323 | Stop Creating AI Slop: Build an AI Content Engine That Sounds Like Your Brand with Brian Piper

0:00Isar Meitis:Hello and welcome to Leveraging AI, the podcast that shares practical ethical ways to leverage AI to improve efficiency, grow your business and advance your career. This is Isar Meitis, your host, and we have an awesome topic for you today. More or less every company in the world right now, and most individuals to be fair, are experimenting with creating content with AI. Now, it doesn't have to be marketing content. It could be emails that you write for work. It could be business documents you're trying to produce. It could be PDFs you're doing for like any kind of created content that you're trying to create.

0:31Isar Meitis:The truth is there's a very, very big gap in quality and the results that you're achieving with that content between those who know how to do it right versus those who just go to Chachi Pitera and say, hey, create this thing for me. Now, the difference comes down to one really simple thing, which is teaching AI about your company, about your goals, about your brand, about your target audience, etc. While this sounds really simple and straightforward. Most people, A, don't know that they need to do it or just don't know how to do it effectively. Either way, they're getting mediocre or even basic results, which is now called AI slop, right?

1:11Isar Meitis:It's just content that you can tell it's AI from a mile away because it kind of looks and feels vanilla versus what your target audience deserves. So in today's episode, we're going to teach you exactly that, how to create a system that will learn basically everything it needs about you, about your organization, about your audience, about the goals you're trying to achieve and so on in order to deliver content that is significantly more specific, effective, and impactful in whatever it is that you're trying to do. Now, our guest today, Brian Piper, has been in content marketing space in more or less every role there is.

1:45Isar Meitis:He worked at large corporations. He worked at small businesses. He was teaching this at university. He worked as a consultant in the field. So basically, he knows a lot about content marketing and how it should be done right. And in the last couple of years, he has dove into the AI space and is now helping companies to use AI to produce content the right way. And it's exactly what he's going to be teaching us today is the process step-by-step on how to work with AI to learn about you, your company, your organization, your brand, your goals, etc. in order to create content. Again, not just marketing content.

2:24Isar Meitis:You can apply this to any kind of business process that you need. Now, since this is something, like I mentioned, more or less every person on the planet with internet connection needs to know right now whether they know they need to know it or not. I am truly excited to welcome Brian to the show. Brian, welcome to Leveraging AI.

2:43Isar Meitis:Well, thank you so much for having me on, sir. I'm really excited. You and I both met speaking at a large conference and we didn't get to see each other talk because we were talking at the same time on two big stages on both sides of a corridor. And so I'm personally very excited about this. I do very similar things. I know you have a lot of experience in doing this. And like I said, I truly believe in what I said. I think the vast majority of people don't know how to do this. Most people don't know they need to do this. And even those who do and done something probably didn't do it as professionally and organized as you.

3:16Isar Meitis:So stage is yours. Please teach us the right way to do this thing. Awesome. Well, thank you so much for the opportunity. And I love talking about this topic. Content marketing is, I believe, the best way to form relationships and create trust with your audiences. But so much of it really goes back to making sure that you have a consistent voice and your brand is consistently aligned with all of the content that you're creating. And so many organizations have multiple different people creating content for all sorts of different audiences they're trying to reach. And, you know, especially in larger corporations, you get a lot of brand fragmentation that causes confusion among your audience.

3:59So what we're going to talk about, and you actually hit it perfectly, you said working with AI to make sure that you're creating the best content. And that's really, you know, the fundamental push that I make for everyone that's doing anything with AI is start working with these tools. Don't just tell it to do things and expect that to work. You really want that back and forth. So let's jump in and talk a little bit about how we're going to make this work. So the first thing I like to do is start with the problem, start with the prompt that you're going to create. Lots of different ways to prompt these tools.

4:34I really like, you know, Jeff Woods was a presenter at Make-On. He He wrote a book called AI Driven Leader, excellent book. And he created this crit framework where basically you give the tool all the context, as much information about the problem you're trying to solve. And then you give it a role. You tell it what part you want it to play. And then you have it interview you. Now, I use this in all sorts of content creation work that I do. I use it in all of my prompting. I have it ask me questions because that's what I want. I want my voice. I want my expertise. I want my knowledge coming across in the content.

5:07I don't want it to go out and grab that from, you know, every generic source on the internet and then put my name on that. So I really want it to pull out my expertise. And then I tell it, you know, use my words, use my phrasing. But having this brand voice knowledge block or this brand voice skill, you know, having your brand voice codified in a way that this is the document, this is the source of truth. When you're creating content or editing content that you're working with me on, always look at this document. Always look at what we define our brand voice as. And a lot of organizations I work with already have like a style guide or a voice and tone guide.

