Transcription
Most people don't get bad results from AI because AI is weak. They get bad results because their prompts have no foundation. So, if this sounds like you, if your prompts feel like kind of hit or miss, maybe they sound generic or they require endless rewrites, this video is going to fix that.
But here's the deal. Not with some magic prompts or 50 pages of prompts, but with structure. So stay with me because by the end of the video, you'll know how you can set up AI so it actually thinks with you and not just respond with you.
Now, here's the shift that most people miss. Good prompts, they're not clever sentences. They are systems. That's my middle name, right? Systems. So the best AI users that I work with, they're not just typing out better words. They're creating better context around everything. And in this video, I'm going to show you the exact traits I see in every strong prompt.
Now, um, these are the same foundations I use when I build the AI workflows. I create custom GPTs and business systems for clients who want speed over all of this slop that we see out there. Now, once you see this, you will stop starting over and over because your AI will remember, it will continue to improve, and it will compound instead of having to reset every time. And then about halfway through, I'm going to show you why the step-by-step prompting is often the thing that's breaking your results and what you need to do instead. And at the end, I will show you how we take this from, oh, that's interesting, to implement it inside of a real business without adding all of this tech chaos in your business.
Okay. Now, the first thing that I want to do, I want to zoom out a little bit and show you the big picture. First, I've got a mind map, and it's going to represent how how these strong prompts are built. Um, not line by line, but by layers, and then as we walk through this, you're going to notice how little of this is about all those clever words, but how much it's about setting up the environment for good thinking.
Okay, so let me share my screen here as always, and hopefully I can get to the right the right uh tab here. Okay, so this is just a little infographic that kind of gives you that 10 10,000 foot view here. So the anatomy of elite AI prompting. What does that mean? The first thing is over here you need to build what we call compounding context around it. The very first mistake that a lot of people make in chat GPT is they don't use projects. And I'll show you at the end how you can set up projects. and they don't give AI the reference files that they need. So because of that, then they're always starting from scratch.
Then you also want to use uh what AI calls extended thinking and search. You want to be able to turn on web searching and that will eliminate uh a lot of the hallucination that uh it sometimes goes down that rabbit hole there. And most importantly here is you want to tell it how you define the success, not the steps. So in other words, what is your final goal and the audience rather than micromanaging uh do this, do this, do this, give it, let it do the thinking and tell it what that end goal is.
Okay, so those are some of the good prompting uh framework that we need to uh put into place. Now what's going to uh create what we call a bad prompt? Well, the first thing is just starting from a blank sheet of paper without any of that reference material. So, you just say, "Uh, give me a good hook for my, uh, Facebook, uh, you know, page or whatever." You're just going to get the same old stuff that you're seeing, uh, everybody else do. Uh, also the vague make it good goals. So, giving AI too much leeway because leeway it AI's whole mission in life is to please you. So, it's just going to go out and grab something, come back like, "Look, mom, see what I did?" Uh, because again, you're not defining what these success steps are. Okay? And then the other thing is what we call contextual bloat. So uh sometimes you talk too much. You try to boss AI around too much. You provide all of this detail that's not necessarily uh relevant for that. Um and so everything just gets super messy.
Okay. So uh we want to give it the right context. We want to build in those constraints, put the boundaries, the quality that we're looking for, and we want to have a conversation with uh chat GPT or AI, whichever tool we're using. Uh when it gives you something, immediately respond back and say, "No, that's not what I was thinking." And give it examples. Here's an example of um YouTube hook u uh headlines that I like and even why you like it. Okay, is that making sense?
Now, uh let's dig into this a little bit more. So, let's talk about these traits. So, the first one is setting up a project and uh in chat GBT, let me hop over here to chatgpt over to the left. Uh and let me say this, if you're using another tool like Gemini, I think they have something called gyms instead of projects. It's the same concept. So, over here to the left, you will want to create a project. So, it's like a folder, but inside of that folder, then that's where you can upload all of the context that's relevant just to this topic, not all of the other millions of things that you've done. Okay? Then you can also get into your settings. I've got some screenshots I'll show you here. Uh, let's see. Do I have it here? Uh, let me see if I can find it. So, if you get into your settings, make sure that you turn on web search. Okay. Uh, I have a video someplace that goes through all of these settings, but you'll want that turned on. And then up here inside of chat GPT, uh, you can have it do deep thinking. So depending on what it is that you're creating, open up that deep thinking so it really goes deep. All right.
