Loading summary
Transcript113 lines
- [00:01]
Michael Stelzner
Hey there, it's Michael Stelzner from Social Media Examiner. Are you a marketer trying to navigate the AI revolution? What if you could create content twice as fast, automate tedious tasks, and become the go to AI expert in your company? The AI Business Society, brought to you by your friends at Social Media examiner, gives you expert led training, a supportive community, and proven frameworks to master AI marketing that will boost your value and your productivity. And here's the big news. Our spring enrollment sale ends Friday, May 16th. Join now and lock in your discounted member pricing@socialmediaexaminer.com AI welcome to the AI.
- [00:52]
Natalie Lambert
Explored podcast, helping you put AI to work.
- [00:57]
Michael Stelzner
And now, here's your host, Michael Stelzner. Hello, hello, hello. Thank you so much for joining me for the AI Explored podcast brought to you by Social Media Examiner. I'm your host, Michael Stelzner, and this is the podcast for marketers, creators and business owners who want to know how to put AI to work. Today I'm going to be joined by Natalie Lambert, and we're going to talk about how to build your own content team of AI staff. And what I mean by that is actual AI people, for lack of better words, not real people, but AI agents that are working for you to help you create your content. And if you've always wanted to have an incredible content team working for you, but you haven't had the budget, this is going to be the episode for you. We're going to talk about some stuff that's going to blow your mind. Also, if you're new to the show, be sure to follow us on whatever app you're listening to us on. And we've got some great content coming your way. And with that, let's transition over to this week's interview with with Natalie Lambert.
- [01:59]
Natalie Lambert
Helping you simplify your AI journey. Here is this week's expert guide.
- [02:06]
Michael Stelzner
Today, I'm very excited to be joined by Natalie Lambert. If you don't know who Natalie is, she is the founder of Gen Edge, a consultancy that helps marketers take advantage of generative AI through customized workshops, training, and consulting. She's formerly the Global Director of Applied AI for Marketing at Google. Natalie, welcome to the show. How you doing today?
- [02:26]
Natalie Lambert
Thank you so much for having me, Michael. I'm really thrilled to be here.
- [02:29]
Michael Stelzner
I'm excited you're here. Today, Natalie and I are going to explore how to build an AI content team. And when I say AI content team, I mean a team of AI staff, for lack of better words, not real people. But before we get there, I want to hear your story. How in the world did you get into AI?
- [02:46]
Natalie Lambert
It starts with, I just have a wide marketing background, was a forest analyst for many years, which I recognize isn't marketing, but it's a lot of those content skills. It's a lot of doing market research and writing reports, writing content. I then spent time at multiple startups where kind of opened the aperture on the different types of marketing tasks and activities people did. And then I spent time at Google kind of going back to my core in content and market research. And it was having that breadth of content and capabilities under my belt that had me wanting to do something new that wasn't quite as focused. And at the same time, there was this amazing opportunity at Google where they were looking for somebody to help figure out how marketers could be best in class using AI across all of the different activities that we did on a daily basis. And I very quickly raised my hand because I felt I had that large, broad look at what marketers do and kind of going, studying AI and how we could apply it to these different activities felt right back in my Forrester analyst days of being research and doing market research. So I had an opportunity to work across Google, 7,000 employees to figure out what were the best tools and best practices that marketers could be using to really improve their impact with AI.
- [04:15]
Michael Stelzner
What year was that?
- [04:15]
Natalie Lambert
2022 and 2023.
- [04:17]
Michael Stelzner
Okay, cool.
- [04:18]
Natalie Lambert
So really starting right after ChatGPT came out and, you know, everyone was trying to figure out how to use it beyond, you know, creating songs, which is what I certainly did. And it was my, certainly my first ask of ChatGPT. It was around that, and it was like, how do we do this? Let's figure out how to look at this across, as I said, all the different tools. And it was just such a incredible opportunity that I felt I wanted to do this on a broader scale. So Google is an incredible company to work for, and they've got a lot of technologies and teams that can help help bring this across the finish line for a lot of the teams. I felt not all organizations had those capabilities. And so the ability to kind of go out, start my own thing and work with companies around the world to figure out how to bring AI to their teams was just really exciting in that moment. So that is when I decided to leave. September 2023 was my last day there, my first day as an entrepreneur helping companies do this at a large scale. So it's been an absolutely incredible journey.
