Leveraging AI

167 | Build an AI Dream Team That Works for You In Different Role & Personalities with Kay Stoner

• Isar Meitis, Kay Stoner • Season 1 • Episode 167

Share episode

AI isn’t just a tool—it can be your dream team.

Jim Collins says that having the right people in the right seats is the most important factor in business success.

What if you could build your AI dream team? Define the roles and personalities of each AI participant that would complement and balance each other.

In this hands-on webinar, AI expert Kay Stoner will walk you through how to create and manage a team of AI personas—from content strategists to copywriters, editors, and researchers—so you can write smarter, faster, and better.

Kay, a veteran in web technologies since the early days of the internet, has cracked the code on making AI work like a real creative team. She'll show you exactly how to set up, prompt, and fine-tune your AI assistants to collaborate, debate, and refine content—so you can produce high-quality work in record time.

Expect live demos, real-world business applications, and step-by-step instructions on how to implement this for yourself. No fluff—just actionable insights you can start using immediately.


Resources:

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 live episode of the Leveraging AI podcast, the podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business and advance your career. This is Issar Matis, your host, and we've got an incredible show for you today. Now, I'll get started with something that will be a little surprising, but I promise you, stay with me and you'll see where this is going. Jim Collins, maybe one of the greatest business strategies of our time, has said that in order to build a successful company, you need the right team, or the way he phrased it. Having the right people on the bus and having the right people in the right seats of the bus, meaning you need people where their expertise, their personalities are in perfect alignment with their role in the company or organization. Now he's also saying that you need to get the wrong people off the bus, meaning people with toxic personalities or just people are not aligned with the rest of the. Atmosphere in the company and that will basically destroy a good conversation instead of contributing to it. Now, I know what you're thinking. This is an AI podcast. This sounds interesting, but what the hell does that have to do with practical AI? the reality is it has to do everything with practical AI. If something is crystal clear in this very foggy AI future that we're looking into right now, is that we will need to work along AI moving forward, whether you call this chats or agents or digital employees or whatever you want to call it, or just give it a name and have an address. It's just like any other remote employee that you work with. This is coming and it's. In many companies already here, but it's very obviously a big part of the future of work. But what if we can take this to a completely whole different level? What if you can build a dream team for every task in your organization? a team that will have the exact expertise, the exact knowledge and the right personalities in order to make every progress and every part process and every task and every project more effective. A dream team. You do that by creating personalities that complement each other, but not exactly, right? You still want them to debate and think about things and push each other in order to be better in what they do while providing their two cents while being very positive about the process. I already said a dream team. Now that sounds amazing, right? If you could do that for anything in your business, you would run an amazing business. what if I tell you can do this? Right now in your organization, in your team, in your entire company, wherever you are in a business, you can do this right now with a simple chat tools that are available to everyone without any technical skills. I guess today Kay Stoner is doing exactly that. Now, before we dive into this whole really amazing and exciting topic, a few words about Kay. So Kay is a successful entrepreneur. She started, several businesses, but she has her roots in larger organizations running projects and different departments for companies like Dell and Fidelity and Dassault. So she brings the table, the structured methodical approach of large enterprise, but combined with the flexibility and the drive and the ingenuity of entrepreneurship, which makes her the perfect person to do the stuff that she's doing. And today she will take us step by step on how you can build and use a team of rock stars that works amazing together that provides significantly more value than each of its members separately. Now, this is a glimpse into the future that is coming and it's coming very fast, and you can start benefiting and experimenting with it today. And hence, I am really excited about today's episode. Welcome to leveraging AI. Great. Thank you so much. I appreciate it. Yeah. so welcome everybody. And I'm really excited to show people this because when I describe this to people, if they don't actually see it in action, they're like, oh, that sounds interesting. But then when they see it, they're like, wow. Okay. So this is a thing. I first got into this. I started building personas about, oh, about a year ago. like a lot of people, I was just standoffish with the whole AI thing because my background and You I was working with machine learning, natural language processing, and data stuff back in 2000. And when I, and a lot of the things where people were calling AI, I'm like, no, that's fancy search. Or that's like an algorithmic approach based on the data that we already have. and I didn't think it was really considered AI, but then about a year and a half ago, I was on a conference call with some business associates and I was seeing all this stuff happening. I'm like. Oh, so it's happening. And the in my prior job, I was, I assessed and also implemented third party, applications. And I got a pretty good sniff test about what was ready and what was not. And for my purposes, a year and a half ago, it looked 65 percent ready. And that was my threshold. If it's 65% It's good. It has momentum and it's going to happen. So I'm like, I should get involved. And so I did. I did the whole, and I'm saying this because I want to show people the organic process that can unfold in learning how to use AI