Stop Wasting Time with Bad AI Outputs: Get Results That Move the Needle with Jonathan Mast

Let's be real, most people are using AI wrong and then blaming the tool when they get garbage results. They throw in vague prompts expecting magic and end up with generic, useless answers. Others believe stuffing their prompts with extra content will somehow make their AI smarter only to get even more irrelevant responses. It's not AI that's broken, it's your prompts. Welcome back to Mastering Tech Growth, where we interview industry leaders to bring you actionable insights and inspiring stories. I'm your host, Mike Sirius, and today's episode is a must-listen if you want to stop wasting time with bad AI outputs and start getting results that actually move the needle. Joining us today is Jonathan Must, founder of Wide Beard Strategies and a leading voice in AI prompting mastery. With over 400,000 followers, Jonathan has helped thousands of business owners and entrepreneurs unlock AI's full potential using his structured 4-step framework. His approach has led to game-changing results, including 5,000% ROI for his clients Today he's here to show us exactly how to take control of AI outputs. Welcome to the show, Jonathan.

Thanks for having me, Mike. I'm excited to be here.

So before we begin, Jonathan, it's a special day for us, which I wanted just to mention. Exactly a year ago, we launched the show. So today's our one-year birthday. We have 50-some episodes in. For no, like, planning or whatever, you just ended up being our birthday guest. Just—

well, I'm honored. That's awesome. Congratulations to you though on a year. That's a major accomplishment in podcasting. Congrats.

I do agree, it's a, it's, it's a lot more difficult than it looks. But let's go and discuss what actually people came here to listen. So we'd love to kick off the show with this question. When I say mastering tech growth, what comes to mind?

I don't know. Um, it's a good question. When we talk about mastering tech, um, To me, I guess it means leveraging technology in a way that makes your life better. You know, it may be different to each one of us, but it's really mastering technology in a way that, again, makes, makes our lives better, makes, makes it a tool that we're using to improve our lives and the lives of those around us.

That's a very noble way to look into it. Let me rephrase this. Let me, let me give this. So you have a business, right? It's a tech venture. Maybe a mobile app, a service, something related to tech. So from all the things you can do to grow your business, which is marketing, you know, sales, people, hiring top performers, thinking about timing, competitive advantage, and all these other things, if you choose one which has the most influence for growth, what would that be?

I think the most influence for growth for most businesses out there is making sure that you're you're adding value to your audience so that your audience is looking to you. In other words, you need to be a subject matter expert. You need to be that person, that brand that people are looking for. And in most cases, I believe that's a person, not a company. And I think that's where a lot of, a lot of business people get confused. Whether you're talking Gary Vaynerchuk, who runs multiple companies worth millions of dollars, But most people don't know about his companies. They know about Gary Vaynerchuk.

Yeah.

And Gary is the brand that drives business to those companies, as opposed to, you know, your local chiropractor or anybody else. So often they're promoting the business and people don't tend to connect as well with businesses. So I think, you know, what's so important, technology can help us do so much better. But at the end of the day, as business people, I think we need to be looking at what is our personal brand? How are we establishing ourselves as the go-to expert in our area so that people are coming to us? Because it's a lot easier to sell our services, our products, whatever, when we have an audience of people that already know, like, and trust us. But we've got to build that. It doesn't come naturally. It doesn't just come through ads. We need to build that that community and that know, like, and trust so that we have an audience ready to sell to.

I love that. I love that. And I love you calling out the fact that people don't connect with businesses and they connect with people. There's so many use cases online where people try to use social media, say Instagram, and they try to push a business brand and it just, just really doesn't work. But the moment they start featuring people and the people's point of view, Suddenly it's a lot better. It's a, it's a, it's a very, very strong shout. Thank you for that, Jonathan. All right. So many users are frustrated when AI systems produce irrelevant or low-quality outputs. This often stems from poorly crafted prompts that lack clarity and specificity. Understanding how to formulate effective prompts is essential for obtaining accurate AI responses. And according to Forbes, the vagueness trap. Is a common pitfall in AI prompt writing leading to garbage results? Jonathan, you've seen many users using, uh, AI getting various outputs from good to bad to garbage. What's the most common mistake in prompt crafting that leads to terrible results or the outputs you just don't want to get?

I think there's two simple answers to that and they're kind of combined. One is not enough context. In other words, we often ask AI to do things and we assume it can read our mind. It can't and it won't, and therefore it gives us garbage results. Tied into that in my mind, and the way to solve that is by when we're prompting, ending our prompts with a very simple phrase, which is, "Ask me any questions you have." Very, very few people actually type in a prompt and then end it with, "Ask me any questions." And that's so important because we do forget to give context. And that gives the AI the permission to ask us clarifying questions. Because here's the good news. If I'm using ChatGPT or Claude or Gemini or any of the tools out there, they generally know if I say I'm a marketer, so I wanted to write a press release. They generally know here are the 7 or 10 things that I really need in order to write a good press release. The problem, most of us doing the prompting don't know what those 7 to 10 things are. And so we give it 3. And then it writes a crappy press release because it didn't have those other 7 things. But if you just end your thing with, ask me any questions, if you didn't give it what it needs, it will then say, oh, hey, Mike, how about this? Can you give me these 3 answers? And then the relevancy of the response, the accuracy of the response literally skyrockets.