- [05:21]
Michael Stelzner
So I want to ask a couple questions. What was it like to be in this role inside of Google? Were you just helping the marketers inside of Google? I mean like help people understand like what it was like in the early days without obviously disclosing anything you're under NDA that you can't talk about. I'm going to, but I'm curious about what that must have been like for you, you know, in the early days of the ChatGPT moment, if you will.
- [05:43]
Natalie Lambert
It was incredible because I could start anywhere and I guess maybe to say it was incredible in that first few days, like where do I. It was too big. But I go back to what I knew, which was content and really starting to play around with the technologies in that content arena. And I had so much fun doing it. I think the other advantage of taking on that type of role at a company like Google is that I could pick up the phone and call any vendor, I could call any company that I had read who was doing really interesting things and say, hi, I'm Natalie, I'm working on a program inside of Google to bring AI to marketers. Will you talk to me about what you've learned? And everybody said yes. I definitely think that is a luxury that comes from having Google on your current business card, so to speak. It was an incredible opportunity. I think one of the things that were interesting that people don't always talk about is at that time, and actually we still see this today, but at that time I definitely had the people who were the hand raisers who wanted to work with me on a project and the other people who were I hope you fail mentality because success as a threat, a threat, an inhibitor to their jobs. And I will tell you it is the people who hoped I'd fail. And I know that I didn't take it personally, but those were the folks I wanted to work with first because I wanted to show them and experience for myself at that point where the limitations were, where were the strengths and so how we could bring the benefits of human and AI together.
- [07:19]
Michael Stelzner
That's so cool. So now you've been out on your own, it's 2025. So kind of bring us up to the present and share with us the kinds of businesses you work with, the kinds of things that you do just so people can wrap their heads around that wide range.
- [07:34]
Natalie Lambert
I work with companies of all sizes and their marketing teams to say how can you just bring AI into your practices all the time? So kind of put that in the training bucket of, you know, training sessions from two hours to a full day where we start playing with these technologies across a wide variety of areas to say, like, where can we change the workflows within your organization for you to really get the benefit of these technologies? On kind of the smaller side, there are companies who, startups right now who haven't hired marketing, they have a blank slate in terms of what they can hire for right now. They don't have kind of the legacy of the, of the folks that are here and trying to learn. They get to kind of bring it in from, from scratch. And so really working with them on what does the marketing team of the future look like and what types of roles should they be bringing in? Who would be best equipped to work with these tools? And then on kind of that third category is I actually build custom models for teams to help them bring in their tone, their voice, their style, their workflows, their practice, like all of those things into individual models that they can use to truly kind of supercharge their work and improve their impact.
- [08:46]
Michael Stelzner
Love it. And I'm assuming you're platform agnostic at this point, right? Whether It's Claude or ChatGPT or Google, whichever tool is best for the task. I mean, I know Google's making massive innovations right now. I'm just curious what your take is on that.
- [08:59]
Natalie Lambert
I am all over them. I'm currently working with one company on a six week project all on Microsoft Copilot. They're a big copilot shop. Same thing with Google and Gemini and working within those tools. And then there are the companies that aren't either, so to speak. And they have the ability to choose what's right for what they're specifically trying to do. So my training sessions definitely range the gamut of all of them. I have large reasons why I love some, hate the others, which I will keep to myself.
- [09:29]
Michael Stelzner
Yeah, they all have their benefits for sure. So, okay, so for the marketers that are listening or other people that are in business world, you know, let's address this question, right? Why should marketers have teams of AI Personas, bots, specialists, whatever the phrase is, right? But why should they have AI, quote unquote employees when it comes to content? Let's address kind of what the upside is and what, what's possible if this is done right.