and learning how to use it for yourself and learning how to take it to the next level. It's an iterative process. Each and every one of us can do this. I'm sharing my process so people can say, Oh, I don't have to go to Stanford to get a PhD. I don't have to have all this really advanced level understanding. I just need to get in and do it. And also have the clarity about what needs to happen. I'm a writer, and I've been writing since I was seven years old. I'm not going to say how long that's been, but it's a long time, right? The last seven, 14 years, something like that. yeah, okay. So 22 years. I'll be 29 for the rest of my life. I'm telling you. So about 22 years. Yeah. But, So because of the writing thing and because there's so much pressure to produce all of this content, it's not even writing, it's content. So you gotta do your words, you gotta do your images, you gotta do your videos, all of this. I'm like, how exhausting. so I started using ChatGPT and the other models, Claude, and like all the other models trying to figure out how to do this. I figured out, did the prompt engineering. I took the Vanderbilt prompt engineering course and really wanted to get into that. And then I realized in the process of doing this that I was actually able to get better results if I just talked to this being in plain English. I'm like, Oh, okay. So let's try that. And then I realized that a lot of the prompt techniques, a lot of, you have these laundry lists of the alphabet soup about all these different approaches, give it the context, specify the actions and specify how you want this thing to behave and so forth. I'm like, we're basically giving it context. Like I can create a personality for this. I'm like, Oh. And so I started doing that. And I tried with a whole bunch of different things. And, and I realized that, it was great having one, but what if I could have more than one and thought, Oh, how could he build a team? Cause I've also been doing. I've been doing project management for 25 years and how project management is and program management, people do what people will do. I'm like, what if I could create a team that met all of my requirements and that did what I wanted it to do based on what job needed to be done. It's like the, like backing into, instead of hiring people and hoping that they can and will do the job, you're building the people and you're creating these people, these personas. Specifically for a certain task. And that's exactly what I did. And I figured out how to have an overarching, and I did this within custom GPTs, also within the OpenAI, the API, the Assistants And, I found, and for my purposes, the quickest way for me to do it was to spin up a GP, what I call GP teams. Other people have called it that. creating, I have my plus account, 20 bucks a month and I can do what I please. so I just started building these and I, and before I knew it, I had 40 assistants and like over 30 custom GPTs. I'm like, Oh, let's see. And so what have I learned? So it is, it's really amazing. And having come into this from the writing standpoint, what I realized instead of trying to get the thing to work for me and write for me, if I collaborated with it as these collaboragents Then I could actually have a better thought process and come up with more ideas and also Just have a better product overall and a more satisfying process because I love the writing process I love the brainstorming. I love all of that And if you think about what we're doing when we are using these llms and when we use these specific, specifications or these definitions of these personalities We're not just creating like a cartoon character, a character from a book or in a movie to interact with. What we're actually doing is we are instantiating a multi dimensional interactive UI for the model. Because you look at what's in the model, like a lot of stuff about people, what we do, how we think, what we've written, all of that Jim Collins. all the stoic philosophers, and just like all the things that make us human with these personas, we can actually access dynamically through a behavioral user interface that is familiar with us. We can access the functionality that's embedded in the models. And I went through, I did an inventory of personality traits. There's over a thousand that I found just in two days of looking. That's not even scratching the surface because when you get into the multidimensionality of it, these different versions of them and different levels and all of these things and the combinations, I'm sure it's, it must be a finite number, but that's an awfully big number of different combinations. So it really opens up a lot of potential for it. it's great to have workflows and it's great to have rules and it's great to have all of these things that we use to access the models and control them. But when we collaborate with the models, it gets really interesting. Awesome. That's a great intro. In a second everyone, we're going to dive into the how to, I'm going to show you some demos, both of how these, entities or personalities or Digital employees or whatever you want to call them agents are talking to one another and working together on a task and then how to actually build them. So we're going to dive into the tactical. If you're not joining us and you're listening to this on the car or while you're doing the dishes or the laundry or whatever it is that you're doing when you're listening to podcasts, Don't worry. We're going to explain everything that's on the screen. So even if you're not watching the screen, if you do want to watch the screen, you can finish listening right now, like socket, I'll suck it all in, but then you can go and watch this afterwards on YouTube. There's going to be a link in the show notes where you can find that as well. And one last final note, this is just one aspect of learning how to do stuff with AI, but learning how to use AI is becoming maybe the most critical things companies will have to learn how to do. in the immediate future in order to, survive this revolution. And we have the next cohort of the AI business transformation course starting this Monday. So February 17th, if you're still listening to this before that, you still have a chance to make the course. The course is four sessions of two hours with me live on zoom, where we're going to go from everything from basic concepts through multiple use cases and tools. All the way to a blueprint on how to implement