Are you juggling five different AI tools just to get through your workday? Trying to remember which login goes with which platform and wondering if you're actually saving time or just creating new headaches? You're not alone.

Today we've got the AI whisperer himself, Dustin W. Stout. This guy isn't just another tech founder with fancy promises—he's been in the trenches as a digital marketing consultant for over a decade before founding Magai, which he brilliantly designed as "the world's best AI tools in one place." Think of Dustin as the Marie Kondo of AI tools—except instead of asking if your tech sparks joy, he's built a platform that actually brings the joy back to your workflow. He's helped countless businesses stop drowning in subscription chaos and start actually leveraging AI for growth. What would your business look like if you could reclaim all those hours spent switching between tools?

The AI Hat Podcast host Mike Allton asked Dustin W. Stout about:

✨ Consolidation Creates Clarity - Bringing your AI tools under one roof eliminates the mental overhead of managing multiple platforms, allowing you to focus on results rather than juggling logins and interfaces.

✨ Integration Drives ROI - The real value of AI comes not from individual tools, but from how they work together in your business processes to save time and enhance decision-making.

✨ Start Simple, Scale Smart - Begin your AI consolidation journey by identifying your most frequent use cases and finding a platform that handles those well, then expand as your needs and confidence grow.

Learn more about Dustin W. Stout

• Connect with Dustin W. Stout on LinkedIn

Resources & Brands mentioned in this episode

• Magai
• My Podcast Showrunner
• AI Work Buddy - Your Own Personalized AI Assistant
• AI Primer: A Comprehensive Guide
• AI Training for Businesses
• Explore past episodes of the The AI Hat Podcast podcast

Full Transcript

(lightly edited)

One Platform to Rule Them All: The Magai Approach to AI Integration with Dustin W. Stout

[00:00:00] Dustin W. Stout: And this is why also a lot of people struggle to get the AI to, to not sound like Ai. We, we've all heard people say like, ah, that sounds like ai. Well, s are, that person probably didn't, didn't give the AI a role to play before having it generate content. Because when you give it this role, it has significant effect on what it generates and how it generates it.

[00:00:22] So personas are our way of allowing you the ability to customize the Ai. Give it a roll, upload it with whatever instructions that you want it to follow. Mike, you mentioned that you have one that follows a very systematic step-by-step approach, which is really good, very high level thinking for a persona.

[00:00:40] Or it could be just as simple as your, our customer support agent and you help the user in X, Y, Z way. You can create personas for any given scenario your business. So for a salesperson or for to act as your customer support agent, or to act as your graphic designer and give you graphic design ideas, by giving the AI this role, you then get significantly better output.

[00:01:05] Mike Allton: One Platform to Rule Them All: The Magai Approach to AI Integration Mike's got The Hat, the AI kind. Unlocking power in business, you'll find. Cut through the hype, explore the tech. Every week, new ideas to check

[00:01:34] Allton leads us on a digital quest. AI secrets, he'll always invest. Tune in tight, don't miss a beat. Knowledge, groove, keep you on your feet. Listen up! To the AI flow. Business growth, here we go. Cutting edge insights, that's the plan. With The AI Hat you'll understand.

[00:02:18] Greetings program. Welcome back to the AI Hat Podcast, where I selfishly use this time to pick the brains of experts at harnessing the power of AI across all areas of business. From streamlining operations and boosting productivity to gaining a competitive edge and driving explosive growth. Oh, and you get to learn, too.

[00:02:34] Subscribe to be shown how to prepare yourself and your brand for this AI revolution. And come on ahead. Now, are you juggling five different AI tools just to get through your workday. Trying to remember which login goes with platform and wondering if you're actually saving time or just creating new headaches.

[00:02:51] You're not alone. Today we've got the AI Whisperer himself, Dustin w Stout. This guy isn't just another tech founder with Fancy Promises. He's been in the trenches as a digital marketing consultant for over a decade before founding Magai, which he brilliantly designed as the world's best AI tools all in one place.

[00:03:09] Think of Dustin as the Marie Kondo of AI tools, except instead of asking if your tech sparks joy, he's built a platform that actually brings the joy back. To your workflow, you helped countless businesses stop drowning in subscription chaos and start actually leveraging AI for growth. What would your business look like if you could reclaim all those hours spent switching between tools?

[00:03:30] Let's find out. Hey Dustin, welcome to the show.

[00:03:34] Dustin W. Stout: Hey, Mike. Can I just say that I think you're the king of analogies. Like that Marie Kondo analogy was like I. The, that was brilliant. I haven't, haven't been introduced like that before and I loved it because that's really my goal in life is to create products that bring joy.

[00:03:54] And I want to build something that not only makes people's lives easier, but makes it easier in a way that they actually enjoy it. So. That, that was just a brilliant, I was really taken by that and I've told some of your analogies before. I'll have to steal that one as well. So thank you.

[00:04:09] Mike Allton: Well, in, in complete transparency, I didn't write that Claude wrote that.

[00:04:13] I used Claude of course. Via Magai did. Claude is my, well, I, I have a show runner persona within Magai and I'll, I'll link in the show notes to how I do that and why, but I've got a whole showrunner assistant for all of my shows. 'cause most folks listening know, I've got four podcasts at Agorapulse plus this one.

[00:04:32] So it takes me step by step. Who's the guest? Dustin Stout. Give me his LinkedIn. Cool. What does Dustin want to talk about? If he filled out the form and told me, which you didn't, but that's okay. I knew exactly what I wanna talk about with you on this show, but most of my guests, they tell me.