So coming back here, uh a good prompt, you're going to start with a project. Now, do I ever just use a one-off chat? Yes. If it's just a one-time thing, I'm never going to come back to it, then that's what I do. But if this is something that I know I'm going to do all the time, so for instance, all of my emails live under a project because it gets to know my tone of voice. It knows different resource documents and so forth. Okay.
Now, when we when we set up that project, uh what we're doing is we don't want a brand new chat every time. If it if it relates to that project, to that theme, we're going to open up that project. We're going to upload all of our reference files, give it whatever custom instructions that we want for that particular project. And what we're doing is we're building this compounding uh context for it. All right.
So, we set up our project. Then, as I mentioned, turn on extended thinking if that makes sense for your uh uh project here because what this will do is it will prioritize reasoning over pattern matching. All of these LLMs are just basically super super trained to look at patterns. When this word is said or in this context, usually this is what happens. If you turn on deep thinking, it will start kind of reasoning. Uh, take it with a grain of of salt, but it's going to go deeper. Okay. When you turn on extended uh thinking, it um allow the backtracking and uncertainty because again, you want to have a conversation. All right? and then um give it very specific instructions uh when you're setting up that project and deep thinking will kind of stick to the grid and do that. Uh, it does take longer. So again depending on what it is you're working on uh it will go deep but it's going to take time to do that.
Then I mentioned make sure you turn on your web search. So, uh, that's the antidote antidote to hallucinations because if it's searching, uh, it won't have to make up stuff. And AI is getting so robust, uh, we're not seeing that as much as we used to. Uh, but again, it will help that. Another thing is when you turn on the web search, it forces citations. So, if there's a citation, then you can click on it and actually check to see what reference they're pulling from, not just making up something. Okay? And then it helps you to verify uh, against reality.
Okay? Upload the reference. We talked about that. So, uh, the more exact examples you can give, that's perfect. Uh, use markdown files. So, markdown files basically, it's kind of a glorified text file. Basically, in the backend like programming um tools out there and stuff, you'll hear people talk about they delivered in a markdown file. So, it just kind of strips out all of the other uh formatting and stuff and just gives the uh, the content that it needs. Okay. Um, when you're uploading the reference, excuse me, the beauty is it's you're showing exactly. So, uh, you're building a website and you love how that hero page, take a picture of it and upload that and say, "This is the kind of hero section that I want built." So, it knows what we're going, you know, it acts like a mind reader, but it's not. And then less is more. So, don't unload all 50 files that you have in your uh, Google Drive here. Just give it what it needs uh, to take care of the job.
Then very important, define success, not the steps. So what does this mean? State the goals. What is the end goal that you're trying to accomplish? What do you want your audience to be able to do or what's that end product that transformation? Make sure that you identify the audience and define the sound and the style. uh I want this presented in a very authoratory authorative uh language or I want this to be uh friendly and engaging and chit chatty type. Okay. And that will help it um deliver what you're actually looking for. Specify constraints. Uh, so that again is when you set boundaries, not rules but boundaries of what you're looking for. And sometimes the boundary might be um this um blog article needs to be between a thousand and 2,000 words. These are the uh keywords that I want to use and so forth. Okay. Uh, again uh define the target in addition to the audience, the target domain. Okay. So um if we're doing research or if we're uh building something, make sure that you give it all of that information so it knows exactly what it is that we're building. brief versus comprehensive quality. Again, um, you know, what are we looking for it to deliver to us? Do we want a one-page executive summary? Uh, do we want a 10-page article that goes three layers deep? Uh, tell it that. And define the exclusions. Basically, we're setting the scope. So, if you've ever done any project management, you're setting the scope of this pro project. This is what we're doing today. Next time we'll we'll get into uh the B part, right? And then the next step, give examples. So, paste good examples. Like I've already uh mentioned before, the more examples you can give it, the better. Quote uh the desired tone that you're trying to get it to um write in and show comparisons because when you say, "Don't give me shallow content." Uh, give it an example. Say, "This is a an example of what I call shallow content. I do not want this." Okay? Again the more explicit you are on the examples then the next step use AI as a critic.