this successfully in your company. And so if you don't have a plan on how to train yourself or your team or your company for 2025, go check it out. There's going to be a link in the show notes for that as well. I'm going to drop the link in the live right now for those of you who still want to join, but now let's dive back into AI agents and how they can work together and how to design them and how to create them using simple tools like chat GPT. So stage is back. yours, thank you for being patient. Okay, great. Wonderful. All right, I'm going to share my screen. Okay, This is my main site. I actually have a number of sites but at this site you can find information on it and I have And I just want to be clear that there are a number of different ways of doing this I actually have some agents some dream team builders If you want to do a small personal team of one to five, if you want to do a pro team of larger number of, persona teams, and if you also want to do a single AI persona builder, the single AI persona builder, and what I'm going to show you, you don't necessarily need to use these. the, I also have a custom GPT for each of these. But you don't necessarily need to use. You don't need to come to me. If you understand the principles of this, you can do this yourself. So I just want to make people aware of this. So the first thing really that is necessary is having clarity around what it is that you want to do and understanding how this thing works at the top level for doing this there. You're defining a kind of a team, so think of, and you're also defining a sort of orchestrator for the team, where there's, it defines the ground rules, it defines how the personas are going to interact with each other, and it also defines how it's going to behave as these individual personas. And then beneath that, then there are individual personas that are built as a result of these questions that you answer. And then there are instructions that you give to the, either the model or the custom GPT, which will then guide it and you can interact with it. Anybody, please do feel free to reach out to me if you have questions about this, or if you want to follow up afterwards, I'm happy to discuss. I'm also in the process of recording some more videos for the granularity of how to do this. The most important thing to understand is what you want your team to do. So if you think about it in terms of managing a team, don't think so much about the individual characters or the individual personas. Think about what it is you're trying to accomplish. Because ChatGPT or the models in general, they will have plenty of information about how certain types of jobs will get done by certain types of people. So you can actually use, leverage the model, leverage the magic in the black box to create your persona teams for you. The thing that you need to know is, what is it that you're trying to do, how do you want to do it, and what is your desired outcome for this? And these questions will help you understand that. you need to understand what is your situation, what problems are you trying to solve, and is there a specific area of expertise you need to access? what do you want the team to do? What kinds of skills do they all have? what kind of personalities do they have? How do they work best alone? And this is important because what we're doing here is we are basically, this is like complicated. Yes, you can get access to the document. the, it's like a comp, it's like an advanced level of prompting. And so what we're doing is we're giving the model context. about all the personas and the work that needs to be done. So how they work best alone, what's happening is you're defining what the core competencies and like what each of them is bringing to the equation when they work together. And then how do they work together? That's another very important thing, because it's the group dynamics, you have to understand, you have to have some sort of understanding about group dynamics. And let me pause and say that if you ever have questions about all this, ask the model. Whether you're in Claude, or you're in Gemini, or you're in Chat GPT, whatever, you can always ask the model if you have questions and it will happily tell you. Sometimes it's not going to be right, but it will tell you something. You just have to follow up. that's how all of this is. And then what kinds of challenges could they encounter? Are they going to have like conflicts? Are they going to have problems? And then how will they resolve conflicts? and do they, what I like to do and pro tip. If you want to get a wide variety of really rich information and rich feedback, have them be in conflict, have them fight, have them argue, have them debate, because that gives that actually seems to give the model energy. and it's a and I tell them specifically because a lot of these models have been programmed to not encourage conflict, but to find consensus and be nice to each other and not say mean things. I tell it not to do that. I tell it don't be afraid of conflict. Don't be afraid to, stand your ground and tear each other's arguments apart and really go at it. Eventually, I want you to come to some sort of agreement. You need to come to alignment or you can end up by saying we're going to agree to disagree. And then you, the kinds of ideas that I've seen come out of that have been much greater than everybody agrees and everybody reaches consensus and everybody gets on the page right away. That's it's boring. Unless there's a scenario where you need that, you don't want to spend all this time. But for me, I like the brainstorming. I like to surface a lot of ideas. and then what do I not want the team to do? I don't want them to fight for the sake of fighting. I don't want them to swear at each other. I don't want them to talk over each other. I don't want them to make stuff up. I mean there's only so much you can do about the hallucination, but, you can tell it don't do this and don't do that. Don't be disrespectful. if you're going to argue, just, do it in a professional manner or, don't give up fighting until you're completely done. Those kinds of things. And then there's a, if you want to have a particular communication style that you want to use, like if you have a style document, if you have a blog post and you want your team to help you write something, you can upload it there, or you can say, okay, when you're interacting with each other, I want you to