[00:04:47] They wanna talk about email marketing or social media, whatever. And I give it all that. And I've already told it how I want to structure the entire show from the description of the episode to the intro of the guest, which I always say, look, I wanna hype 'em up. I wanna make them feel fantastic. And so I love it when folks compliment me on the intro or specific things like you did.

[00:05:09] And I'll say, Hey. Claude wrote that you're the one who did all the work to become the Marie Kondo of the AI Digital Revolution.

[00:05:18] Dustin W. Stout: Well, I, I definitely appreciate the, the analogy and yeah, I, I suppose I am for any podcast host out there, I am the, the best and the worst guest you could possibly have.

[00:05:29] Because I don't come with demands which means I don't give you any ideas. That's true. Don't ask me for questions. I'll just show up and we'll just have a good time.

[00:05:40] Mike Allton: That's right. That's right. Well, for, for folks who aren't, aren't familiar with Magai, give us the elevator pitch. What exactly is Magai and and what problem were you trying to solve when you created it?

[00:05:52] Dustin W. Stout: Yeah. So when Chachi bt hit the scene, I was in between trying to figure out what my next project was gonna be. I had two SaaS two failed SaaS projects behind me over the last prior two years that just weren't taking off. And I. I was really at a low point in my life just wondering what am I gonna offer to the world?

[00:06:09] I feel like I have nothing left. I've spent the last two years trying to build these two products that I really had passion for, but just didn't have the resources to see all the way to where I really wanted them to be. And I was just at this point where, okay, I guess I have to start taking clients again.

[00:06:23] And took on a few clients and then Chat GPT came along. And it completely upended everything I thought was possible with ai Up to that point. We had some copywriting AI apps, right? We had the, the Jaspers before it was Jasper and we had Copy AI and some of these other template-based tools out there to help us do AI copywriting things.

[00:06:46] But they were all pretty terrible. If we're honest, they were not very good. They were more work than, than they were reward and. I don't know. I just wasn't impressed and, and I hated using them, but ChatGPT was something different. I didn't have to go through a templated system and fit my ideas into somebody else's box.

[00:07:06] I could just tell the AI exactly what I wanted and it gave it to me. That was magical to me. And as I started using it for both myself, my personal brand, my business, and my clients, very quickly things started to pile up. I'm like, why doesn't it do this? And why is this so ugly, and why is this not fun to use and how can I make this easier to organize my chats and how can I make it easier to work with my different clients and separate things out?

[00:07:30] And so all these quality of life things started to pop up, and I wanted to have an interface where I could leverage Ai in all its capacities, but also stay organized and, and in a, in a way that doesn't require me to switch tabs every five seconds. And on top of that, there were all these other tools popping up, Gemini well before it was called Gemini, it was called Palm.

[00:07:55] There, there early ai generative model. And then there was Claude by Anthropic popping up. And I'm seeing all these AI tools and these image generators and these video generators starting to get rumors and audio generators. I'm like. I have to have 15 different subscriptions just to get all the AI things.

[00:08:11] And that's not something that a bi busy business owner needs, right? That's not something that a busy consultant needs. They don't need more subscriptions, more cost. They need a centralized location that will leverage all of the great AI things in a way that is useful and helpful, organized. And so I just scratching my own itch was like, I think I could build a better version of Chat GPT.

[00:08:37] And leverage some other tools into it as well that that Chat GPT will never have because they're a single company and they're only working with their own products and so. Long story short, I scratched my own itch. I built the first version of Magai about eight weeks, start to finish, and was really happy with it, was really excited, decided to share it with the world, and lo and behold, the world actually liked it.

[00:09:01] And so I kept building and making it better and better. And here we are today 30 different AI tools. A whole lot of organizational tools to help teams work together and collaborate with all these AI tools in one place. Make it easy to find very sophisticated layers to how you can integrate 'em, and it's really become much more than just a better, prettier version of Chat GBT.

[00:09:25] Mike Allton: For sure. Now, as someone who's known you for, man, I'd say 15 years I've, I've, I've watched your journey through several different businesses, felt for you the past few years, excited for you over the past year to see what you're doing with Magai, so that's fantastic. Now, you, you talked about how there's, what did you say, 50 different AI tools within Magai.

[00:09:46] There's definitely more

[00:09:47] Dustin W. Stout: 30. I haven't counted 30 recently, and there's like four more. I know there's a lot I need to add today. So,

[00:09:54] Mike Allton: so there's a lot of tools in there, but a, a moment ago you also talked about the fact that we don't need more tools as business owners we're already overwhelmed with the sheer number of tools and subscriptions.

[00:10:03] Talk to me a little bit more about how Magai consolidated approach helps solve that subscription fatigue that a lot of business leaders and marketers are feeling right now.

[00:10:13] Dustin W. Stout: Yeah, well, number one, it's, it's the pure financials of it. The, the fact that you would otherwise have to have a subscription to Chat GPT plus a subscription to Claude Pro, a subscription to Gemini Pro, a subscription to well, I guess you get DALL-E as part of your chat, GPT Pro, but you would need a subscription to maybe another image generator like a Midjourney or like a Leonardo, or like an ideogram.

[00:10:36] And then you need video generators, right? You need a subscription to Runway. You need a subscription to Kling and all these video generators out there. And now. While that might be cool if, if the finances doesn't really bother you that much. Here's what bothers me more than the actual finances. It's the fragmentation of the workflows.

[00:10:57] So if I want to generate copy for a blog post, I go to this one and maybe play it against this other one, and I have to switch back and forth between tabs and log into different accounts. And then if I want to image generated, I have to go over to this other app. And now I've got three different apps.