- [09:59]
Natalie Lambert
The reason I love this concept of an AI content team or AI specialists or whatever you want to call it is that the way these models work is that you can train them to do one thing and that one thing really well. You can train them to learn about style, tone, voice. You can get them to learn about what does well and what doesn't. And it's just such a great way to use this technology to truly supplement and enhance what we as humans can do by ourselves. And so when you can create this AI team, and when I talk about an AI content team, like, let's get really specific on what I mean here. I mean, you have an AI specialist that helps you with your blogs, you have an AI specialist that helps you with your LinkedIn posts, you have an AI specialist that helps you with your Twitter slash X posts. Notice I'm not saying you have a writer, because when you have a content writer, the way you speak across those channels differs dramatically based on that channel. And so that ability for AI to learn a tone, style and voice and the type of content that gets posted across those different channels is just ripe for creating these individual specialists that can help you across all of the different areas that a content team needs.
- [11:28]
Michael Stelzner
What I find really fascinating about this, especially for the smaller businesses that are listening, most of them would never hire a specialist just to write LinkedIn posts or just to write X posts or just to create an article. They might hire a contractor to do this. But this is the equivalent of having, if I hear you correctly, very inexpensive, very high quality staff member who's always going to be positive, always is going to be consistent, generally speaking, right. And can produce really high quality output for a very low cost. Is that true?
- [12:03]
Natalie Lambert
That's absolutely true. And I would say it can go above and beyond that because, you know, to your point of hiring somebody, we hire writers who can do kind of one thing well, and we hire social folks who can do social well. But imagine having the data down to a particular post that outperform everything else and being able to do that on the blog and being able to do that and just do data analysis that really no individual can do at a level of scale that we're talking about. And being able to train each of those specialists to just outperform in their respective areas so you truly have the best in class of folks that can help you across all of these things. And my goal when I talk about specialists is I want them to get you to 80%. They might get you slightly more, but if they can get you to 80%, and so you as the human can review that content, make sure it fits your style and your message and all of the things that you will do to make it great, that to me is a success.
- [13:11]
Michael Stelzner
And I really love this because what I'm hearing you say is, hey, you could have a full time content person that works inside your company and their job will be to take it the last 20%. Right. Or the last mile or whatever, whatever, right to completion. And they'll be getting stuff that's pretty close to good. And they're going to take it from good to great. And that's where you can have now all of a sudden somebody in, in house. Maybe it's half their job, maybe it's all their job depending on the amount of content you produce. But that's kind of exciting.
- [13:39]
Natalie Lambert
They're managing their team of specialists. Like they have to be a leader. They have to be able to be clear about what they want across all of these pieces of content. I mean, they have to feed their team.
- [13:50]
Michael Stelzner
Love it.
- [13:51]
Natalie Lambert
Just like you would if you have individuals. But now their role, their job, their goals, their outcomes should be based on posting amazing content. Like don't wrap your role in how much you wrote today. Take that out of what your objectives are. Your objectives are creating incredible content that your customers want to read. And so at that point it doesn't matter how it's created, so to speak, but they have to make sure what gets posted is fantastic.
- [14:19]
Michael Stelzner
I love it. So let's assume that we all want to hire a virtual team of AI team, whatever we call it, of AI specialists. AI content specialists. What are the things that we need to be processing or thinking about before we do anything else?
- [14:36]
Natalie Lambert
I think you have to think about what content you're looking for. And I kind of think about this as like, what are the roles you want to hire for? Right. You know, are you do, what are the different types of content that is responsible under this, this one leader? Is it blogs? Is it thought leadership? Is it email nurture programs? Is it bylines? Like whatever you're either hiring for or probably better put, what are the individual micro tasks that each person would do? That's how I would start. Because the more granular you can get with who you want to hire, that is how granular you can get with the specialists that you train. And you know, just, I want to put this into a bit of color. Let's just take social as an example. Because the type of content that a person would post on LinkedIn is very different than the type of post they would do on Twitter. X.
- [15:36]
Michael Stelzner
Right.
- [15:37]
Natalie Lambert
Or on Facebook. Like they're, they're different. Typically the audience is different. But even if it's not, there are character differences, there's stylistic differences. You know, someone who's, you don't want to train A model to do both, train them to be amazing in the channel that they're posting to. So this is why I talk about that micro task. So break up kind of, you have the roles you want to hire, the tasks they're going to do and get those into those models, micro tasks. And those are the examples or those are the types of specialists you can create when you can get that specific and you can even kind of take that level of granularity, that specificity, that specialist, and give them their own knowledge, you know? And so now not only are all of these specialists trained to do one thing well, but they're trained with a set of knowledge that is unique to the channel which they're posting on.