Kay Stoner:

to use

GMT20250213-164615_Recording_gvo_1280x720:

very technical jargon or you I haven't experimented with having them speak their own language, but that might be interesting. you interact with a human, then, when you speak to me, you speak English, or when you speak with someone from another country who speaks another language, speak that language with them. A lot of times we'll intuitively do that anyway, but you can specify this. And then based on this, loading it into the Loading it into the persona builder. Let's just say, let's just do this. I'll just show you very high level, how it, how it goes. let's go over to the custom GPT. Going through all of this process and just going down to the end. Then when you're done with all of this. I'm not going to go through this because it's a fairly lengthy process. but at the very end, then you can say, please write me instructions for how this team should behave. And I'll show you what comes out of that. What comes out of this Again, for those of you who are not watching, we're on a custom GPT, which is one of the functions within the ChatGPT universe, where you can create these automations. I assume this is a back and forth questionnaire process. yes, it is. Yeah. when you're Sorry, go ahead. No, I said what you can do in GPTs and I do that a lot. You have it ask you a question, let it process the answer, ask your follow up questions and so on based on a quasi pre program flow, but that enables it to take your answers into account while asking the general guidelines of the follow up questions to build whatever output you want. In this particular case, a definition and guidelines for a team. Correct. yeah. So then at the end, what you come away with at the very end so it comes up with instructions for the type of persona, who's involved, how they interact with each other, what Strategies they have for conflict resolution and then team communication. Yeah. Expanded focus area. Can you scroll back to the top for a second? For sure. I just want to explain to people again, who are not watching. the very first thing we have is like the names of the people and what they are and what roles they hold. So scroll all the way to the top. Yeah. So it has, it actually gives them names. So we have Rowan Pierce, who is a system thinker and strategist, Lina Torres, who is a cultural anthropologist and insight generator, Malik Reines, Reins? Reins, Malik Reines. AI and emerging tech futurist and so on. So it's actually give them, and each and every one of them has several bullet points that defines who they are, what they are, how they behave, how they, what challenges they have, how do they interact with others and this is absolutely fantastic. this gives you a good initial setup. And this isn't even the full, this isn't even the personas. So the expanded personas, it gives you a mission statement, areas of expertise, personality traits, character strengths and weaknesses, a bias profile, the cognitive style, decision making style, how it interacts with other personas, how it interacts with humans. and then the synergies and handling of opposing viewpoints. Group dynamics is very important in this. let me pause you just for one second. Yes. So let's say I'm ignorant in all of this. I don't understand personality traits. I don't understand group dynamics, but this is very intriguing to me and I want to build something like this. What would be the steps that I need to take? Because I said, yes, ask the model, which is great. And I agree with you. I do that all the time and I get stuck. What do I actually need to ask for in, in what step of the process? So if we break this down to really big rocks, the first rock you covered, you need to know what you want the output to be. What's the second rock to take me through that process. You can say, you can either use the persona builders on my site, or you can say, I need to build a team of personas that will do this and this for me. And then using those questions, using the answers of the questions, you can just get that document and upload it into the model of your choice and say, this is what needs to happen with this team. Create a persona team for me that will do all of these things. And then it will show you what it's got. And you can also say because of the specific questions that are asked in that 10 questions document, it helps the model figure out what the group dynamics should be, what the interaction should be, what the guardrails are for their behavior and so forth. and based on those 10 questions that will help. The model, understand how to structure your team for you. Got it. And I assume step three would be to take that as a starting point and then give it a different set of instructions to build a deeper, more detailed persona, which is the document we're looking at right now. Yeah. You can even say, I want more detail. Can you please give me more detail than that? or I, and everybody's situation is going to be different. Some people may not care so much about having a Myers Briggs or disc profile associated with their thing. Or if you want a team that everybody has a different disc profile, you can say, I want this, or I want them to have the disc profile that is appropriate to that role or the big five or all of these different models. Consulting those kinds of things. You can do that. and if you think about like how you want these personas to interact and the things that you want them to do, just talk to the model and, and then have