[00:11:13] With three pieces to the singular piece of content that I'm creating, and I have to somehow wrangle them all into one centralized place by myself. I have to come up with my own system, keeping track of them. I need to, find a way to, to, to store them in a way that I can find them later. So it's this fragmentation to me is is is harder to deal with than the, the pure finances of it. All in all, I think I was paying for, I don't know, eight different AI apps before Magai was fully formed and it was maybe costing me a hundred, $120 a month. Not, not a huge deal, but the fact that I had to spend hours a week just organizing and, and intermixing and, and trying to duct tape things together was costing me in the thousands of dollars.

[00:12:05] So for me, it's, it's, it's both the, the cost savings that's makes it nice, right? To, instead of paying $230 a month, it's just your same $20 a month that you'd be paying to ChatGPT, but then having it all centralized. All your images, all your text generations all in one place easily, organizable easily, easily accessible, and easily collaborated.

[00:12:30] Collaborat able, is that a word? I don't think that's a word.

[00:12:33] Mike Allton: It is now

[00:12:34] Dustin W. Stout: easy to collaborate with, with your team members because you can invite team members into your Magai workspace as well. That means everybody has access and it's just easier to to streamline all the workflows that you're trying to execute on with this content.

[00:12:49] So it's, you're not generating content for the sake of content. You're usually trying to do something. Well with it, you're trying to piece it together and put it to a, a, whether it's a blog post or a social media campaign or something. So centralization really, really helps with the, just getting stuff done faster and more efficiently.

[00:13:07] So I think that is to me the, the most important piece.

[00:13:13] Mike Allton: One of the really subtle benefits that I've discovered, I was doing a a AI bootcamp training and there's, I dunno, 20, 30 people in the room from, mostly the Pennsylvania area, but we had some folks from further out and they were all different kinds of businesses, all different roles.

[00:13:27] So it wasn't just marketers, there's hr, there were legal in the room and so on. And the question that kept popping up was, which. LLM am I supposed to use? At what time? Which image generator should I use? Like you talked about video generation and then you dropped three tools that folks listening may have never even heard of before.

[00:13:47] Which one are they supposed to use? And one of the things that I loved, and I showed them this in the training, was that if you've got Magai, well, you can just test each one. Over and over and over again, right? Like if you're not sure which image generator is gonna give you the kind of image you're looking for, for a particular use case, put in your prompt.

[00:14:05] Pick an image generator model. Get the output and then change the model and get another output, and change the model and get an output. I showed them where I'd done that like 20 different times for one image, just to see and play around a little bit and see which models gave me, the kind of output I was looking for and, and I should stop for a second and just transparently say.

[00:14:25] This is not a sponsored episode. Dustin's not paying me to come on here. I've been using Magai for over a year, and like I mentioned, Dustin and I go way back. I wanted to take this opportunity to get the man to come on here and, and talk to us more about what Magai does than what I typically mention in some of these shows.

[00:14:41] Longtime listeners will know. I usually have video spots in here for Magai. I'm an affiliate of Magai, but I really just, I honestly love the tool. I'm in it every single day for Agorapulse for my own work. But walk us through, if you would, kind of the, the typical use case for a small business owner who might be using Magai.

[00:15:01] What might their workflow look like? Maybe like before Magai and then versus like an after implementation.

[00:15:08] Dustin W. Stout: Sure. Well, let me be a jerk and, and put a pin in that and we'll come back to it because you said something that I think is extremely important. You touched on probably the one question that I get asked several times a day, every single day.

[00:15:23] Over and over and over again. And it's this question of how do I know which one to choose? And I don't think it can be overstated that having them all in one place so that you can experiment and test is really a, a superhuman unlock because otherwise you're having to log into these different things.

[00:15:45] Different interfaces again, switch tabs, have an account here, have an account there, which for, all the IT people in the room, that's a nightmare, right? Having to manage employee accounts that are spread across 12 different apps, multiply that times 50 employees, the IT people in the room, or splitting migraines.

[00:16:05] So having them in one place, being able to. Just play them off of one another or just see how one responds differently to the same exact prompt is super powerful. And the reason for that is because the question on everybody's mind is, which one's the best? And you cannot tell which one's the best. I can sit here and I can tell you I.

[00:16:24] What my tests have revealed and what the strengths and weaknesses of different LLMs versus others, and the strengths and weaknesses of different ai image generators or video generators. I can go through all that, but at the end of the day, what, what it really comes down to is these models are changing so quickly and each one has different types of things that it's good at.

[00:16:46] And the way that it responds to my query about my type of copywriting task. Might be best executed by Claude because I've tried it against GPT and Claude and Gemini and all the others, and I just feel like Claude does it the best. But you might put the same thing into Gemini and it might speak to your language, your the way that you build things much better.

[00:17:10] So it really is a matter of playing and experimenting and figuring out your tasks and which AI seems to fit best. Your outcomes and that that's something that I really encourage and it's, it's the reason why schools like the University of Notre Dame love Magai. They use it in class. So they actually have a generative AI program.

[00:17:30] Many schools are starting to do this now. They have a generative AI program and the, the teacher, John Baron, professor John Barons, what he does is he allows the students in class to explore these different models and he's teaching them prompting and he's teaching them, all the fundamentals of working with ai.

[00:17:46] And you can't do that. You can't experiment to that degree as easily anywhere else. You have to have it in one place to be able to go back and forth and, and analyze and, and, and do it all in the same interface. And furthermore, when you start to understand the different aspects of this, AI is better for this, and this one's better for this, you can actually play them against each other.