Okay. So effectively we're saying the number 1 problem is that people just write a sentence and expect good results. They go like, hey, you know, I'm, I'm, I'm writing this blog post. Can, can you just write this for me? So I think that there's no structure that is saying as they would be asking another person to do something for them. And that is not the right approach. We wanna have a structure within the prompt, which effectively is, you know, giving the role, giving the context and, you know, asking the question. And I feel like I'm stepping into your formula somehow, but—

Oh, that's, you're, you're, you're doing it perfectly. It's exactly it. It's, and it's, it's so simple. It is. It's, tell it what type of expert that it should be. You know, imagine we have to remember that AI has been trained on essentially every book in every library. And while it doesn't exist, if we can close our eyes and imagine for a moment, Mike, that we're walking into a world library that contains a copy of every single book ever written, that would be so mammoth. There would be no way if we were trying to find one book, there'd be no easy way to find it because the facility would have to be so mammoth. Now, AI is fast and it knows that, but it's the same thing. We need to give it some context. So if we don't come in and we don't tell it, hey, I need you to act as an expert copywriter who specializes in writing blog posts because I want to write a blog post, it may or may not be able to figure that out. Now, it's getting better, and so it does a better job of figuring that out, but it's kind of like having a map when you're driving on a trip and not having a map. You can probably get to the beach without a map, but you may not get to the right beach and you may not get there as quickly as you will if you have the map with you to tell you how to get there. And so that's why we recommend starting by telling it what type of expert it needs to act as.

Okay. So am I right saying that this is a step 1 out of your 4-step framework, is that you need to define what role AI should take in this conversation while you craft the prompt? Correct. Okay.

Absolutely.

So, so we give it a role and you mentioned, you know, social media writer or a blog writer or a copywriter. So these are, what are these like job titles or like how do you know like what is the role when you give that to AI?

Well, you really want to think about what you want to accomplish. So let's just for, and it doesn't have to be terribly specific, but we want to give it some direction. So let's say that again, you mentioned a blog post. So if I want somebody to write a blog post, I would say at a minimum, if I was, let's think you're gonna go hire somebody. Who do I wanna hire? Well, I wanna hire somebody who's an expert blog post writer. You could start with that and go, you are an expert blog post writer. Maybe I'm gonna be doing data analysis. And so instead of it being a blog post writer, I wanna tell it you are an expert data analyst who specializes in, you know, extracting relevant data from a certain type of report. Or something like that.

Oh, so you can go longer like that. It's not like, hey, be a leadership coach. It's like, be a leadership coach for, you know, men in CEO positions which manages businesses up to 40 people. And so you can—

that's the more specificity you give it, the better. But as long as we head in the right direction, we're gonna help. But absolutely, you know, if I'm writing a blog post And by the way, writing a blog post is not necessarily a great idea with AI just because most AI models can't write an entire blog post in one prompt. They can help you write it. But going back, I'm gonna, I'm sorry, I'm taking a step back. The second mistake people make after not giving it context is that they assume that it can do too much. I can do one prompt and I, for example, I'm doing a course coming up on how to write a book with AI. You can use AI to help you write a book. One of the most common questions I get is, what prompt do you use to do that? It's not one prompt. It's hundreds of prompts. It's an entire framework that you have to follow. It's a process that you have to follow because ChatGPT or any of the AI tools, you know, Mike, I can't just go and say, hey, I was on a great podcast with Mike. We were talking about technology. Write me a book about that. I'm going to get about 400 to 500 words. Not even enough to be a blog post. So again, quite a short segue. That's the second big mistake people make. But going back to your question, again, as far as, you know, how do we determine who that expert is? Think of it like I'm hiring somebody for a job. If I were hiring somebody to do this for me, what skills and characteristics would I want them to have?

I love that. That's, that's a very good way to think about this. So effectively, almost like a bullet point of skills you would be looking in someone. You just put that in a context of AI prompt when you're defining a role for it.

Correct.

Excellent.

Correct.

Can I specify a human being which has these sort of skills? Can I go like, hey, be Tom Cruise and, you know, teach me his acting style? Like, is that a good role definition?

You absolutely could do— you could absolutely do that, especially, you know, we're— as we're talking about writing content for blogs and things like that, you can even do it there. If you like somebody somebody's tone and style, if I can get that phrase out. You can literally go in and let's say, you know, for example, that I really like, I'll pick on Russell Brunson, who owns ClickFunnels. If I'm a marketer, I may like his tone and style, and I can literally go in and say, use the tone and style of Russell Brunson. And it knows a lot about who Russell is, and so therefore it will use that. And if you're not sure, you can always ask AI, you know, tell me about the tone and style of Russell Brunson, or like you said, I aspire to be an actor and I look up to Tom Cruise and like to be an actor like him. What things about Tom Cruise can you teach me so that I can be an actor in the same style that he is? And AI will then analyze Tom Cruise because it knows a lot about him. He's famous. And it would then be able to teach you the characteristics that would lead you down that direction to be, to learn as if you were learning acting from Tom Cruise himself.

Okay. So that's step 1, defining a role. So we put that in, and is there any specific way you want to do this, Jonathan? Is this like act as blah blah blah, or it just doesn't matter? It's just that the first paragraph of our prompt or a sentence or 2 sentences are very clearly given role.

Yeah. Yeah. Exactly. I just make it the first sentence. You can do it any way you want, but it does, it does tend to help, again, the AI give you the best possible response.

Okay. So we effectively giving AI an avatar in this prompt, in this conversation, to be someone which effectively narrows its focus to the skills for that avatar, right? So what is this step 2?

Step 2 is the context that we talked about. It's giving the AI relevant context to what you're doing. So carrying on through in the blog post example, if I wanted it to write a blog and I wanted to write maybe a short blog, which you could do, I would then want to tell it who's my audience. What's the topic I want the blog post about? What perspective do I want to take on that? All those things. So, and a good example would be, uh, I just recently put out a, a post on social media that's getting a lot of praise and hate at the same time. I love those. Those are always the best ones.

Mm-hmm.

And it was basically that certification, AI certification programs from accredited universities are bunk. They're not worth anything. And it's because universities are too slow. It's nothing against universities, but— and it's only related to AI. So if I was going to write a blog post about that, I would say, you are an expert blog post writer and you specialize in creating engaging content. Second step, I want to write a blog post about how AI certificates and certification programs from accredited universities are not valuable because of the fact that universities operate too slowly in an AI marketplace that is literally changing every day. So I'm, I need to give it that context. And then I would probably even add, I want the blog post to prove my perspective that AI certification programs from universities are not as valuable as, as certifications from universities in other areas. So that might be the whole You know, I might do 2 or 3 sentences for context then.