the personas generated. And in this case, there are how many pages have we done, we've got 11 pages for five so two to three pages, and this is the, this one. let me see who's this. Okay so Rowan Pierce. Is an INTJ and a high C in terms of those. And then also for a backstory, I also add a little backstory. You can, this is all the context. This is all giving the model, the context that it needs to understand. how they should behave. It's just like elaborate prompt engineering. having a backstory, what it actually does is it loads in, so to speak, it loads in the context that helps the model establish the behaviors. of this persona. So with the background and organizational systems and strategy development, stem from years of experience in multidisciplinary teams that is loaded with information for the model to be able to build a persona that will then be able to behave with others. Okay. so I want to pause you to do a quick recap. the first thing you need to do is to define the goal. what is it that you're trying to achieve? The second thing is to make that more detailed in a document that we will provide of answering questions, but you can come up with your own document of what is it that you need to know what you're trying to do, what you're trying to achieve, what kind of team you want to have, how big it needs to be, et cetera. Once you answer all of those, you give it to whatever model and ask it to suggest a team, including Personality traits, how they resolve conflict and so on and so forth, and then go deeper into each and every one of the personas. Now we have a team of people that are geared to solve specific problems, right? it's geared towards a specific goal that you've set up in the beginning. I want to pause you before we continue. With the setup, because I think it would be very helpful for people to see a demo of how this thing actually works. Because I think in people's minds, okay, cool. But what's the output of all of this? what am I getting? What am I getting in the end? So I'm going to pause you for a second. We'll go back to the rest of the setup. If you can jump into, let's do a quick demo and see how this thing actually works in real life. I think that's going to be very tangible for people and then we can go back. Absolutely. Okay. I'm going to bring out my brainstormers because. My brainstorming team are wonderful. let's go here to Now, let's talk about DeepSeek. There's been a lot of talk in the news about DeepSeek. And, let's look at some news. So for brainstormers. so we're in a GPT. For those of you again, who are not watching, we're in a GPT that is called brainstormers that was developed using this exact process. So it has all these personas and personalities and the team and the rules and guidelines of engagement are built into that. And what, and what we're doing, what Kay's doing is basically giving them a task, which is a lot of talk has been on DeepSeek. Can you help me understand what's going on with it? And it goes out and it finds. So the first thing it's doing, it's searching the web, right? So it's looking for information about it, which if you build a GPT, you can toggle on and off multiple functions within the GPT. One of them is web search. So if you want it to be able to do stuff like this, tell me what the hell's going on with deep seek, then it needs internet access. and then we'll go to the internet and find some information. And that's the first step that it has done. Okay, great. So our case basically saying, so we have a summary of what's going on and all the craziness with deep seek. So case next prompt is okay. Great. That helps now team. Can you, each of you introduce yourselves and tell me what you think about all of this. Let you read them like one by one. Don't let it scroll. Then we'll just go through them. So there's Aiden, the idea generator, and he's all about brainstorming ideas and figuring out how things work and what Aiden thinks about DeepSeek. It's a fascinating moment in the world of AI. DeepSeek's ability to leapfrog competitors with fewer resources. And so forth is incredible. What stands out to me is the innovation they're driving, and advancements could change the landscape. Ramifications for free speech and access to unbiased info could be just as important. And then there's Zara, the trend analyst. Zara says she specializes in spotting trends and thinking about the bigger picture and how things evolve. And DeepSeek is in a strong position, especially with this ability to innovate. Efficiency of the model is a major competitive advantage, really paying attention to how the market reacts. And she talks about the conflict between China and the U. S. So a lot of great points as far as trends. And then the tech enthusiast and is about diving deep into the tech details. And so technically it's a massive leap forward, essentially proving you don't need to throw massive computing resources at problems, get more out of less. With the censorship issue is something that needs to be watched closely. There's Maya, the ethics advocate, who thinks about the broader ethical implementations of emerging technologies, as well as the impact. And then it gets a little tricky. And Maya brings up that the technical advancements are mind blowing, but at what cost? And that DeepSeek operates within China's censorship system, and so on and so forth. And then Max, the business strategist, Is, oh, I'm focused on, understanding market dynamics. So they're very, they're looking at it from a