media Marketing World 2026. Grab your tickets right now by visiting social media marketingworld.info.