5:49Oftentimes that doesn't actually align with the content that they have out in their digital environment, their digital footprint. So one of the first things I always do is I run an audit. Like what is your content actually saying? What is your brand voice that's already out there in the market? And is that what you want it to be? If not, then we need to make adjustments. But that's a great place to start. So what I'll usually do is I'll start off by doing a brand audit. Now, I have a prompt that I use for this, but I'm going to walk you through the process that I use to create that prompt because I use this on almost every prompt that I create.

6:25So basically what I'm going to do, I use a tool called WhisperFlow. So I talk my prompt into the model. So I'm basically just going to stream of thought following the CRIT framework. I'm going to go in and say, I need you to help me create a prompt that will look at what an organization's brand voice is all about. I want you to look at things like the mission, the vision, the values, look at all the content that they have online. I'll give you some links, but I want you to do an extensive look at all the content that they are creating and then infer from that what their brand voice is. You are an expert brand marketer with a deep understanding of different audiences and what their needs and questions are.

7:07You understand the repercussions of having a disjointed brand voice or a voice that doesn't align with the values of an organization. Before you create a response, ask any questions that are going to provide a more comprehensive and impactful output. And I want the output to be in a markdown file. So I just blasted all of that in there. It throws that into the prompt. And now I'm going to use Claude. I'm going to use Cowork. And I like that because it is able to go access the internet, is able to go use my browser and go out and find all the content around my particular brand. But the first thing I want to do is just create that prompt.

7:46So I'm going to run that. And I actually have already run that. So it'll ask me some questions. It'll come back and say, like, what should the output be? And I'm like, I want it to be a full brand voice guide. What inputs? I'm going to give it some URLs to analyze. I might upload some documents. Where does it need to live? I want this to be portable. I want this to be able to be used on any large language model, not just specific to Claude or Gemini or whatever model you like to use. And then, you know, how should it handle that gap between what an organization claims to have as their brand voice and how it actually sounds.

8:21And I wanted to call those out specifically. So then, you know, it'll, it'll come back, it'll give me some, some things, and then it will create a draft of this prompt that I can use. And basically all I have to give it is the name. So it can be a personal name. It can be an organization's name and then a URL, and then it will go out and find everything it can based on the information you provided. Now, obviously like anything with AI, the more data you can give it, the better, right? The more good examples you have of writing that you know is brand specific. You give it links to your social channels because it can read some social posts.

8:59It can use the browser to read some of those. Any newsletters you have, podcasts, videos, all of that sort of information, any links you can give it to content you have, it can at least go try to pull information from those sources about your brand voice.

9:14Isar Meitis:Yeah, I want to add a few things. Well, I could probably add 50, but there's a few critical things I want to highlight from the things you said, because there's a lot of gold between the stuff that you said. The first thing is, I'll say something for those who are not watching the screen, the output, what Brian was doing, he was prompting AI to build a prompt, right? So he didn't ask AI to do the thing. He asked AI, write me a more detailed prompt than I can. And what you're not seeing is that this prompt is probably two to three pages long-ish. It's a really, really long prompt that you probably cannot write on your own, but it knows how to really build a process, a step-by-step with all the details that it requires in order to achieve the goal that Brian was clearly stating.

9:58Isar Meitis:That's the trick. The value of doing that is that now he has a prompt. he can use anywhere that will deliver, and this is the critical thing, consistent results, right? You want different people using the same prompt, getting the same exact outcome. The other thing that he does, and you're free to test that because I show this when I talk on stages, is when you have a prompt that is that detailed, that is three pages long, you can go to ChatGPT and put it on ChatGPT instant, which by the way, does not cost you any tokens. So you can run it forever. And you can run this prompt on high as far as the effort it's going to run, and you will get very similar results because the prompt is very, very detailed.

10:39Isar Meitis:And again, I do these exercises all the time. Now, when it comes to creating content, it will be a little different. When it comes to analyzing numeric data, you will get exactly the same output because it's analyzing quantitative data. But having a really detailed prompt that you cannot write yourself, but AI can help you write, is a huge deal. The other thing that Brian said that is critical is having AI ask you questions. Brian said very quickly that he's using a process and the first step is giving it context. What the hell is context? Context is telling you as much as possible about the thing you're trying to do.

11:10Isar Meitis:The thing is, we never think about all the things AI needs to know in order to do the thing that it needs to do. And giving AI the permission or even better guidance rules, you must ask me at least X number of questions. You need to ask them one by one. You need to recontextualize after it's questioned to ask the next, like all these things help AI pull out of you the right context that it needs in order to do a better job in the task. So I'll stop now and let you continue. But these are really critical things that I want people to take out of this. Yes, absolutely. Great points. And, you know, another thing I've found, and you've probably also noticed this, that the more you use AI, the more reading you do.