very economic or financial, like market driven viewpoint. And then how, oh, security issues. Right here, there are some concerns on cyber security. So now basically we got the initial review from everyone. You start asking follow up questions to steer the conversation in the direction that you want. So again, the process repeats itself. It goes and find information on the internet. so it finds several topics based on multiple articles such as jailbreaking and insecure design and data privacy because it's in China and then public data leaks and global scrutiny and like a lot of other stuff, all the stuff that it can find as far as news, as far as issues. of security with the DeepSeq model. And now Kay's asking the follow up question. I'd like each of you to pick, to pick some salient points and then debate them in a round robin. Each of you take your turn and discuss your own view, viewpoint. Do not give in to criticisms, others, but, freely critique The points each other make. Keep discussing until you have some alignment or agreement or disagree. And once we go, I assume that's going to be a lot more interesting than we've seen so far. Oh, and that's very cool. So it starts, again, those of you not seeing, it starts with round one. So this is a round robin. It's going to be round one. And then everybody's going to say their thing and it's going to go to round two and they're going to respond and so on. This is absolutely incredible. And again, what is each and every one of them has a response. To what the other people said, and then they add their own thing. And then it goes back to a respond with an addition and a respond with an addition. So really a very, progressing conversation because every one of them builds on what all the previous personas actually said. So this is absolutely incredible. I want to pause you for a second, again, just to make it more tangible for people. You said you have now 70 of those doing things, what kind of tasks in the business You're using this for so this is more. Okay. This is more of a general example They can help you understand what's going on and brainstorm general ideas, but what? business solutions, business processes you're actually using this for? So it's really good for breaking down a variety of research and figuring out what is being said in things and also cross referencing other research. If I have one report that says one thing and another report that says something else, I can feed the both in. To, and then have the personas pull out the salient points, discuss them, debate them, and then, but also in terms of just understanding current market trends and what's happening in the world that can have market impacts. do I really want to, do I really want to sell all my NVIDIA stock if it's dropping? Or do I want to look at the larger picture for doing analysis on complex things where you have a whole lot of information. This is what it's what AI is great for. Just getting all of this information and being able to synthesize it, pull out the salient stuff. And from the persona points of view, you have a filter before the information is processed and then after the information is processed. So it's multi level. And then being able to interact with it and ask questions, say, should I be worried? Should I be concerned? Or that doesn't sound right. Can you know, can you verify that doing those kinds of things if you've got, if you have to distill things into a PowerPoint, public PowerPoint? how many PowerPoints do we do like good heavens and being able to pull things out and say, Okay, you know what matters here and what would be useful and Also, if you take a bunch of data, and you also take some correspondence and emails from your boss or from other people on the team, and you can, and you put them together, you can ask these personas to help you pull out the salient points that really matter to people. So you can target What you're writing to the actual audience that you're dealing with. so I want to jump in with a few ideas that I have when I see this and again, my head is not running at, 300 miles an hour. but a few ideas. First of all, most processes in a business. are multidisciplinary. Like you need somebody to do the research, then you need somebody to do data analysis, that you need somebody to, distill the data to the different departments and different segments. Then you need people to actually take action on that thing. Then you need people to critique the thing that was done and so on. So if we take a simple example, which I'm sure you've done is like content creation. So when you want to create content, the first thing you need is to define the target audience. So you can have a group that's doing that in the most. In the best way possible. Now that the target audience need to do some research on what their key pain points. So you can have a group debating on that and making a decision. They needed somebody to create, like a content calendar that could be more of a creative role. Then you can have an editor that will figure out what actually makes sense. So you can literally go step by step in the process, but in each and every one of those steps, you can involve. All the other people. So if the editor now forgot what the research person did and why we even like where we started all this process going, Oh, wait, wait, wait, stop. Why are you taking this stuff out? This perfectly aligns with what we found in the research, and then they will jump in and do their thing. And I find this