- [16:32]
Michael Stelzner
Okay, I love this and I want to break down these micro tasks a little bit more because maybe hypothetically, someone listening has never done this before. Maybe they don't know what actually micro tasks might be like. Let's just take hypothetically LinkedIn. Okay, let's talk about LinkedIn. If we were to build this quote, unquote, job description, as we're going to call it, or this role, how do we discern what those micro tasks are if we've never actually processed this before? Can AI help us with this? Or do we need to like reverse engineer some existing stuff like how do we know what to put in these micro tasks? That's the part I'm. And by the way, I write on LinkedIn all the time. I've just never thought about it in micro tasks.
- [17:16]
Natalie Lambert
I mean, just so you know, I have a hundred specialists of my micro tasks. So like I break things into them, the micro, micro. Because you can train them really well. What I would say, if you're just starting, you can start big. And by big, I still mean just LinkedIn. I don't mean social, but what you can, if you're thinking about kind of just getting started, LinkedIn is your platform. Think through or look through the most recent LinkedIn post you've done. Okay, what is the tone? And you can ask AI, copy them all in there and say what is the tone, style and voice that you see in these posts? And write this up as a style guide for me. I mean, you can start as simple as there. You can then take it a step further and go through your analytics and figure out what social posts have done the best. Maybe just pull those together and ask it to analyze it. Maybe then do the same thing for your worst performing social posts.
- [18:12]
Michael Stelzner
Oh, okay.
- [18:12]
Natalie Lambert
Which ones have performed badly and have IT analyze them so you can start to slowly. While at first you might just have the generic replicate this voice, style and tone, but then you can start adding more information around what works well and what doesn't work well. Things that as you get start adding more and more components to, you can start getting go back to the really micro. So for example, when I am thinking about posting on LinkedIn and I want to post it based on an article that either I've written or I've seen that's very interesting, I might have a different micro task that says read this article, what are the three most controversial things in it and then use those to be the topic starters for the individual posts. And so now you've gone from creating LinkedIn posts that are based on tone, style and voice as well as kind of content that performs well and content that doesn't perform well to creating specialists that help you decide what you actually post there based on another type of specialist that you train. So you can see like as you start to use them, you're going to realize these small things that you do that you might not have even thought were things but you realize are and you can start to create kind of additional specialists to help you with those tasks.
- [19:39]
Michael Stelzner
You know, I love this, I was telling you before we hit start that I just got back from Social media marketing world and if I start to think about all the things one can do when they're posting content on LinkedIn, there's a lot, you know, you could actually like you said, you could talk about some news that you're excited about. So for example, in the AI world there's a lot of people that when new features come out, they do little videos about the features as a post on LinkedIn and that can be written and or video, right? Or if there's a controversial take that someone has, maybe saying AI is a threat, you could use AI to identify and have a counter take to it or an agreement take to it. Right? Or if you just want to kind of journey your like I do, I write about all the things I'm working on related to the conference. That's another kind of post where it's storytelling, right? So there's so many different kinds of content. And what I love about what you're saying is you could have an AI specialist that's just helping you discern angles to take on a controversial article or another one to help you come up with talking points that you might do in a video. I mean like in the list goes on and on and on. So when we're identifying these micro tasks, that's where we have to kind of reverse engineer the way we normally create content. Is the idea here that we're going to just document all this in the beginning and it's all going to be fought for. Creating unique roles, is that kind of where we're going with this?