[00:18:10] So, Magai, you could easily start with Claude. Switch over to GPT, do a few things. Switch over to Gemini, have it take over for a while and it re, it continues the continuity of the conversation as if they were all in the room together already collaborating. And that way you get the strengths of every single model and none of the weaknesses.

[00:18:30] And so there's really a lot to that. And for everybody who's listening to this, you, you have undoubtedly at some point said, I don't know which model to choose which one is best. Well. Go experiment and find a way to experiment them, and you will absolutely find the one that that works best for your given outcomes.

[00:18:48] That was your actual question.

[00:18:50] Mike Allton: Well, I No, no, I love, I love that point because I know you just posted in your community a, a a a post about this very topic, and I'll enter that in the show notes. Dustin's got a fantastic community for Magai users. Yeah. And the other thing that I talked about a lot is this fact of having, I think everyone should have an AI assistant. Not a persona or a custom GPT that's for a specific task, but one that has been trained on what you do, who you are, what your goals are, your challenges. So it's that AI Work Buddy that I may have mentioned on a previous show, and I'll link to that also in the show notes. My AI work buddy, is my most important persona inside of Magai.

[00:19:29] And to your point, I love how I can. Pick that as my persona, so that's who I'm talking to no matter what. But I can start with Gemini mm-hmm. As the LLM, to help me think strategically about something like a marketing campaign. But then when it gets to the point where we're actually gonna generate some copy switch to Claude, same chat.

[00:19:47] Mm-hmm. So I don't have to give Claude all new context. It's right there in the chat. Mm-hmm. But I can switch to Claude, which most people generally would accept is the better writer mm-hmm. For creative writing. So, absolutely love that. Yeah. So the question was though, help the small business leaders who are listening understand how to inject this into their, their day, their workflows in, in some way.

[00:20:11] How would they use Magai?

[00:20:12] Dustin W. Stout: Yeah. So the first thing you wanna start with is, is understanding. You want it to under, you want any AI that you work with to understand your business from a fundamental level, right? Without the proper context. Whether that's, you, the type of business category, product services, name of your business, location of your business, target audience, all of this core fundamental information, brand voice.

[00:20:37] Your, your messaging that you, is ideal, your ideal messaging, all of this needs to be passed over to the ai if you have it, generate anything for you, right? It needs to know this vital information. And so, step one, whenever we walk an agency or, consultant or enterprise into using Magai for the first time, the first thing we introduce them to is the concept of workspaces.

[00:21:00] And a workspace essentially represents, a, a silo. Where all of your chats, all of your images, everything you generate is going to be contained. And if ever you want to create another silo where that is separate, where other chats and other generated items are contained separately, you can create another workspace.

[00:21:20] And so this allows you to create, for co consultants, like each client has their own workspace, or for enterprises, each department might have its own workspace. That way the work isn't intermingled and you can keep things organized from the top down. So a workspace allows you to have what's called custom context.

[00:21:39] And anybody's used chat, GPT. Probably knows this concept where you're giving context that is fundamental foundational that you want the AI to know every time you start a conversation with it. And so within a workspace, you can create your custom context so that every time, whether you or one of your team members creates a chat inside of that workspace, the AI will already be preempted given that information.

[00:22:05] During the conversation, and it will always be front and center in the AI's memory. We'll never, go outside the AI's memory. You can also in a workspace upload knowledge. So upload PDFs, documents, things that are vital to your your company ethos or proprietary data workflows, standard operating procedures, stuff that, again, you want the AI to always know, but it may not be necessary in every chat.

[00:22:27] So you do it in a file and upload that file so that when necessary. The AI can go, oh he asked me for this. And so that requires this file. And that way you have very high level customized information and data given to the AI in every conversation. So that's, that's step one. Prep your workspace, make sure the AI knows everything that you need it to know in every given chat.

[00:22:52] Now furthermore, we offer, we've had this from day one, even before GPTs were a thing. We've had what we call personas. And a persona is essentially just a assigning the AI a role and a lot of prompting his, his. Been prompting, guides and so forth have been built around this concept that when you assign the AI a role, when you tell it you are an expert copywriter, for example, the copywriting that it comes up with is actually way better because instead of having the AI start as a generalist and just pulling whatever it generates from its massive swath of information and making all these calculations in a generalized manner by saying you are an expert copywriter and narrowing its focus, you are getting much more specific output and much more higher quality output because it's not diluted.

[00:23:44] And this is why also a lot of people struggle to get the AI to, to not sound like Ai. We, we've all heard people say like, oh, that sounds like Ai. Well, the odds are that person probably didn't, didn't give the AI a role to play before having it generate content, because when you give it this role, it has significant effect on what it generates and how it generates it.

[00:24:06] So personas are our way of allowing you the ability to customize the Ai, give it a roll upload it with whatever instructions that you want it to follow. Mike, you mentioned that you have one that follows a very systematic step-by-step approach, which is really good, very high level thinking for a persona.

[00:24:25] Or it could be just as simple as, your our customer support agent and you help the user in x, y, Z way. You can create personas for any given scenario. In your business. So for a sales person or for to act as your customer support agent, or to act as your graphic designer and give you graphic design ideas, by giving the AI this role, you then get significantly better output.

[00:24:50] So in Magai, we allow you to create as many personas as you need to fulfill your different roles and different outcomes that you're looking for. Just like you can create custom GPTs in Chat GPT. Those live within a workspace. And so you will then have on the, for very first layer of the intelligence that you're giving the Ai, you have any context that you've given at the workspace, and then you activate a persona within the chat, which then gives it even, more specific, instruction as to you know what it's supposed to do in that chat. And you can even upload custom knowledge, just like in the workspace, custom files or documents that that persona alone has access to. And so again, just really refining the, the level of intelligence and specificity for every single conversation or every single outcome that you might have.