11:49Because now I spend my time reading this prompt, this three-page prompt, to see if this really does give all the information that I need. Is this looking in all the places that I want it to look? Is it doing with the, you know, with the data what I want it to do? And, you know, the nice thing is you can craft this prompt to be specifically what you want. But this is going to have way more information in there than I would have thought about if I would have tried to write it from scratch. I mean, it even goes into details like, you know, look at your website, but then look at all of these possible pages that might exist on your website.

12:23That might be places where the brand voice might live and might be, you know, your about page, obviously, you know, those are places where you're really going to be more specific to your brand and to your voice. So then what I'll do is I will take this prompt and I will go run it in different tools. So I like running the same prompt in multiple different places. So I'll run it in Claude Cowork. I'll run it in ChatGPT Work. I will run it in Gemini Deep Research. And each one of those tools will do this full exhaustive brand audit for me. And each one is a little different. um so you know here are the results from when i just copied and pasted that prompt and then i went into you know add the details so i i said you know you're looking at me i want you to audit my website look at my linkedin look at my newsletter right i gave it links to all of those places where i wanted it to specifically go look and then since this was um in co-work in claude co-work it asked me at each stage do you give me access to go look at this website and then so then it came back with all these findings, right?

13:31And it noticed places where there was a difference in my website, the tone of voice that I use on my website versus the tone of voice I use in my newsletters or on my podcast. My website's a few years old. So there are some things that need to be updated on there. So it's already finding like discrepancies in my brand voice out there that I need to go fix. And it gives me this full, and I usually ask for things as like Word documents and markdown files so that I can use them. It's easier for me to go clean up a Word document. And, you know, then I can use that markdown file for a reference document for other AI prompts and other large language models.

14:09But it gives me this full breakdown of all the different things that it found, all the different places it looked, it rates the tone and the style, the temperature audit for all sorts of different categories of my content, and really helps me look at what I've got out there in the market and say, is this what I want my brand voice to be? Do I want to be formal? Do I want to be more casual? Maybe I should think about adjusting that, or maybe I lean into that because that's what all of my content is currently saying. So then I'm also going to take that same prompt. I'll put that in Gemini and I will run that as a deep research report.

14:44So then it will go and it will look through hundreds of different sources and it will come back and it will also give the same results and it'll break the categories down. in the same way. So now you've got multiple different data sets that all have slight variations in the way that they are doing this brand audit. So whenever I have a situation like that, I will take those and put that into now a Gemini notebook used to be notebook LM. Now it's Gemini notebook. So you can just add those sources into a new notebook and I use the markdown files in these, and then you can ask it questions and you can say, you know, where, what are the discrepancies between these different reports and what are the numbers that are different and what is one report finding that the other one is not finding so now you really get to start narrowing down what do you want your brand voice to be what are you seeing out there in the market so it's really insightful and then of course you know gemini notebook you can do things like have it build you uh you know a mind map so that then you can look at all sorts of you know what are the rules that you, your brand voice needs to, you know, subscribe to like always, I always want to write in the first person.

15:55I always want, you know, the instructionals to be in the second person where I'm saying you, you, so it can break all those down. It can give you individual insights into, you know, how your openers and your sentence flow and all sorts of different things around your brand voice. And for an organization, this is a great resource document that you can create or or a resource that you can create. So if anybody has a question about your brand voice, they can just jump into your notebook and ask it that question. And also, you want to distribute the final document, right? This is a knowledge block that should be shared across your organization so that anybody that's working on content can always dump that PDF file or markdown file, whatever format you want to use, into their prompt at the end and say, does this piece of content align with our brand voice standards?

16:43You can also use it to do like a full audit of your website and say, where does our brand voice drift within our website? Do our different departments or divisions have different voices that need to be more consistent?

16:57Isar Meitis:I want to pause you just for one second for a few things that are important again. One is use multiple tools. You will get slightly different results when you use ChatGPT versus Gemini versus Grok versus Claude. That's just the nature of it. It's not good or bad. It's just different. And they all have their value. Now I know what some of you are saying. Oh, I don't have access to all of those. Then fine. Use the ones you do have access to. And if your organization limits you to just one, then use just one, but know that you can get different variations and flavors because these tools are wired slightly differently.

17:29Isar Meitis:They're going to be roughly the same, but they're going to give you nuances that one will give you that the other won't and vice versa. And looking at the accumulated outcome will give you just better results. Two is going back to why Gemini notebook. Gemini notebook is really good at already pre-canned outputs, meaning you could have taken all these outputs and throw them into any of those things and ask questions. So the back and forth as far as asking questions would have been the same. But Gemini notebook has, I don't know, six or eight different outputs that are set up. They're ready to go.

18:05Isar Meitis:You click a button, you get a PowerPoint presentation, click a button, you get a detailed document, you click a button, you get a mind map, you click a button, you get a podcast, you click a button, you get a video summary explaining like, and you can use all these formats for different use cases inside your organization. And so the value in using those is just knowing what's available on all these different tools to then get the best output with the least amount of effort. That's the whole point of all of this. The other thing that is really cool here is that what Brian is doing is he's looking at internal and external sources of information, right?