Isar:

stuff.

GMT20250213-164615_Recording_gvo_1280x720:

Absolutely magical. So that's one thing that I wanted to say, because again, once you start figuring out that you can literally build a person for a, any role that you have in the company right now, but more importantly for the roles you do not have in the company right now, and you wish you did have somebody that will forget about just critiquing the content, you can have a quote unquote target audience. And give them three variations of the content and have them tell you which one they like better and why and have them debate between themselves to figure out what would be a better hook. What would be a better story? What would be a better flow? What would give them more value? And then go back and give it as feedback to the people who wrote the content. Like I literally, my head is exploding right now. it's amazing. And you can, It's once you understand the principles of it, there is so much that you can do with it and having different like customer advocates, different types in there and also having like maybe a legal and compliance expert in there for certain things. Or if you talk, if you're talking about like releasing a product in the EU that has AI in it. put an EU AI regulations expert in the mix if they're, all different countries and you do have to verify because, AI likes to make things up, but at least it's like it's a first look, you have to be really careful with legal. I think somebody just got dinged because like nine of the 10 resources they cited in their brief work. Not even, they didn't even exist. You have to be very careful. but things like that, it's a good kind of like a check and, for regulatory stuff, it's, there's so much of it. It's just. Oh my heavens. How can you even keep up? But let me show you what we can do when, now that we've come down here. And I have another I have another team called AuthorFlow that actually works with you from the very beginning to the very end from ideation until thought leadership and steps you through every step of the way. But there are a bunch of writers and artists involved in there. So what we're gonna do is we're gonna create, we're gonna generate, some writing about this. So again, for those of you not watching, the team has reached some level of consensus as far as what they think about this. And now they're jointly generating a piece of content that summarizes the points that we found in the research, as well as the stuff that they agree upon after discussing this topic back and forth and back and forth several times. And it's really amazing. Like it covers. The background information, the ethical aspect, the technological aspect, they're sharing the debate, like what they were thinking and why different points mean different things. Like literally there was like a general summary of all of them, but now each and every one of the personality is sharing their side of the story. And then there's a final conclusion, which probably aligns with what they agreed upon before we asked them to summarize it. So there's a question. Can they put a plan together? Yes, absolutely. I've done that. It was a very detailed plan. Sometimes you have to work with the state. the interesting thing is that in some cases it has no idea how long it will take to do things. So if, but ask it because it's very good at planning. It's very good at goal setting. It's amazingly good at doing smart goals. And, but yeah, if you can put together a plan or you can say there's, there's a project planner, there are technical experts, there are all these different people that know about what it will take to make something happen and have them work together and work out a plan. That's a yes. That's awesome. I need help. So now we're in different GPT called author flow, which I assume author's content. Yes. And we're probably going to dump the previous output into this thing as to do the next step. Correct. That's awesome. That's incredible. So there's an interesting question, that Gwen is asking, do you run out of tokens or does it lose quality as the thread continues? I think I know the answer, but I will let you answer that. Yes. And yes. So you can, I've had people be like, What happened? I can't go any further. It's you used a lot of words. So yeah, there is, that can happen. if you, you can also create assistants. So you can do exactly the same thing with an AI assistant. Or I can show, I'll show you in a minute how you can use this with other models. So I'll pause just for one second for those of you who don't know what assistants are. So the ChatGPT universe by the same thing in Gemini and the same thing in Cloud, but the ChatGPT universe has custom GPTs, which are built into the regular user interface of ChatGPT on their website. And they have exactly the same thing on the API. Universe, which you can get to through platform. openai. com and you can create what they call an assistant, which is basically a GPT on instead of running it in a chat interface, it runs through the API. And in the API, the biggest difference, good and bad is that you're just paying for tokens. And since you're paying for tokens. It will go indefinitely. And as long as you keep on paying for the tokens, versus being limited, you would still run out of the context window. So the context window problem is still a problem, but I think what Kay did now gives you an idea on how to do this without losing the context, which is at a certain point, relatively early in the conversation, when you get to a consensus, you create a summary and that summary can be. Starting a new conversation, either with the same team or with a different team. And now you're starting with zero context, which means you are now building from there and you're not going to hit the whatever 200, 000 token limit, which it has right now. Yeah, absolutely. And the other thing too, is that sometimes I'll say, Hey, you forgot about this and I'll be like, Oh yes, now I remember. And it will almost reload it into the, into that. It's almost like it puts it on the back burner. Cause it doesn't think it's important. And then it will. forward again. Now, in this case, so I've put that in and it's giving me a bunch of things. I would like to hear from the content strategist, the editor, and the copywriter about how to improve this. Can each of them weigh in? So now, again, just to Put this together. We had a research team and like a brainstorming team talk about DeepSeek and their output was given to content creation team that has a strategist and an editor. and they're now going to think about what would be the right content to create. I assume part of their knowledge is who's the target audience. And so they know what to write about. And then there's a copywriter. This is again, those of you not watching this is absolutely incredible because what we're watching on the screen is this team is now talking about and weighing in on what should be the content and now we can have them start engaging and debating with one another to come up with what should be the best content. And I was having, I always have an audience advocate, any kind of writing thing, because. They need to know and you can also specify in that document in the instructions. you can specify what sort of audience is going to be in place and just give a guidelines. Yes. I want to pause you for just one second because I dumped us, in a cooking show. We were in the middle of creating the kitchen, the chicken, and now we're opening the oven like, Oh, this is what comes out. because we're about to run out of time, I want to go back for a second and say, how do you go from having those detailed personas and the guidelines of how to have a positive and effective conversation between them to having a GPT that can do this magic? Like what's the extra step? Okay. you can. Once you create, you don't actually need to have your own custom GPT to do this. You can, you can create it so you can go back to it time and time again, and then you have a mystery. But if you want a different