- [21:02]
Natalie Lambert
I think we might get there. I see it right now is the opposite. We kind of build for the broad and we get narrow. And so yes, maybe once you've done this for LinkedIn and you've realized your one little LinkedIn specialist has now become 10, making up a number depending on the situation, you know, maybe you've documented that. And so the next time you go create like a blog specialist or an email campaign or event, email, what have you, you have all of these ideas there that you can build that out from the start. But yes, I mean, I will literally sit down and think through every step and I miss steps, don't get me wrong. But I'll think through the steps so that those are the ones that I am creating. And oftentimes I'll be using one of my specialists and realize that I asked the same question or how did it do something, something? And it's like the third time I've asked it to do that after using them like that needs to be a specialist. Like the fact that I'm always asking it to do X. I love that I should make X its own thing. It's just one of the great benefits of using AI and specifically GPTs to do this.
- [22:12]
Michael Stelzner
Let's talk about what's keeping most marketers from truly mastering AI. It's for sure not a lack of interest. Instead its reliable guidance and structured learning. The AI Business Society solves this with monthly live training and virtual meetups with marketing experts and peers who are actually applying AI to their daily work. Member John Marie Pearson said, quote, I'm so excited to be part of the AI Society and see how I can learn to embrace the world of AI and have a community to help me stay up to date with everything going on in the AI world. I'm ready for it. Here's what's happening on the inside. Our members are creating persuasive sales pages in half the time, generating stunning visuals without any design skills, and developing AI enhanced strategies that get real results. Spring Enrollment closes Friday, May 16. Lock in your discounted rate at socialmediaexaminer.com AI your AI transformation starts now. So at this point in the process, we have sat down, we've identified some AI roles, for lack of better words, and the tasks that are typically performed by that person, that role. And maybe we started with just a few things, but over time we get more granular. But let's say we've identified three different roles that we want to create. Okay, so the next step is to build something and you kind of hinted at it a little bit. So what do we do with that information once we've identified these roles and tasks?
- [23:56]
Natalie Lambert
So this is where you go build, you actually go build that specialist. There are a bunch of tools that offer the capabilities to build these. So ChatGPT has their GPTs, Google has their gems, Copilot has their agents, PO has their bots. All of these are that idea of codifying a set of knowledge, a set of instructions into a bot specialist, whatever you want to call it, to repeatably do something. Personally, I like chatgpts. The reason I like GPT specifically and.
- [24:33]
Michael Stelzner
You'Re talking about custom GPT is right?
- [24:34]
Natalie Lambert
Is that what I am? I'm talking about GPT specifically?
- [24:37]
Michael Stelzner
Yep.
- [24:38]
Natalie Lambert
And the reason being is unlike some of the other tools, copilot agents do this as well. You can ask it to do something. So, you know, maybe. Let's go back to my example. I have an article and I want my controversial GPT to come in and pick some of the three most controversial statements in it. And maybe it, it provides me three points of view to have on it as an option. Once I have that and it provides that output, I can simply call in by using the at symbol, the next GPT. And so start imagining this at scale. Being able to ask a question with one GPT that AT symbol, call in the next one that does the next step. Next step, next step, step. I mean, this is where you start getting into agents. This is not full automation because you are having to call agency each agent. But at that point you're getting to a place where you can call 10 different specialists into a conversation in less than a minute. And you get this entire workflow done point by point in a way that you've wanted it to, you've trained it to behave, and you have a result that I would argue is 80, 90% there, but 500% better than what you would have gotten had you just asked the AI to write a LinkedIn post on X.
- [26:02]
Michael Stelzner
So just. And I've not done this before, but my understanding is that when you create custom GPTs, I've done one custom GPT, but I've never tagged it. When you put the sign in there, is it pulling up the name of the GPT based on how you named it.
- [26:16]
Natalie Lambert
Correct.
- [26:17]
Michael Stelzner
Okay, so you've got a thread open in ChatGPT, right?
- [26:20]
Natalie Lambert
Correct.
- [26:20]
Michael Stelzner
And the idea is that you can you put the tag in there and then you put the, the text in there and somehow it knows to use that custom GPT to do the action. But then in that same thread, the answer can be brought into the next GPT.
- [26:35]
Natalie Lambert
It's the next.
- [26:36]
Michael Stelzner
Got it. So. So this is the magic of. It's almost like you're assigning manually tasks to the different GPTs, but it is all the prior context all like transferred to each GPT.
- [26:47]
Natalie Lambert
It is.
- [26:47]
Michael Stelzner
That's magical.