[00:25:39] One, one of the most popular uses of a persona is a brand voice. So you may not need every single chat to, to have your brand voice or your writing style in it. But sometimes you're writing marketing copy or you're writing website copy, or you're writing blog posts, and you do need that brand voice in there.

[00:25:56] So you can create a brand voice persona by simply telling the Ai, Hey, you are our brand voice writer, and this is our brand voice. And just give it all the juicy descriptions of what your brand voice is and what your. What your writing style is or what have you, and then activate that, that persona only when you need it.

[00:26:15] And just like with the models, you have the ability to play these personas one after the other too. So you can start with your high level strategist persona, get all the strategy laid down and, and all of that. And then switch over to your, say, social media. Persona that is that is instructed on how your social media posts should be.

[00:26:34] So a, a, a lot of really intelligent focusing and, and, and ability to hone in on very specific outcomes.

[00:26:45] Mike Allton: Got it. Now, there was a conversation in the community recently where someone had uploaded multiple files mm-hmm. To their persona, and, and they were struggling with that. And, and you'd made a comment, and I'm probably gonna get it slightly wrong, so feel free to correct me, but it was something along the lines of, it's only gonna read one file at a time.

[00:27:02] Mm-hmm. So walk me through them. What's the best practice? If I want to have a brand voice document, a target audience document some other kinds of. Really smart, good business documents attached. What's the best practice or, or use case for instructing the persona or the ai, when to use those files?

[00:27:21] Dustin W. Stout: Yeah, so this is actually a limitation of how AI actually works.

[00:27:25] And, and no normal human being is expected to understand this or know it outright, which is why it's our job as product creators to, to educate. So there's, there's two methods that you can fill the AI with. With, with data or understanding or knowledge, right? The first way is in the instruction or in the custom context.

[00:27:46] Now, for all intents and purposes, instructions, custom context and persona instructions are ingested. The same way for an ai, it's just a bunch of text that gets in inserted before the chat and is always present. We just can't see it, but it's always there, so it's just a bunch of texts.

[00:28:04] Uploaded files uses a technology called RAG or retrieval augmented generation, which essentially means you've uploaded a file or a multiple files, and these files are turned into stored in a what's called a vector database. And without boring you to death on what a vector database is, basically turns all those words into math.

[00:28:27] Excuse me, but it assigns each of those files a designation and assigns a bunch of mathematical equations to help the AI understand when it needs to retrieve information specific to a file. It knows what file to go to and it. Retrieve that information. Hence retrieval augmented generation. It's augmenting its generation with the retrieval of very specific data.

[00:28:51] Now, again, just like you said, the AI only has the capability to parse one document at a time. So in the AI's mind's eye, what it sees is I'm a. Ai, I have this persona. I've been given these instructions. I've also been given access to this list of files, and it sees essentially the file names.

[00:29:14] And when you make a request, one of the things it has to make a decis decision about is, does the users request require me to reference one of these files? If so, which of these files do I need to retrieve and parse to extract whatever I need to reference? And in the, in the age of current AI and current AI models it doesn't have the capacity.

[00:29:40] To do multiple takes. Just doesn't have that yet. Now we're working, we're in a world where age agentic AI models are a, a reality and they're getting very good. But until we have that, until, agent can come in, an agent is an AI model that can perform multiple steps. One after the other and decide whe whether or not it needs to perform another step.

[00:30:04] But, but as of now, most models are not age agentic. And so they can only do this one thing. And so you have to be very specific in how you structure this, this data that you're giving to the ai. Think through this scenarios yourself. You know what, what items in this data that I wanna upload should be.

[00:30:24] Combined to create one large document or which, which of these documents should be separated? What, what do I not need contained in one file? And, and really the, the, the real art and mastery of working with AI and, and knowledge is understanding how much context is enough context, because you never wanna overload.

[00:30:49] The context of an AI model. The reason for that is because you give it too much to think about. It doesn't come up with as good a results, and it's diluting its ability to focus on a given problem. For example, the, the person you were talking about, she's a, a novel writer, she's writing stories.

[00:31:06] And so this is a very difficult challenge for her because as a writer you wanna have all of the information of the story. Present at any given time as you're building these scenes and building these characters and the character arcs. So it's a very challenging thing, but the truth is, if you're writing a scene for Frodo and Lord of the Rings.

[00:31:28] You don't necessarily need, if you had an AI writing assistant to help you write Lord of the Rings or Yeah. Lord of the Rings. Sorry, I thought I got the book wrong for a second. You don't necessarily need, the character profile of an ORC over in this early scene to inform how Frodo responds to, seeing an elf for the first time.

[00:31:52] At the Elf. Layer, I forget what it was called. But you, you can start to piece together very quickly. Okay. I understand. Like you don't wanna give it too much context because you don't want to give it irrelevant context because the irrelevant context will color any context you give.

[00:32:08] It will color the output. And so you want only the most specific context necessary to get the job done. And that can be a, just a learning process and, and it's gonna be a learning process until, we get quantum computers who can just make all the decisions for us.

[00:32:26] Mike Allton: Yeah, so this just really unlocked something for me.

[00:32:29] 'cause like I mentioned before, I've got an AI chief of staff. It's a persona. I probably have seven files attached to a persona. There's a voice there's a target audience. I a whole research and created a document targeting my audience which for the AI hat. Well, obviously I'll, I'll, consult anybody in the world.