Okay, so effectively you're almost like giving your big idea, right? Like, like, here's my big idea, this is what I want to do, this is what, this is what I want to express. You give context about that. Do you want to specify what it actually returns or how it returns, like styles and, and, uh, writing situation, or is that more for a role?

No, you can, it doesn't really matter. You can put it in either spot. It's, it's, um, AI is a natural language tool, so it can be much like we're talking today. You know, you and I, Mike, may be working on a project and I might not get everything in the perfect order, but that's all right. You're going to understand what I was talking about. And as long as the questions get answered, it doesn't really, the order itself isn't particularly important. So yes, if I wanted to write that, say for example, in the tone and style of Socrates, not that I know what Socrates sounded like, but I could say I want to write this in the tone and style of Socrates. And what it would do is it would then use the writings of Socrates that it has analyzed to determine what Socrates' tone and style would be, and then would write this article or blog post about what I wanted, making the points that I wanted it to make in the style of Socrates.

Yeah, it just blows your mind when you think you can go in and you go like, hey, You know, be like 800 AD poet, you know, write some poems for me about nature. You know, you just give some ideas and just launch a poet, you know, book. It's just, when you think about it, it just blows your mind how much of this will happen later in our lives still. And it's going to be so hard to actually distinguish where there's a human creativity and skill, where there's AI. Drifting, it just took me somewhere. The moment you mentioned Socrates, I was like, okay. Ideas came to my mind. So let's go, let, let's come back to the context. So can we give, can we give context in the sense of like books? And I know not all AI, uh, models do that, but let's say if we take ChatGPT as an example, can I just go in and go like, I'm gonna give you some documents, right? I'm gonna upload to my prompt some documents. And what I'm saying to you is that anything we gonna be generating, is going to be based on these documents or from these documents. So you're not allowed to pull in any information you have online or internet or, you know, but just my, you know, the scope of my documents I gave to you. Would that work?

Maybe. You can do it. I mean, you can absolutely upload information to a ChatGPT or any AI model at that point and have it do it. Whether or not it actually complies and only uses that, is where it gets to be difficult because it's likely still going to use other aspects of its knowledge base to help out. What I find is if you tell it that the uploaded data should be its primary data, that it does do a good job of sticking to that probably 95% of the time.

Yeah, I was sort of struggling with that approach and that's where the comment is coming from. What really works for me to understand if it's using my information or not was to say, use my information as primary, just as you mentioned, Jonathan, and then, you know, put in the brackets if you're using not my information, as in tell me when you're pulling something in not from my resources in the brackets, and it says like internet or source or whatever.

Right.

Then you can actually distinguish, because when it gives you output and you're giving, let's say, you know, 10 documents, which are a lot of input of information, you don't even know what's in these documents, right? Like you didn't read them, you don't know them. By memory. So when you get output from AI, you're like, is this coming from, you know, my set of documents or is it getting filled? So that little trick allowed me to sort of, you know, draft the responses in a way where I can—

That's a great tip.

—understand that. Okay, so that's step 2, right? So we've got step 1, the role, the step 2, the context. What is the step 3?

Step 3 is the simple part. It's what most of us type in as our prompt. Write me a blog post or whatever your question is. That's again, it's what most people put in as your original blog post. So whatever that command instruction or whatever you wanna say is, that's step 3.

So that's effectively the command, the task, whatever we wanna get. Exactly. What we're saying. Is there any bad way of doing this? For example, giving too much of commands with one prompt? Or, you know, not saying— Certainly, yeah.

AI, much like humans, if, you know, if I walk into my office and my assistant gives me 20, a list of 27 things to do, I can promise you I'm forgetting everything but maybe 2 of them just because it's more than I can take. And AI is very similar to that. If we give it too many things to do, it's simply not going to follow through on that. So it works best in a prompt if we ask it to do 1, 2, or at the most 3 things. I find once we get beyond about 3 tasks, it really starts to have trouble following our instructions. And we have to keep in mind that while we're dealing with computers that shouldn't have that problem, these computers and these applications still have a limited amount of memory allocated to a conversation. And so if I use up too much in my instruction, it doesn't have enough memory to actually execute on that. So it's much better to do what they call a chain of prompts at that point, or a series of prompts. And my first prompt might be, Write me an outline for the blog post. The second one might be, go through the outline and write me a summary paragraph for each section of the outline. Then the third prompt might be, for each section, one at a time, write me that section of the blog post. And each one of those would be a separate prompt. And then at the end of it, I might say, great, now analyze the entire blog post you've written and write me a key takeaway section to put at the beginning of the blog for for people that are skimming. And then when I'm done, I would probably then say, and now analyze everything and write me an FAQ section of 5 FAQs and answers related to this blog. But that's probably 12 to 15 prompts in sequence as opposed to one prompt. And the result that I would get would be mind-blowingly different. Yeah.

And then you don't go at the end like, hey, 'Now return the whole block for me,' because it's just— the output is just way too big. So you're gonna get too big. It's much shorter version. So that means anytime you finish with that segment of conversation, which is your prompt, you, you want to probably take and copy that output in a document which you're working on, and then continue, uh, compounding on these prompts and getting these other segments to, to grow a much longer content. Is that, is that correct assumption? Yeah.

Yeah, the one exception to that is ChatGPT has relatively recently added a feature they called Canvas. And I stress Canvas as opposed to Canva. A lot of people think it's the graphic tool Canva, but it's not. It's called a tool called Canvas. And Canvas actually creates a working collaborative document that allows you to make multiple prompts and it will then continue adding to and rewriting that document with you. And you can even go in and make changes to it and then ask it to make changes. So you can write an entire blog post and have it compile it in one canvas, as they call it. But it's faster, in my opinion, to go and do it just the way you described, which is, here's my outline, copy paste into a Google Doc. All right, here's section 1, write that, copy paste. And then when I'm done, when I'm ready, for example, like the key takeaway in the FAQ section, I can then download that compiled document that I have re-upload that and go, here's the blog post you just helped me read. Now, based upon that, now write me key takeaways. And then it would analyze the whole blog post for you.