- [26:48]
Natalie Lambert
Absolutely incredible. And while I name them all and I always try to give them funny, I tell them kind of what their job is and all of that as their name. But I was actually watching a webinar. This is what actually gave me the idea that a little over a year ago, and it was someone in media and a movie production studio and he had told me his process and he was explaining it on the webinar and he just named them step one, step two, step three.
- [27:14]
Michael Stelzner
Oh, really?
- [27:15]
Natalie Lambert
Okay, that crazy because he did the same exact steps every single day. And so he would literally wake up in the morning, go onto his computer step one and just call them individually one by one to work his workflow. Now, you know, for those of us who do a lot of different things that probably won't work as, as cleanly, but it's the same idea. I have the Natalie outliner versus the Natalie first draft versus the Natalie final draft versus the copy editor. Like each of these in turn being called to help me get to that final result.
- [27:49]
Michael Stelzner
Okay, so for heavy ChatGPT users like me, and I'm sure you, Natalie, there's a lot of different models that you can pick from. Right. As of Today, you've got 4.5, you've got all these different. 0, 1, 0, whatever. So does it matter which of these models we're using when we're pulling in these custom GPTs? What's your thoughts on that?
- [28:09]
Natalie Lambert
You don't get to say, I guess is my better point.
- [28:12]
Michael Stelzner
Oh, I see. So basically it's going to use whatever the correct.
- [28:16]
Natalie Lambert
And I think right now it's four. Zero is the main model.
- [28:19]
Michael Stelzner
Got it. So it doesn't matter what thread you're in, it's going to default to whatever the standard is.
- [28:23]
Natalie Lambert
It's going to go to what the default your GPT has been programmed.
- [28:26]
Michael Stelzner
Ah, okay. That's really important for people to understand. Okay, that's perfect. That's really good to know. So that makes sense. You're right. I had forgotten that. So once you create a custom GPT, it's just going to go with whatever the default is. Right now it's 4.0. It's probably going to be 4 or 5 by the time this comes out. Who knows, right? But it's constantly changing.
- [28:44]
Natalie Lambert
It is. But one of the interesting thing is to your point though, about being able to mix match. Like imagine you've done something with deep research, okay. And that is its own, its own model. You've asked some questions, you've built out a research report, it creates this 20 page thing. You could then create a GPT that is designed to read through that and find the things that are most interesting to you based on you've trained it on what you want. So now you went from a deep research model to. You're still leveraging the model that the GPT is based on, but it can exist within that thread.
- [29:23]
Michael Stelzner
Love it. Okay, so when we're creating custom GPTs, let's talk about, we're going to call them content specialists, right? When we're creating these content specialists, which are custom GPTs, in this example, let's talk about like some tips on how to increase the likelihood you get from it, what you're hoping, right? Because obviously my understanding is the more you give it, the greater likelihood you're going to get something good out of it. Is that, I mean, as far as like what we're about to talk about.
- [29:53]
Natalie Lambert
Yeah, I think that's totally fair. So one of the things that I'll do when I am doing a kind of style, tone and voice, I will say I will generally upload maybe 10 different blogs that I've written or 10 different LinkedIn posts, bucket the content by channel by the specialist you want to create. So don't put like LinkedIn posts together with emails together with blogs. If you're trying to create a blog.
- [30:19]
Michael Stelzner
How are you uploading those as a PDF or something?
- [30:21]
Natalie Lambert
You can do Google Doc, Microsoft Word, PDF matter. So I just go in, I open up Google Doc for me and I'll just write blog one, Paste blog two, Paste Blog three. So they're separated by headers so it knows where the delineation is and then upload them. You can literally through ChatGPT, add content or directly integrate it with OneDrive or Google Drive and just select that document and I'll literally say to it like you are a writing analyzer or blog analyzer or, you know, whatever content it is and Your job is to analyze the style, tone and voice. You know, what do you see stylistically? How are sentences kind of, what is the overall feeling people get? Like, I'll ask a bunch of questions that try to dig out how that writing is perceived and how it's written. I will also say, do not make a note of the content of this, the topics, because I don't care about the topics. I'm trying to create a style.