[00:32:47] My focus is specifically St. Louis businesses and so I did a ton of research on St. Louis businesses and documented what their particular business needs and pains are and how AI can, can solve for those. And so my AI assistant, my AI Chief of Staff understands that I'm targeting St. Louis businesses. And so when it and I are brainstorming content to write, it's naturally pushing me towards St.

[00:33:12] Louis businesses. Mm-hmm. And. I was assuming all this time that when I'm strategizing and then going to the actual creation of content, it has all the information in all these files stored in its current memory at once, and evidently it doesn't. And since I don't have an operator yet who could do all these steps at once, I need to walk it through the steps of let's strategize who we're talking to.

[00:33:37] Then let's look at the document that has that target audiences. Pain points and their solutions to figure out what we're gonna write about. And then let's move on to the next file, which has my voice to actually create the content. And so I would either have to specify those files mm-hmm. In the prompt as I step through that, or I guess at least have it stated in the instructions whenever we're going to create content use the voice PDF file or something like that? Yes. Is that right? Absolutely. Okay.

[00:34:09] Dustin W. Stout: That is exactly how you should approach it and, and you should of course, definitely put that inside the persona instructions. One great thing to do is to tell the AI in the instructions you have access to these files, and this file is for this, this file is for that.

[00:34:25] This file is for this. And so just describe those scenarios where the AI would need to reference. Mm-hmm. That file because again, like most people make the, they make the assumption that by uploading these files, and this is just because there are a lot of AI experts out there who they don't work technically with the models.

[00:34:44] They just have a knack for in having intuition about how things work. But they may not fully understand. So there's this misconception that uploaded files are training the ai and that word is tossed around quite a lot because it's, it's the word that just makes the most sense to us. And so I don't fault anyone for doing it, but from a technical standpoint, it is not training.

[00:35:07] And, and just as I outlined before, it is just a retrieval process. You're giving the ai think of it like a human sitting, sitting in an office. And that human has a bunch of different file folders on their desk, and these file folders are labeled different things. And when you give that person a task, the only thing that they're capable of doing and go is going.

[00:35:29] All right. Which one of these folders do I need? I think this one. And that's it. And then they're gonna respond based on that one pulling of that folder. So, that's how it works. And so, there is a way to train the AI on your data and on your knowledge, and then have it become a part of how it just generates content all the time, but that's a very expensive and intensive process and, and not accessible to, to most users.

[00:35:57] Mike Allton: Right, right. Folks, we're talking with Dustin Stout about how a platform like Magai can help you leverage and make sense of ai, and I've got several more questions you're not gonna wanna miss, but first let me share with you how I'm using Magai every day.

[00:36:09] This episode of AI for Business Leaders, The AI Hat Podcast is brought to you by Magai. Tired of juggling multiple AI subscriptions and trying to figure out which tool to use and when? Magai simplifies AI for businesses by offering access to all the leading language models and image and video generators for one low price.

[00:36:30] With Magai, you can create custom instructions to personalize AI for every business need, and use them across all LLMs, including Chat GPT, Gemini, Claude, and even Deep Seek. No more building custom GPTs or gems that are limited to just one platform. Magai also lets you create customized workspaces for your entire team, giving everyone in your organization access to the same AI assistance and tools and prompts ready to streamline your AI workflow and empower your team.

[00:36:59] Tap the link in the show notes to visit Magai today to learn more, and now back to the rest of today's episode.

[00:37:07] So Dustin, for for business owners who might not be very tech savvy, what's the learning curve like within Magai? How quickly would somebody get up and running?

[00:37:17] Dustin W. Stout: Well see, that's where my heart is. When I build products, I don't build them for developers. I don't build them for highly technical people who could probably go out and build something similar to themselves.

[00:37:27] I build them for real people doing real work in, in everyday roles. So Magai is designed to be just as simple as signing up, clicking the new chat button and. Talking to the ai, you don't have to go through the, the process of up of creating a workspace that has custom knowledge and uploading files. You don't have to create a persona to get started.

[00:37:51] You can literally sign up, press the new chat button and start talking to any AI model that you want. And, and that's the heart of everything we build. That's the heart of how we build and what decisions we make on what to build next. It's always with this idea of can a person who knows absolutely nothing about AI get started immediately and start working?

[00:38:14] But also at the same time, we wanna have those hidden little sophistication when and where it's needed so that power users, like you can go in and, customize to the high heavens.

[00:38:28] Mike Allton: And and one thing I've known about you and always appreciated for years is your extremely sophisticated design sense and that that really comes through with Magai.

[00:38:37] It's honestly a very easy to use a very intuitive program. It's a very beautiful looking interface, unlike many the other. Tools, not necessarily the lms. They're, they're, they're pretty good open. Ai Gemini is gorgeous, but some of the other AI startups that are very focused on a very specific kind of task, I get into some of these tools and I'm like, oh my God, who designed this?

[00:38:56] Dustin W. Stout: Yeah, they make your eyes bleed because

[00:38:57] Mike Allton: it's awful. Yeah, it's terrible. So, but speaking of some of these, these, these all, all-in-one platforms. Oftentimes, I think like they'll sacrifice depth for breadth. How do you ensure Magai provides, powerful, comprehensive capabilities rather than just surface level functionality?

[00:39:16] Dustin W. Stout: I love that question so much because I, through and through, I'm a pragmatic person and, and to a fault. I am not impressed with the shiniest, newest, most mind blowing AI feature out there. I want to see how it's used in everyday work in life and, and I will not even, I. Entertain a new feature or capability of an AI model until I've seen very clear pragmatic uses for the average person in whether in their day-to-day work or in their day-to-day life.