18:42Isar Meitis:So you're using existing documents you've written and existing content that you've created, but you're also looking at websites and review sites and what other people are saying about you and social media, because you started with the internal stuff, go check one, two, three, four, but you tell the AI, go find other places to do this. So the review becomes a lot more comprehensive when it comes to actually understanding what is your brand right now, and then you can decide what you actually want it to be if there is any kind of consistency. But this is absolutely fantastic. So again, going back to context, you're now giving AI different kinds of context in order to do this effort and give you better results.

19:25Absolutely. And the nice thing about having this as a reusable prompt that you can use, or, you know, in a moment, we're going to show you how to turn it into a skill, is that not only can you run this on your own brand voice, but you can run it on competitive brand voices and see what the voice and tone is of some of the other organizations in your market. And then you can look at, well, you know, are they going to be the tight buttoned up professional company, and we're going to be the looser, more, you know, friendly, conversational type of company, so that, you know, You're not competing with someone else who has the same kind of look and feel that you do and also talks the same way and has content that's focused on the same audiences.

20:10You can help find your niche in a market by looking and seeing what other opportunities are there out there for adopting a different brand voice.

20:19Isar Meitis:Yeah, I love that. I'll add again my two cents. My content machine actually does both and it does it in real time all the time. So it actually looks at what my competitors, quote unquote, are putting out there on social media. And it's trying to analyze what's working. And how do you know what's working? Not because they're getting more likes, but what is getting more engagement beyond their average. So if their average post is getting 10 ,000 likes, then getting 10 ,000 likes doesn't mean anything. That's what they're always getting. That's the benchmark. The fact that they're getting 10 ,000, you're getting 50, doesn't mean that that post of theirs is better.

20:55Isar Meitis:It just means they have a bigger audience that's following them right now. So it's looking for the signals that all the posts that actually outperform their benchmark and then said, okay, these are unique. And then it tries to analyze how long is the post? What was the tone? Did they use an image? Did they use a video? Did they use a carousel? They're like, what was the actual content? How, what did they write? What was the hook? Like the first line, like all these things. And then it tries to say, okay, this is what's working on LinkedIn right now. This is what's working on Facebook right now, whatever the case may be.

21:27Isar Meitis:And then it goes back to my content, the things that I say on my podcast, the things that I say on my social media posts, the things that I say on my community meetups, like all of that is recorded and it has access to that. And it has access to my tone of voice. And it's like, oh, I suggest we do this. So now you're doing basically, you can take everything that Brian is talking about. So the looking at your competitors, knowing what your brand guidelines are and your own content that you're creating in different ways. And people are like, well, I don't have content. Yes, you do. You have client calls.

21:59Isar Meitis:You explain your product, right? That's content. It's not rigged and tailored to be marketing, but AI knows how to do that very well. But it is still you explaining about your products and services. All you have to do is record those calls that you're having anyway. So you can take all these golden nuggets that Brian is sharing and tie them together to a process. Absolutely. And so, you know, that's really the next step is once you have, you know, your brand voice where you want it. And once you have this prompt that really seems to be working well for you, you've tested it out a couple of times, you've refined it.

22:33So it really gathers exactly what you're looking for out of this brand voice. You know, if you're working in Claude, then there is a super powerful thing you can do called building skills, right? And to build a skill is so easy. Basically, it's a reusable prompt that works with inside Claude's system so that you don't have to go find that prompt every time. You don't have to customize it every time to work the way you want it to work. So within your customized menu, you have these skills and connectors. And if you add a new skill, if you browse the skills, the first skills that Anthropic has like already prepackaged and built for you is the skill creator.

23:16so you install that you just tell it yep activate this right it's it installs it and now all you have to do is say to claude whenever you're done with a prompt that you really like just say create a skill for this voice checker right and then it'll ask you some questions about like well what do you want to do with this skill how do you want it to work where do you you know do you want it to look and expose contradictions at like we did and then so it will build that skill for you It'll build all the supporting files. And then you just have to add that skill into your cloud. So now all you have to say at any point in any prompt that you're working on is say, run a brand voice audit.

23:53And it knows that is the skill that it will need to use. So now you can use that over and over again. And the nice thing about doing this, and you can do this in ChatGPT, used to be custom GPTs. Now it's within skills within ChatGPT work. There's not really a way. I don't know if there's a way to do this in Gemini because I don't usually, I usually use Gemini for a lot of the research, but I don't use it for a lot of the building. But so now what I like to do is I look at, okay, this brand voice audit, I can use it all the time. I can use it on pieces of content. I can use it to go look at my website.