- [31:24]
Michael Stelzner
Okay, that's good. So you tell it to ignore the topics.
- [31:26]
Natalie Lambert
Okay, ignore the topics. And yeah, ignore that. But really analyze and write up a style guide of what you find in this.
- [31:36]
Michael Stelzner
Now, real quick question. Are we doing this in the custom GPT or are we doing this in a regular ChatGPT thread at this point?
- [31:42]
Natalie Lambert
Regular ChatGPT.
- [31:42]
Michael Stelzner
Okay, good. Keep.
- [31:43]
Natalie Lambert
And so I write that in regular chatgpt, have it kind of go through and it will write a style guide. And that's when I'll come in over the top and be either I disagree with this, please change this, or please add this. Things that really matter to me are from stylistic standpoint, like I like the Oxford comma, I like sentence case.
- [32:05]
Michael Stelzner
You could take this into a Google Doc and add that context, I would imagine if you wanted to. Right?
- [32:09]
Natalie Lambert
You could totally do that. Yeah. So you're just, you're having it write up this style guide on everything it finds. Once you're happy with it, I just literally copy it and that's where I go to the, you know, top right corner. My GPT is created. Chat GPT and I'll name it Natalie's blog assistant. LinkedIn assistant. Whatever your content type that you had it analyze. And just say you are a writing assistant for blogs for what have you. Here is how you should write all content when asked. And I would just paste those instructions in there and then you save it and then it becomes something that. When you're going back to that example where we had some piece of content that we want to turn into a LinkedIn post, it has the content, but now when it's writing it, it is going to follow those style guidelines and it's going to follow the writing guidelines that were put together in the, in the guide. That's how you get from kind of the more generic content to having it write something that is more personalized to what you're looking for.
- [33:14]
Michael Stelzner
What I like a lot about this approach is it seems as if you are using chat GPT standard and at this point, I guess are using the reasoning models at this point, or does it really matter. It doesn't matter. Okay, but you're giving it examples of your best work and presumably also giving it examples of work you do not like because we talked about that earlier or are you not doing that at this point?
- [33:38]
Natalie Lambert
In this case I'm not, because this where I'm doing a style guide. In this particular case, I want it to focus on just the style, focus on the good. I'll put more of the examples of things that I don't like in the kind of more granular. Maybe it's a reviewer of a Twitter post, like make it like this, not like this. And that's. Those are the instructions like that can those can be in some of the final checks or it's all going to be personal preference at the end of the day and how it works for you. But my style guide is strictly dues.
- [34:12]
Michael Stelzner
Now when we were prepping, you had a tip about not attaching examples into custom GPTs. And tell us a little bit about like the pros and cons of that.
- [34:21]
Natalie Lambert
Yes. So when I first started playing around and building GPTs and for anybody who has, you'll see that idea of adding a knowledge base. And for me I was like, ooh, I can add the knowledge base of all these blogs I've written in the past and tell it to reference them when finding my style. And that worked well for maybe a month and a half. The problem with it came as the market shifted. So a really kind of ridiculous example, but was so powerful in me noticing this problem was at some point Google changed Google Bard to Gemini. So it was a name change. Unfortunately, all of my knowledge base articles, because they were written prior to that name change, all referenced Google Bard for the technology. And so every time it would write something for me, it would always talk about Google Bard. And I couldn't understand why, even though I would write Gemini or you know, I'd given an article about Gemini and I'd want to have it refer to that, it would always go back to Google Bard. And so it was kind of that and then a couple other examples that were very similar that made me realize I really have to abstract the content from the instructions. And that's where I really started focusing on building that comprehensive guide. And you know, I made it somewhat simple for the sake of this conversation around just like the stone, the style, tone and voice. But I mean my prompt to get a style is like two pages of the things to look for. So that when it builds this style guide and that's a two or three page guide of how to write. Right. Like Natalie, so that it genuinely comes off the way I want to is going to be perfect. No, but 80%, yes.
- [36:18]
Michael Stelzner
Interesting. So when we are creating custom GPT is if we've previously done this and we've attached the knowledge base data, what I'm hearing you say is it's going to try to stringently follow the data that's in the knowledge base. And if you go back into that knowledge base, is it kind of difficult to edit that data or update that data? I mean, you don't really see it, so you don't even know where it is. Right.