[00:39:53] I'll give you a great example. There's a lot of buzz around video generation and specifically cloning yourself. Giving, getting an AI video avatar of yourself, and then training a model on your voice so you can potentially create an entire AI video version of you that will talk and, and do whatever you say just by typing in text.

[00:40:18] This technology's fascinating and is, it's just mind blowing that it's even possible, and some of it is very, very good, but it's not quite good enough. For me, I look at every single one I've ever seen, every YouTube channel that is now using these AI avatars, every TikTok account, and I go, yeah, I can tell.

[00:40:40] And it's pretty cringe. Like, I wouldn't ever do that or want myself represented in that way. And I can't see any businesses that would. Wanna risk their personal brand in that way, or, or their business brand in that way. So I scrutinize every new thing, even with chat, g Chat, GPT tasks is another one.

[00:41:01] So when this came out, it was like, people, whenever, okay, between me and you, Mike. Don't tell anybody this, but man, it gets on my nerves. How? How the open AI fanfare just no matter what they do. It is like, yeah, you get like 15 different people on X that are like OpenAI just released a game changing new feature.

[00:41:25] Here are seven seven magical examples. You won't believe it. And ChatGPTs tasks. For me were one of those things. It was like, oh, you can schedule a push notification. Okay. Anybody else need more push notifications in their life? Anybody? There is some, it is not completely ridiculous, right?

[00:41:48] Because you, you're basically scheduling a prompt. Right. Yeah. That's all it is. It's scheduling a prompt and the, the app will either email you when it happens or it will send you a push, push notification, and some people find that easier to use than their reminders app. I have five different to-do and project management management apps, which will send me same kind of push notifications, but they won't execute a prompt, which to some people has become very useful.

[00:42:19] As I've surveyed my users and the people that I know get the most out of Magai, I haven't found a really great example of the utility that that brings the pragmatic day-to-day thing that actually is worth investing in. I am by default, practical, and pragmatic, and I will not add a new feature or a new shiny object to Magai unless I see very clear business and personal value out of it.

[00:42:52] Mike Allton: That makes a lot of sense. The example that I keep seeing over and over again with tasks is, oh, I programmed it to send me the latest news about ai. Like, so we've had Google alerts for, I don't even know how long, I didn't need to pay 20 bucks a month to have that sent to me. And, and yeah, it, to your point, like deep research, I get that it's pretty good.

[00:43:15] Mm-hmm. But. It's also extremely limited. I have a, a Google AI account. I was already paying Google 20 bucks a month for storage and I'm like, why? I just switched to Google ai. Now I've got Gemini with notebook, LLM and, and deep research, which I absolutely adore, and I used Gemini Deep research probably twice a day minimum.

[00:43:40] It's just, it's so good. It's so easy, and it's. G giving me fantastic information. Now, sometimes, it, sometimes I'm not doing deep, deep research on something. It, it might be as simple as, help me understand how to repair my vacuum, which I had to do recently. And so it helped me figure that out without me having to sort through a bunch of like Reddit threads and other people's random questions and even more random advice.

[00:44:05] But I appreciate that you're looking into those and, and carefully weighing what to add, what not. You've really impressed me with the models that you've brought on. Not just, the most famous ones and, and the biggest ones, but the breadth of models. So people can try laa if they want to, or, gosh, what's Amazon's Nema something?

[00:44:25] Dustin W. Stout: Oh Novo.

[00:44:27] Mike Allton: Novo. Yeah. So you can Novo track these models. Novo

[00:44:29] Dustin W. Stout: Macro and Novo Pro. Wonderful scheme. Yeah.

[00:44:34] Mike Allton: Or Mytral, which I don't think doesn't give enough attention. Mytral seems like a really nice model. No, not nearly. They're doing great things. Yeah.

[00:44:41] Dustin W. Stout: And if I can speak to that, something you said too.

[00:44:43] Yeah. 'cause it, it just dives deeper into this philosophy of number one, like, how do I know the right model to use? But number two, like shiny new things don't necessarily mean. You should be using that. So one of the conversations we recently had and Magai that I wanted to educate people on is this idea of, you don't, you don't use a jackhammer to punch a nail into a board, right. You, you need the right tool for the right job. And so when deep research came out, this was, everybody went crazy. 'cause the, it's really phenomenal. They can go out, gather a bunch of different information all on its own. And it's not just doing one web search, it's doing several iterations of searches as it feels it needs to, and it's gathering, sources from.

[00:45:31] 2030, I've seen up to 60 different websites and then it puts it all in a nice little, basically a, a report for you and organizes all that information that it found, and it is really useful for some things. Now, problem with that, that I saw immediately was number one, AI can. Make stuff up, right? So if it's pulling factual information, number one, you're gonna have to fact check it anyways.

[00:45:56] And so that seems like more work for me. Number two, you don't, AI doesn't have this ability that we have to sort the good information from the bad information. Like you and I, we can visit a few different Google search results and go, oh, this site's not trustworthy. Like, this is, this looks super shady, but somehow they managed to game the, the search results, right?

[00:46:17] AI doesn't have that ability to sort through good and bad information. It's all just information, and so it's doing exactly what a human would do. It's taking the top results from all these different search queries, and it's just putting all that information and treating it all equally. Which is very dangerous for, for some types of research, right?

[00:46:36] So again, you're gonna have to do more fact checking for this. And so. For me, the utility of it, the power of it is very promising, but the reality and the, the practical application of it really falls short in most cases. Unless it's a very low stakes type of research that you're doing, then I guess it's, it's.