So effectively, we said— what we're saying is that's two different things, right? We would make a split two different ways. Yeah. So in one, we would finish with the document and just get a PDF and start the new prompt, upload that PDF, and it's like a from scratch, effectively redefine the role. The PDF is the context now, and then just continue prompting with an FAQs, as you mentioned, or anything else we need. Yep, exactly. Okay. Right. So we got 1, 2, 3.

What's the 4th one? 4th one is very simple. Ask me any questions you have. Just literally end your prompt with that statement to the AI model. Ask me any questions you have.

What if it goes, I don't have any questions?

That it will— it'll just then proceed to answer your question or your prompt.

Okay. So if AI— effectively what we're saying, AI will work out if it needs more context to make sure that it delivers against your task by defined role. And if you allow it to, to sort of ask these questions, it will ask you. So you will provide more context and they will use that to just better its output for you. Mm-hmm. Okay, excellent. All right. So we've got role, context, tasks, and ask me anything you need to know to basically do better. Is it like, is there a phrase for this or is it literally goes like, ask me how you said that, Jonathan? Ask me.

I literally say, ask me any questions you have.

That's it. Ask me any questions you have.

Okay. You could, you could type it in different ways, but that's all I would put in place.

Okay, so now we have many of these tools, right? So we have, I wanna say, I mean, I said many, like it sounded like it's probably like in hundreds. It's, it's not, uh, it's probably like in tens, but still there's, there's, you know, we got ChatGPT, we got Gemini, we got Claude, we got Deepseek now coming out. And you know, I'm, I'm sure there's more when it comes to content generation, not image, but text. And I mean, they do images as well. Grok as well on Twitter. So there's like, you know, bigger companies, big tech giants, let's say like this, you know, they're releasing their own tooling. And the idea is that some of these are better for certain things and, you know, they're better for, let's say, long content, you know, maybe blogs. Others are better for maybe shorts and social media, whatnot. How do you make selection? Like which are which, like how do you actually use the tools for the right job, as they say, like even, you know, to get more leverage?

Well, I think, you know, one of the biggest things that we struggle with across the board is there's so many tools to choose from. Which one do I pick? And I want to encourage everybody that's listening here that in most cases you only need a handful of tools and that it's probably much like going to buy a new automobile. We all have brand preferences and you may buy one brand of car, I might buy another brand of car. But if we're driving to dinner this evening, chances are we're both going to get there. And AI models really work very similarly. There's a number of what I would call major models like ChatGPT, Claude, Gemini. Those are probably the 3 biggest ones. Grok now from X or Twitter is becoming very popular as well. I'm a big proponent of starting with ChatGPT. It has the most functionality. It is very good across the board. Is it perfect? No. And, you know, I can argue that Claude may write a better book post for me or whatever, but the differences that we're talking about are minute differences for the most part. They're not massive differences. It's not the difference between I can't drive my car and I can drive my car. It's the difference of All right, I really, you know, I drive Hondas and therefore I like how Hondas drive because I'm familiar with them. And you may drive Toyotas and go, you love Toyotas because you're familiar with them and like the way they fit. We could switch cars and we'd both still get where we wanted to go. We might not like it quite as much because we weren't as familiar with it, but we'd get where we wanted to go. So I encourage all business owners and entrepreneurs and everybody with AI, and this is hard for me as an ADD entrepreneur, but try not to get caught in the bright shiny object syndrome because there's always new tools coming out and the marketing always makes it sound like it's better than anything that's ever been released before and you need this tool. No, you don't. If you have ChatGPT, maybe Perplexity for search, which I like really well, and then pick an image tool that you like because I'm not a big fan of DALL-E within ChatGPT. So I use Midjourney. Today, Those 3 tools, ChatGPT, Perplexity, and Midjourney, probably make up 95% of the AI use that I use across the board.

Do you use anything more, Jonathan? So I got Perplexity, ChatGPT, and Midjourney, which is image generation. Do you use anything right now in your toolbox?

Because of my role and the fact that I educate a lot of people about AI, I do use almost all the tools some of the time. And I shouldn't say almost all, almost all the major tools. So, you know, I use ChatGPT, I use Claude, I use Groq, I use Perplexity, I use Llama from Meta, I use Mistral out of France, I use Deepseek out of China, I use Qwen and Kiwi or Kymee out of China, but not exclusively and not all the time. But I literally use all of those tools. I use Google Gemini. I don't think I mentioned Gemini. I use all of those tools. Because I need to be aware of what their capabilities are in that so that I can help guide my clients, because that's what I do. I do AI coaching. So I need to give them guidance. But what are the tools that I use on a daily basis? ChatGPT, Perplexity, Midjourney, and then a tool that a lot of us don't talk about, but is a— I use a note-taker tool, and there are a number of them. I happen to use Read.ai. There are other tools like Otter and Fathom and probably many, many others. But that has actually been one of the biggest productivity hacks that I've had, has been having an AI note-taker now attend my meetings with me, not in place of me, but with me, so that I can keep track of what's going on in those meetings. And I use that every day, every, at least every workday.

Oh my God, these are excellent. And I love you called them out. I was, I was waiting until you finished listing and I was like, Will I give this or will you tell? Will I give this or will you tell? But these are game changers. I mean, honestly, jumping on a call without thinking, they just follow your meetings because you give them the rights. And then you come in, you chat, and boom, at the end you have your minutes of the meeting, actions captured, you know, good splits. Like, they do great job. And yeah, it's a huge, huge productivity boost. If people listening right now, you're a business owner, you have meetings and you're still taking your notes manually, or you have a virtual assistant taking notes for you? No, just, just check these tools. Okay, so we got 3 tools, and effectively what we're saying— and Jonathan being the expert over here— what he is saying is that you only need those 3 tools. That's gonna cover 90% of the use cases. Don't get distracted, have a focus, choose one, become better at prompting, you'll get the results. It's not a tool, it's basically your prompts. That's what we see. It is. Okay.