So what do we mean by that? So ask for a draft critique. So a lot of times I will create something myself and then I will upload it. So for instance my emails all my emails are written for me. The the first pass is me thinking brainstorming then I will upload it to my project my email project. So it knows me it knows how I like to write it knows my common CTAs. Okay. Then I will ask it to critique it and then it will come back and diagnose some of the issues it sees and then so do you see how uh we're giving it that foundation. We're not just say saying write my email to my list this week. Uh, it's becoming an evaluator saying um let's tweak it here. Let's make this call to action stronger and so forth.
Then the ne the next framework here chat and correct mistakes. So what do we mean by that? So again, give it immediate error correction just like training your pet. You know, that's what you have to do immediately. If you see that it's uh going down a rabbit hole you don't want, just click the stop button and say, "No, this is not what I want." And again, give it examples of why you don't want that. Uh, use concise messaging. Uh, it doesn't need a dissertation. And then keep pushing it and make sure that you don't have this bloated context. If you are in this uh project back and forth, you know, for an hour, it might really just be because that's how our brain is, right? I mean, we just, oh, what about this and what about this? And we just go down all of these different rabbit holes. So, we might need to stop, take a step back.
Now, that's all of the um the good things. Now, let's talk about 11 signs of a bad prompt. Although, it's not showing 11 here, but I've got another chart here I'll show you. So, again, just a a blank chat. Uh, it's not a project, you're just a white paper. You type something in and think you're going to get something back. Uh, you don't have the thinking mode turned on. You've got very vague goals. One shot prompting. I mean AI will probably get there, but at this point, you know, we have to put some thought into what is it that we're trying to create and then describe the steps over the outcomes.
So I have a little table here, I think someplace. Did I delete it? Oh, let me come back here. So, here's uh the big picture and then here's the little mind map. And then I've got this table here. Let me shrink this down just a little bit. And if you want a copy of the table, just mention uh in the comments below and I can uh I can share it with you. So, here it just is. Can I move this over? Sometimes this is a little finicky. Okay. So, basically uh the prompt, the category, the actionable advice, and then what happens on the negative side. So I'm not going to read all of it, but let me just read the first one so you get the idea. So uh what's the first uh uh framework basically that you need to do? U make sure that you set up projects. Okay. And uh what this does uh it I mean the very first time it's going to be a fresh start, but you're going to be uploading all of these AI reference files. You're going to be building some custom instructions. So then every time you go back to that project, you're not starting over. Now the the negative sign is if you just start a blank project every time then you're killing yourself. You're having to start excuse me over and over and you're just getting a vanilla output. Okay. And then the expected outcome improve continuity and contextual relevance. All right. So that's how you uh look at the little chart there. So let me come back here and hide this.
Okay. So that was a lot of information today. So, if this uh video has helped you to see AI a little differently, um here here's the deal. Put it into practice. Okay? So, as we talked about, it's not uh copying these uh top 50 prompts and stuff. Put this into practice and um make sure that you have the support and the structure and the real implementation. And have to do a little plug for DBA uh digital business academy because we work on this together. We build the projects, the prompts, the workflows, the systems. As I'm always saying, it's all about systems. That's what you need in your business. And then you just play around with it.
So, if you would like to get that live support and help in turning AI into something that actually saves time and grows your business, then uh you should join them. So, if you're ready to stop this prompting randomly and creating just a vanilla slop out there and start building AI tools um that really will grow your business, then uh join us inside of DBA. And uh if you if you want that little uh spreadsheet thing, just comment below and I'll share it with you. And hey, would you do me a favor? Share this video with other people that would benefit and let's all become uh great AI frameworkers, not prompters, but build that uh frame out. And as always, thanks for watching. I'll see you next time. Take care.