[00:46:54] Pretty good. Like I've used it several times for myself just doing market research. Like market research I think is a fairly good type of deep, deep research thing to do. Competitive landscape type stuff where you're just gathering information on other companies. So that kind of thing is fairly safe, again, you have to understand, have.

[00:47:11] A level of understanding about how AI models work and what they can and can't do, and what their limitations are in order to make the best decision about this. And so in most cases, I would say just a simple web search function. That AI has had for months over a year now we, we've got it built into Magai.

[00:47:29] You can have any of these LLMs perform a web search before it generates, and you'll get up to five, or I think sometimes 10 results depending. And that will be sufficient and you won't need to have, 60 different resources to fact check. And then you won't have like a 2000 word article. To sift through, right?

[00:47:48] You'll have those results and you'll have the generated outcome and it's much more efficient. Not only is it much more efficient for you to audit what it comes up with, but it's more efficient from a power consumption standpoint. Like the thing with these these newer models, the thinking models in particular, they are very energy.

[00:48:11] Intensive to perform this high level thinking that, that they do like the O one model or the NewLaw 3.7 sonnet, Gemini's thinking models. They are consuming massive amounts of power to process all of this, and I imagine how much power is consumed by this deep research functionality. This is also why they're limiting it even to the highest tier users of Chat GPT.

[00:48:36] They're limiting it because it's such an resource intensive task and understandably so, right? It's doing who knows how many websites, searches, and scraping that content, processing it and scraping more and processing more. It's a lot of energy that's being consumed, and so I think we as consumers of this technology also need to be more cognizant of the effect of using these high powered functions.

[00:49:03] When we don't necessarily need to, right? We're not, nobody's filling their water glass with a fire hose, right? You just don't need that level of power sometimes. And so, I, I encourage people to, to make decisions on what features and functions they use, based on what they need to accomplish and ask themselves, do I need the most powerful, most intensive, most energy consuming model to.

[00:49:29] Write a summary of this YouTube video. No. You don't need a thinking model to write a summary of a YouTube video. Yeah, just one of the basic models. We'll do just fine. So, but even at that point, it's just a matter of educating right.

[00:49:44] Mike Allton: That's a great message. In fact, I need to do like an entire episode about that because it's one of the challenges as a user, my electric bill doesn't go up.

[00:49:53] Mm-hmm. My fee for using the large language model or Magai doesn't go up just because I choose a thinking model. It's. How much energy is being used on the backend that I don't see. And so that's the quandary, I think, for consumers and why so much more education is needed. Dustin, in the moments we've got left, I'm wondering if you can share with us, maybe just you and I, what's coming up with Magai.

[00:50:18] What are you working on? What can we look forward to beyond the, the features that are already there that we've already talked about?

[00:50:25] Dustin W. Stout: Right now we're, we have a heavy focus on our marketplace. We just released the persona marketplace where creators can create these really high, powerful highly powerful personas, and they can distribute them to our community.

[00:50:38] Anybody in Magai can now go through our persona marketplace and see personas that were created either by our team or by handpicked persona creators. We have some being created now that are really fantastic from people like Jeff Hunter. I think Julia McCoy is, is working on one really smart people.

[00:50:57] I'm sure Mike, you probably have some really amazing personas that you could probably contribute to the marketplace, but the idea is to have an app store. There are a lot of people solving a lot of different really high level problems with these personas. And we want to give the, the community access to all of these and to lift up and, and put those persona creators in front of other people to get their names out there as well.

[00:51:18] Because eventually what we'll have in the coming months is the ability to sell these personas to actually make a living by creating these highly powered personas that, that bring clear value to people. And so there will be this marketplace where. People can buy and sell high level personas, and that's our current focus right now, figuring all the, the mechanics of that out.

[00:51:42] So right now it's only free personas, but soon we'll be adding that layer of e-commerce into it as well. After that, we are heavily focused on agents. There are a lot of breakthroughs happening in agentic ai. And it's super exciting. And so our next big build phase of Magai will be exactly that. Figuring out how to tie these personas into agentic workflows and build systematic workflows where you can say, start with persona A, execute prompt one, two, and three, and then go over to persona B.

[00:52:18] Execute prompt four and five and then go to persona B, et cetera, et cetera, and create these very sophisticated autonomous workflows with your personas inside of Magai.

[00:52:29] Mike Allton: Love it. I can't wait for that. That all sounds fantastic. Dustin, you have been amazing as always. For folks who want to learn more from you that maybe got tons more questions.

[00:52:39] Where should they go to connect with you?

[00:52:41] Dustin W. Stout: You can find everything you need on Magai.co, MAGA i.co. And I'm at Dustin W Stout on all the social things, so you can probably find me wherever your favorite social platform is.

[00:52:55] Mike Allton: Terrific. Thanks Dustin. Thank you, all of you for listening. Of course, as always, we'll have all the links in the show notes below.

[00:53:01] But remember, if this is your first foray into the world of AI in business, I've got the perfect resource for you in the show notes below the AI primer designed to quickly get up to speed on terms the tactics and the underlying technology that's powering this AI revolution. Check it out and I'll see you in our next episode.

[00:53:18] Welcome to the grid.

[00:53:19] That's a wrap on another episode of The AI Hat Podcast. I hope you've gained some valuable insights and actionable strategies to boost your business with ai. If you enjoyed this show, please subscribe, leave a review, and share it with your network. And don't forget to visit the ai hat.com for more resources and to connect with me.

[00:53:38] Until next time, keep innovating and remember, the future of business is human and it's powered by ai.