Yeah. Yeah. And, and, you know, while I have 3 tools that I like, if, if somebody's listening and they're like, Jonathan, I'm just getting started. I don't want to have a bunch of subscriptions. What do I do? Literally, my recommendation is to start with ChatGPT because it can do virtually everything that somebody needs and it's super easy and super effective and you can get started for free. You don't have to pay it anything. What you're going to find out in any of the free tools is that you'll just run into a usage cap at some point. When they'll be like, Mike, you need to wait 3 hours before you can do any more work, and or you need to give them $20, in which case then you can do more work. So there's nothing wrong with trying those and starting off with free as well.

I get usage cap even on my paid accounts.

Yes, I do too. I do too. There are usage caps across the board. Yeah, especially on weekends. I actually— can I share an advanced tip, Mike? Can I share a little more advanced tip? Go for it. So if like you and I, you run into that where you're— because I'm doing hundreds of prompts a day, you run into usage caps on the paid plans. There, there is a great aggregator tool out there called TypingMind, and I, I don't get any money for supporting them, but they're a great tool. What's different about TypingMind is that it connects to the API or the actual backend engine of all of these different AI tools. Now, you're not paying a monthly fee, you're paying per usage. It's like putting fuel in your automobile. The more you drive, the more gas you need. The same here. The more you use it, the more you're going to pay. But I can go to TypingMind and I don't have any limits. I may spend more money, but I don't have any limits because I'm paying for my usage as I go with it. So within that, I generally use Claude and ChatGPT, at least their engines in the background. But that's just a little advanced tip. If you don't know what I'm talking about, an API makes you think, oh, I don't know. Don't worry. Just don't go there. But if you start using tools and like Mike said, you start running in the usage caps, TypingMind is a great tool to look at.

Yeah, it takes time to set it up, but once it does— yeah, once you—

once you— it's an advanced tool. It's definitely not a beginner tool.

Yeah. Then you walk into your automation and then you start asking questions. Well, okay, what if combined this and this?

And then you start creating this rabbit— this rabbit hole gets very deep very quickly. Yes.

Yeah, I do agree. Okay, so I wanted to follow up on the tools, right? So, so we saying the ChatGPT, Midjourney, and, was it Perplexity for the search, right? Is there any selection if we're thinking about speed and cost? Like, are we saying basically ChatGPT, Midjourney, and Perplexity's cost is, is well balanced with the market? And if we care more about how much we spending because we want to give the tool to maybe 10 people within the company or maybe 100 or maybe, you know, 200. And as well, we care about the speed of, you know, returns coming in. If we factor these things in, is it still ChatGPT, Midjourney, and Perplexity which sits at the top?

It's, it's probably still going to be there, but I think then, you know, especially if we're rolling it out to a company, we need to take a look at what type of functionality they need. And at that point in time, it may make sense again, depending upon usage, but understanding that most of our team is not going to run hundreds of prompts a day. They might run a dozen a day. If that's the case, you may want to take a look at an aggregator model, and there are a number of good ones out there. One that I'll mention is called Magi, M-A-G-I, and it gives you access to like all of the various AI models for a monthly cost. The downside to a Magi, if there is one, and I know the CEO, it's a great company, is that there's a usage cap because they're paying the fuel essentially as you go along the way, so it's not unlimited. But the nice thing about it is I can, for example, as a business owner, I think, and we use Magi in my company as well, I think I'm spending $50 or $60 a month, but now I can allocate usage to different members of my team. So I don't have to buy everybody on my team a $20 a month or a $30 a month license. I can buy Magi, pay them X number of dollars, and then I can assign usage to each member of my team. And so that may be a solution. I'm not saying it's the right solution, but it is definitely one of the aggregators that I recommend if people are looking for that type of a solution that can work really well.

Okay. So if we're bringing this to the companies and let's say I'm a founder or a business owner, CEO, and I'm like, okay, well, everyone's talking about AI, everyone's talking about productivity boost, you know, We know it works for the meetings, for the emails, for the document analysis, social media. It works on a lot of things. So I want to roll this out in the company. At the same time, I know that people will not utilize it well, or at least, you know, they will start playing. But how do you actually embed AI usage within the companies? Like, do you have your go-to strategies of how you help companies to go like, we totally noobs to like, we're going to utilize this and get this advantage on market?

So one of the biggest benefits of using AI throughout your team is the fact that they're going to save time. They're going to also probably deliver a better quality of product, but they're going to be able to save time. And I was just out in San Diego, California talking to a group of people about this last week. And one of the things that I recommend to business owners is as you're beginning to implement AI within your company, make it a win for your team. And the best way to do that is tell them that the first 30 minutes of time that they can save every day because they're using AI is their time to do what they want with. Not, Mike, if you save 30 minutes, I want 28 of that back. I want you to do more work. No, the first 30 minutes you save. So this week we're going to roll out AI. This month we're going to roll out AI. If you save yourself 30 minutes a day, you get to put your feet up. You get to have another cup of coffee. You get to take a deep breath. Whatever it is, I'm not expecting you to get more work done. That'll be next month, by the way. So owners, don't freak out. We're not gonna, we are gonna go there. But the first 30 minutes that they save, give it back to your team. Let them take a deep breath. All of our team have more tasks to do in a day than there is time to do them. So give them that time to do those tasks better and with lower stress. Then month 2, month 3 down the road, as they start getting used to using it, they're gonna start saving a lot more than just 30 minutes a day. And as they do so, then we start leaning into the fact that, okay, Mike, because of your using AI, maybe we can get 20% more work done, but you're still getting that done at a less stressful or frantic pace. So my advice, when you're rolling it out at first, make it a win for your team, not just a win for the company. And the other benefit, by making it a win for your team and giving them that chance to take that deep breath, They're going to love working for you that much more. They're going to be happier. They're going to be more productive. You're going to end up with a better workforce. And then month 2 and month 3, as we start saving more time, then we can start recapturing those efficiency gains to add to our bottom line.