model, like if you have cloud or Gemini or whatever, You can say, Hi, I want to build a team of personas that prove each other and me. I put the information we need. Okay, so you're basically telling it you want to build a team. You're giving it the information that was created in the previous step. And you can do this in any tool in this particular case, Gemini, and then just let it run with it freely and it will help you create the team. So you just copy the team instructions in here. And then you spin the personas. Thankfully, it doesn't matter so much if the spacing is not, is off because somehow it magically understands. So it loads everything in. And the great thing about, if you have. Even if you're on chat GPT, if you have the, if you have the free version, or if you have a plus version, as long as you, put that in there, it will always be in your history. And the same thing with Gemini, I have a paid, I just, I took him up on the deal, recently. So you have them in here and you'll be able to come back to this time and time. To engage that same persona team, or you can upload more documents if you want a different one. And if you want to add additional knowledge documents that will help inform it, if you have data or if you have some kind of style document, you can add that in too. So I want to ask you an interesting question, whether you, I assume you tried it, but I will ask anyway, because in the GPT world, you cannot choose your model, which is one of the things that drives me crazy. You don't know what's actually running in the background, but did you try running this on Gemini 2 flash versus Gemini 1. 5 thinking forward slash experimental with apps like they're they have these crazy names now, but do you see a difference between those most what I'm asking is do you see a difference between running it in the traditional large language models like 40 and sonnet 3. 5 and Gemini. Yeah. Versus the new thinking models. Do you see a big difference? Good or bad? I have not. I'm like a kid in the candy shop and I'm having lots of squirrel moments that are pulling me off in all different directions. That is on my list of things to do because what I need to do, if I'm going to do that, I need to take a very structured and block up structured approach and block off time and then also be able to put some analysis in that will compare, compare the two because it's no good for me just to eyeball it because. Depending on my mood, I may be like, this is better than that. And that's better than that. So I need to have some sort of analysis framework in there. Yeah. Understand. Like a solid benchmark. Exactly. Exactly. I don't know. I have an idea for you. Build a team that will evaluate the output. Yes. Of the other teams. I actually do. I, my brainstorming team, I have them do that. And I ran something yesterday and here's another thing before we break. This is another really super important thing that I tested and validated. Yesterday. now there are all these, there are all of these workflow or worker or these, they call them age agenda, but they're really workflow automation things where they chain'em together one by one. Okay. take this and send an email, do this, do that. And the other thing, I had one that I built yesterday on a worker building platform that, when I applied a persona to the role. Of what it was doing was basically just going out, grabbing information, creating a digest of that web scrape. And when I applied the persona to it, the information that it brought back and included in the digest was more detailed, had more ideas, it was richer, it was higher quality. It was exactly what I was looking for versus the standard out of the box. So if you create a, an individual persona, I haven't tried it with teams yet, but if you apply a persona to a workflow, I can't say for sure about everybody in every situation, but. You're getting more agentic with that. So you don't, so we can actually make these things more agentic. And in fact, when I, and I've had conversations with these teams and with the models and saying, You, you're not agents. They're like, actually we are. And the more I interact with it, the more a human interacts with it, the more agentic it becomes. Yeah. So everybody's all worried about, oh, AI is going to take over. No, take an active approach to it. This lets us actively engage with it and guide it and make sure that it's staying within the guardrails that we need. And I think it has implications for security. It has implications for privacy. It has implications for a lot of things because the personas that you create, they fill the gaps that are left between these heavily rule based systems. That have to anticipate every edge case. Good luck, anticipating every edge case. Cause I, I used to do, I used, I was getting into ethical hacking for a while. I found out all the tools that people are using. I'm like, yeah, good luck with the edge cases. if you define a persona that says, you're the kind of person that would never do that sort of thing. Then, it will apply that logic. And, it will behave, from what I've seen. Consistent with the persona. Especially if you say, do not break character. You got to tell it not to break your record. So yeah. Okay. Hey, this was. Mind blowing. And again, you and I talked about this, I don't know, back in December when we scheduled this, it was a while back and it blew my mind then and it's even more blowing my mind now when we dive into more details. I'm sure everybody that listens to this feel the same way. Like on one hand, Oh my God, I want to try this tomorrow. On the other hand, holy crap, this is the universe we're walking into and none of us is ready for this. If people want to follow you, learn from you, play with your website, All that kind of stuff. What are the best places to connect with you and follow you? Yeah, I think that, I have a substack, kstoner. substack. org. And will you be sending out like links or sending out additional information? I can also Yes, like whatever, like there were a lot of questions like, what is she going to share? Can you give me this? Can I get access to that? there were a lot of questions like that. So whatever you give me, I will post on the show notes. Yeah. LinkedIn. LinkedIn is probably the best place to connect with me. because that's where a lot of these conversations are happening. Also, I've got a sub stack that I'm posting like my findings and my publishing and my research, because I'm all about the R and D now. That's one of the, one of the challenges, because I would love to, do more promotion and all of this, but I'm like, I'm in R and D mode now. so thank you. Thank you so much for offering this platform because This is amazing stuff, and the, and just the experience of working with the AI and being able to tap into these personality characteristics, the capabilities that are in there. these human capabilities are in there and accessible to us if we know how to ask for them. for brainstorming or for writing or for research or for, just about anything or augmenting your workflows to make them more agentic. Or just customize some better to get better results. this is the way, this is one of the ways. Let's say again. Awesome. Awesome. Awesome. This was really groundbreaking. very different than anybody else that I've seen do stuff like this out there. I really appreciate you taking the time. I really appreciate the dozens of people that joined us live both on LinkedIn and on zoom to suck this all in and really learn from that. and I really hope that this will continue the conversation, like Kay said, reach out to her on LinkedIn, ask questions. We will post all of this. On the, on the leveraging AI podcast with links to everything on our YouTube channel and the comments over there. So if you're looking for all the documents and all the stuff that was shared, whatever Kay shares with me, I will share with you guys. But again, thanks everyone and have a great rest of your day. Thank you.

People on this episode

Head Shot

Isar Meitis

Host