I love that. It's a great rollout plan. Love for sharing that. So we are using— so I'm an employee, right? I'm getting this tool. And the first thing I'm thinking is it's another skill I need to have, right? It's like, here's a new service or here's a new technology. A good example is like learning spreadsheeting well, right? So it's like, there's something I need to learn now. You know, there's prompting, there's utilization and things like that. And then I need to manage these new developed skills to actually get better at it. So to be an actual value add to the business, let's say like this. So when that happens and I'm— I am employee, I want to get that leverage. How do I sort of maintain that skill? Where do I go to understand where the market is moving, where AI is moving, what's relevant to this now? Because this is a totally new tool in my toolbox, which I now need to manage as employee.

Yes, absolutely. I think, you know, if you're an employee, I think one of the most common things we see is the fear that an employee has that, uh-oh, if I embrace this AI, my boss isn't going to need me. I'm going to lose my job. I believe just the opposite is true. And I think that's where you're headed with this, Mike, is that you, you as an employee are not likely to be replaced by AI. I'm not going to say it couldn't happen, but it's very unlikely. On the other hand, you are very likely to be replaced or overlooked because somebody like you has learned how to leverage AI. So as an employee, I think you know, one of the best things to— I encourage people to do is start using AI regularly. My advice in that is one new thing a day. Don't try to do everything today. We can't do that. I mean, I'd love to do that, but it doesn't work out that way. One new thing a day. So maybe today it's, you know what, I've got an email that I need to write. And I'm not sure— I need to set some guidelines. I need to be straightforward. But I need to do this in a nice professional way. And I'm not sure the best way to say that. Copy and paste the email you got from your client or your boss or whomever. Paste it into ChatGPT and say, I need to respond professionally. I need you to help me write a response. And I want the response to basically say the following. And you'll be amazed. Within seconds, you're going to have an incredibly well-written email that you may need to edit just a little bit. But instead of spending 20 or 30 minutes fretting about this email that you needed to write. You're going to have it done in 2 to 3 minutes. Maybe you've got a large data file and the boss comes to you and says, hey, here's, here's a report we just got. Maybe even worse yet, maybe it's a government report. So, you know, it's massive. And we need you to go through and figure out how this is going to impact our, our, this part of our business. And maybe it's 300 pages and you're like, great, I actually had plans this weekend, but now upload that report to ChatGPT and say, all right, I'm in this role and I need to analyze this report and determine how it's going to impact us in these capacities. And guess what? In about a minute and a half, you're going to have an entire list. It's going to scan and read that entire PDF. It's going to give you an entire list of the impact, and then you can ask it further queries. And that's take the next step to go, okay, now that you've said that, give me more detail about point 1 or point 2 or whatever it is. And you can now analyze that entire PDF in maybe 30 minutes instead of 3 or 4 hours.

The instant question I get, can I trust it? Like, can I trust if I didn't read that document? It's, it's very important. It's 300 pages. I don't want to read it. So I will upload it and I will start, you know, prompting. But how do I know that I can trust what I'm getting as output? What if it's hallucinating, you know, and giving me something from online in the document and I'm like, I will see that.

And it can do that. Yeah, spot check. So this is, I refer to as trust but verify. In other words, trust that you're likely getting the right out, but then verify a few things. So ask it when it's doing that to give you the pages in the PDF where it's finding this information. And then let's say it gives you 30 points, go check 5 of them and make sure they line up. And the last part of it is be candid and open with your report and whoever you're providing it to that you used AI to help with that. So, you know, one of the things that I do, for example, is I do training courses and we then take the transcripts from those trainings and we feed it to AI to summarize. I'm very open and honest when you go to there. Here is an AI summary of the 2-hour transcript of our training. You'd be better to watch the whole 2-hour training. But the reality is I know that you may not have time to do that. I may not have time to do that. So chances are the transcript, even which I'm going to tell everybody is AI-generated, is likely going to get us enough of the value. It's kind of that old 80/20 rule. If it gets me 80% of the way there in 20% of the time, then I've really won in that process. But trust but verify.

I love that. Trust but verify. Exactly that. Do you think there will be a point in time where, where the trust is just baked into the tool and people will have No reasons to not to trust it?

I think that people will begin to trust it, but I hope they don't ever trust it blindly because then it gives the opportunity for people to program the AI in ways that would lead us down the wrong path and give us false information. So I hope we never get to the point that we blindly trust it. I think that for, again, much of what we can use it for in business and as employees, many of the things that we can use it for, we don't need to do a lot of verification. If I'm, you know, Mike, if I'm having it help me write an email to you, all I need to do is I need to go over that email and make sure it didn't say something stupid. And if it did, I delete it. So it's pretty straightforward. The other part of that is if you add in that last part of the prompt framework, step 4, ask me any questions you have, then the AI model, when it's analyzing, can come back and go, hey, I have a question about this or about this, and it will get much better information Normally when it gives you what we would call hallucinations or bad information, it's because the model is assuming. And we know what happens when people assume. We become and look like the south end of a northbound mule. And so we don't want to assume. But when we let AI assume, the same things happen. So by giving it permission to ask you questions, it'll avoid most of those hallucinations. I love that.

I think that's a missing bit. I'm just thinking about resources I read, tips I got. That one, to ask AI to ask the questions, that's, uh, overlooked. I put it like that. That's definitely the overlooked one. So we mentioned that unlikely that your roles would be replaced with AI, but it's very likely that you might be overlooked because you don't have skills to use AI tools, which becoming— Correct. A lot more important, right? So with that said, this is where we are right now. This is 2025. This is where we are right now. What will happen in 5 years?

I can't predict. I know that as humans and employees and employers, we all need to be nimble to be responsive. I can't predict where AI is going to be at in 5 years. I— 5 years ago, I wouldn't have predicted we would be where we are today. Um, but the good news is, is that we're still, we're still running the show as humans. We're still controlling, and we can always unplug AI if we need to. It's a tool. It's nothing more than a tool. It's a tool that amplifies our skill and our experience. And if you're out there wondering about it, I, my best encouragement is just learn how to use it as a tool so that you can excel and deliver more value, whether that's more value to your community, to your family, to your business, to the employer, to the clients you serve. Learn how to use it to add more value and you will, you will thrive through the changes that come up, not be caught flat-footed and, and have problems.

That's great advice, Jonathan. That's great advice. I, I, I'm fearful that at some point we're not going to be able to unplug it. You said like, we, you know, we still can unplug it, and I'm like, you know, could you unplug internet right now? Like, as a, you know, it's like almost unplugable, right? So if AI becomes a valid point—

yeah, it's a valid point. I, uh, again, I probably just operate in a different sense of, of perspective. I don't, I don't want to live in fear about what's going to happen. And so, you know, when I look at major technological advancements that have happened over the last century or so. Every time that's happened, whether that was the internal combustion engine, whether it was the PC getting on a desk, whether it was the fax machine, whether it was the internet, any of those things, all of those, by the way, when they came out, fearmongers were out there telling us how it was going to ruin society, it was going to ruin our lives. You know, we were, oh my goodness, we can't do All of those created far more jobs than they did destroy. Yeah. Did it mean some people were displaced? Yeah, there were definitely people that may have been displaced in the process, but those, those were few compared to those that had the opportunity to get new opportunities. And those that were displaced weren't forced to sit at home and do nothing. They also had opportunities to grow and learn and make changes based upon the new technology. We're going to have those same opportunities. I do agree 100%, Jonathan.

And the case which always comes to my mind is when, when personal computers appeared, you know, the first Apples, and everybody was saying the same thing is like, you know, it will take our jobs, like, you know, you don't need to write now and all of that. And it just created so much more. None of the companies who bought a lot of personal computers at that point in time actually fired people. It's one of similar situations which I fully agree that you need to readjust yourself and your skills to just adapt to the evolving market and you will just be, you know, an employee people want rather than the one which loses job because AI came.

So, yeah, you know, it's the same reason we went to university and things like that. We wanted to learn to be the best possible in our field. AI is just another tool now that we need to learn how to use. Yep.

I love that. Can we do a rapid question type of round, Jonathan? So I got like 10 questions and you got 30 to 60 seconds to answer them. I will not dig deep.

I'll do my best to keep them short.

Here you go. Don't keep them too short because then, you know, it's just like, well, actually, fair enough. Whatever comes natural. Okay.

All right. I appreciate it. Let's go.

What's the most misleading advice about AI prompting you heard?

That it's going away and we'll never need to learn it. Okay.

For me, it's talk it like you would be talking with a human. Doesn't really work. If AI prompts fail, what's the first thing you check?

Generally, I look back at what I asked because it's probably my— like I made a request to you and I didn't give you enough information. So Chance, the biggest advice I can give there is have a conversation with AI. Don't expect to give it one prompt and get perfection. If you give it a prompt, you ask it a question, and it doesn't give you the right response, clarify what you were looking for, clarify what you were asking, just like I would with you if we were working on a project together.

Okay. What's the hardest type of problem to solve with AI prompting?

I think many people look at AI as a magic wand that's going to be able to do everything, including going to the store to buy the roast cooking the dinner, washing the dishes, and getting it all done. AI is not an autonomous agent. It is something that can do tasks very well, but even those tasks need to be directed. I often refer to that we need to be the conductor of the orchestra. The AI may be our orchestra, but we still need to conduct it. So there are a lot of problems that AI cannot solve at this point in time, especially those that require autonomous action.

Okay. What's the prompting technique that works that nobody talks about? Let me answer this, Jonathan. It's the 4-step formula of Jonathan Must. That's my belief.

I'm biased. I think, you know, you will find a lot of them that the first 3 are the same. I have not found anybody else that teaches like I do that 4th step, which was ask me any questions you have. I really do believe that's the game changer. I know that's an AI phrase now, but That's the game changer in prompting, is to end your prompt with, ask me any questions you have. And it will— then when you take a moment to do that, it'll really change the quality of your responses.

Okay. If you had to train someone in AI prompting in 5 minutes— let me rephrase. If you try to train someone of AI prompting in 60 seconds, well, how these 60 seconds would sound?

Tell the AI how it— what type of expert it should be. Give it relevant context, ask your question, and end with, ask me any questions you have.

Brilliant. What's the biggest mistake even AI experts make when they crafting their prompts?

We make too many assumptions that AI will figure it out. In other words, we don't give it enough context.

Yeah, I fall short with this one as well. I mean, my prompt is almost like a A full page. I just, I, it's just way too big and I, it just explodes all the answers. Yeah, that's 100%. All right. How do you get AI to generate complete original insights and not just regurgitated data?

Well, I don't know that, I don't know that I believe that anyone can generate completely new insights on anything. We are all influenced by those around us, by the news we see, the books we read, the music we listen to, everything. It's all inputs that influence us as individuals. AI is influenced the same way. I don't believe I can come up with a truly original thought because I've been influenced by others for 55 years. I don't believe any of us can. So I think asking AI to come up with a truly original thought is a losing game because I don't think it's ever going to be possible.

Okay, so when the market talks about AI creativity, you know, as a thing, we as humans, we could create new hypotheses in science or, you know, or draw a picture which never existed. You know, we call this new. So we're saying that AI wouldn't be able to do that.

Well, I can say if I ask AI, and I use it all the time to help me do artwork and things like that, it's not— it may come up with something that is unique. But it's not necessarily original, if that makes sense. It does make sense because it's going to be, it's going to be based. For example, if I'm basing it on a prompt, I'm going to tell it I like Disney Pixar style. So be inspired by Disney Pixar. That means immediately it can't be original because it's already being inspired by something else. It can be unique and we're probably talking semantics, but it can be unique without being original.

I love that. So are we saying as well, because you said I don't believe anybody can create a totally new thing. So we're saying even as humans, we always use something we know from before, and it's just always sort of like compounding. It does.

Hmm. Now look, my belief is very simple, that we're, we are a result of our inputs, just like AI will be, and that we all have inputs. So even though someone developed AI technology, they were inspired to do so by other things in their life. And so it may have been a unique tool, but it doesn't mean that no one had ever had that thought before. No one had ever had that. I mean, we can go back and look at sci-fi stuff of things that they were, you know, predicting and that they, they were talking in Star Trek about devices that we now have and use. Does that mean that Apple was original when they came out with the iPad? Maybe, but, or I'm sorry, the, the iPod. Maybe I would contest that no, they weren't original because Sony had already come out with the Walkman and they were inspired by the Walkman to create something unique, a digital version, but not original, if that makes sense.

It does make sense. Well put. What's a real use case where AI prompts saved you serious time or money?

I think probably more than anything else, it is in the content creation that I do. Um, I create a lot of content. I want my content to be incredibly helpful, and I could not create a lot of that content at the capacity that I do if I didn't have AI to collaborate with to help me in that process. Okay. And that results in making me money because I've literally built my business around that content that I've created to become an influencer, to become a leading voice, to become that go-to expert.

In my niche. Are we saying that people utilizing AI can become niche influencers fairly easy?

Um, can become, not fairly easily. So my journey has been almost 2 years in the process, but, and, you know, I'm, I'm producing hundreds of pieces of content every week from social media posts to videos to blogs and that type of stuff. It's not easy. I work 12 to 14 hours a day. I love what I do, but it's not easy, if that makes sense. I don't find it arduous, so I enjoy it. But it is, you know, we are not at a point, nor do I see we're ever going to be at a point where AI is that proverbial money tree in the backyard that we can go, oh, I'm going to use AI and it's going to make money for me. No, it's not. We still need to direct it. We need it. We're going to collaborate again. It's a tool that amplifies skill and experience.

Well put. What's the weirdest AI prompt experience that actually worked, if there is one?

I think probably the weirdest ones that have worked is when I'm brainstorming with it. I find that it works most effectively when I use my voice inputs. In other words, literally my microphone. I turn on transcription on my computer, and in a brainstorming session, instead of trying to type out prompts, that I can literally have a stream of consciousness, a thought that I can talk, give that to AI and let it process and come back. Sometimes those don't follow my perfect prompt framework at all. It's literally just, I'm sharing information, I'm sharing context, and it's then taking all of that and sorting it. AI does a great job, especially for someone with an ADHD brain like mine. Of organizing and contextualizing what is sometimes a very random stream of thought coming out of my mouth.

So how you do that, Jonathan, is that you just have a microphone and you talk straight to, let's take ChatGPT as an example. So you talk straight, that just taking the transcript of what you're saying, and then you just click enter and it just sends to ChatGPT.

Kind of. So on both a PC and a Mac, and I'm not talking about using their voice interface where it talks back to you. I'm literally talking that I will click inside the prompt box, and on my— I have a Control key for it. I press it and it turns on my microphone, and at that point in time, it will transcribe in real time as I'm talking. Both a PC and a Mac now can transcribe that in real time, and so I can, I can have all that done. I do the same for most of my messaging, most of my emails. I rarely type that anymore. I typically will click in the box, press my Control key, and then I'll be, hey Mike, I'm looking for the following, is this something you can help me out with, you know, I really love your feedback, thanks so much. Much, much faster to talk that or say that than it is to try to type that.

Okay, and so you just give 1 and 2 and 3 and 4 to ChatGPT as an example, and then what do you get back? Like, what do you ask to give it back? Like, Just a summary of your thoughts, putting it more in a structure?

It's like, depends on, depends on what I'm doing. I mean, I do even a lot of my basic prompting I'll do by voice. So again, as we talked about that blog post earlier, okay, instead of typing it, I might say, you're an expert copywriter who specializes in writing blog posts. I want to write, I want to write a blog post about how I believe that AI accreditations from universities don't have the same value as other accreditations from universities based upon the speed of change of AI. I'd like you to go ahead and write me a comprehensive blog post that's targeted towards entrepreneurs and students who may be considering AI certification courses from universities. And I'd like you to ask me any questions that you have. Please begin by writing the outline.

And you just say that and then boom, enter.

I'd say it just like that, and then it would respond accordingly, and it will all appear as text inside of ChatGPT, and it'll give me a text-based response.

Would you read it, or would you click a voice to tell you back the response? Generally, I'll read it because I can read it faster than And it says, yeah, so, okay, so we're effectively, we're saying the efficiency here is that you can say that prompt quicker than you can type it. Correct. Yeah. Okay. Yeah. What's the single best habit for improving AI prompting skills over time?

Use it consistently. Most people are looking for a shortcut. They're looking for a list, the perfect prompt. The perfect prompt doesn't exist. I do believe in my framework. I believe that's a great prompt framework, but the perfect prompt doesn't exist because even if we're doing the same thing, Mike, your needs and my needs are going to be different. So using AI in my mind is the absolute best way to get better at it.

Okay. And what's a valid— not a valid— what's enough usage to actually see your skills developing? Are we saying 1 prompt a day? Are we saying 10 prompts a day? We're saying, you know, spend 15 minutes, uh, using ChatGPT or just make sure you're using it daily for whatever amount of time, just never quit using it.

I think if you, if you're using it daily and you're trying to do one new thing a day with AI, you're gonna learn very, very fast.

Mm-hmm. I love that. All right, Jonathan. So if our listeners want to dive deeper into AI prompting, where should they go and find you?

I'm easy to reach. Just remember my name. It's JonathanMast.com/Linktree. So JonathanMast.com/Linktree. I've got all my contacts there. I've got free groups, tons of free resources. There are some paid classes there too, but most of it's free. Just go ahead and connect, reach out, and I'd love to help people figure out how they can really use AI to save time.

Excellent. And for the ones For your convenience, just look at the show notes. If you're watching this or listening, there will be a link to Jonathan's Linktree so you can find him easier. And if today's conversation helped you to rethink how you approach AI, don't just let it be another episode you listen. Try out Jonathan's 4-step framework and get much, much better results. Also, if you find value in episode, you know that we would love you to subscribe. Send us a comment. We do love to help. And as always, until the next time, stay safe and see you online.