- [36:38]
Natalie Lambert
You can't, per se. It is a snapshot in time when you have a knowledge base.
- [36:43]
Michael Stelzner
So you almost have to copy it and start again or whatever.
- [36:45]
Natalie Lambert
You would start again or you'd upload a new one.
- [36:47]
Michael Stelzner
Oh, that's interesting.
- [36:48]
Natalie Lambert
But I'll tell you, like, where knowledge bases are really good is, you know, let's say you're a product marketer and you're getting ready to do a product launch, and that launch is going to basically be the product for the next year, putting all of the messaging for that product. If this GPT's job is to write product content, knowing how to stay on message and the positioning of the product is a very fair knowledge base to upload because you do want it referencing that all the time. It's just things that change. And yes, I mean, arguably product information could change too, but that's probably used enough for at least six months or so that you would just duplicate, just switch out the knowledge base when there's a product update to make sure that that knowledge always stays fresh.
- [37:38]
Michael Stelzner
Love it. Okay, so at this point, if we've been following everything we've been learning from you, we have created a bunch of custom GPTs that are essentially these content specialists doing all these various tasks. And ideally, we're creating a thread in ChatGPT and we're bringing them into the conversation. We're moving it down the pipeline of the things that we need to do. Let's now talk about, like, next, leveling this with Personas, interactive Personas, I think, is really what we're talking about here. Right. So talk to me about this.
- [38:06]
Natalie Lambert
This is where I believe everything I've talked about a great marketer can do. Might take a while, no doubt, but if you are skilled in your channels and you know your content, you can do everything we've just talked about. AI will help you do it faster, but you can do all that when you start to layer on Personas. I genuinely believe this is where everything gets better. And the reason is, is that today the number of content folks out there, I was definitely one of the many times you write a white paper, you write a blog, you write something and leadership will come in and say, well, have you shared this with a customer? What do they think? It is not that easy to have customers like, do you want to review this blog? Do you want to review this white paper? And the number of times it's like, well, go build a customer advisory board for content. Great idea. If you have the time and the resources, like, power to you. The reality is the vast majority of marketing organizations do not have the time, the resources to pull together that type of program. And so let's do it then, but let's do it through AI. So this is where you can start to build out Personas that to some extent mirror your audience. I think perplexity is a great tool here. Google Deep research is a really good tool. So is chatgpt Deep Research. But this is where you can go ask AI to say, build me a Persona of, let's take me. I sell to marketing leaders. I go talk to marketing leaders to convince them that they really need to be upskilling their teams with AI. So build me a Persona of a marketing leader and I'll probably start that generic and all of the tools do it. It's hilarious. They will actually, like, meet Jane, not whatever, like, they name them and they build out a Persona. They'll typically include, like, how much schooling, where they might have gone to school, what size company they work for, and it will build that all out. It will say in my prompt, I'll also say things like, what keeps them up at night, what are their goals, you know, what are their aspirations like, and have it build each of those out by section. And let's say it builds me a Persona of all retail. This person works at a retail company. I can edit and say, no, can you make this more focused on B2B? Like, get that Persona right. Have the conversation back and forth to build a Persona that feels almost right to you. When I started doing this, initially, it didn't always put in commentary about AI. So I would literally ask, what does this marketing leader feel about AI? And it would then start to answer questions around, want to include it. They're being pushed, pressured to do it. They don't know how to use their tune, they're worried about job security, all of those things. So get it. Follow up, ask questions, get it to a point where you feel Comfortable. And then I'll typically say, can you just summarize this into one clean Persona? Once it does that for me, I copy and paste it. And this is where we go back to creating a GPT. Go back to create a GPT and I'll name this Persona. So my Persona's name is Genevieve Gen Edge. I don't know, I thought it was fun. And so Genevieve is my marketing leader and so I name her Genevieve and it says my marketing leader Persona. And I will in the description just say, you are a marketing leader that follows this guidance. And I just paste that in and you hit save.
- [41:36]
Michael Stelzner
So when we're prepping the Persona, there's a really good chance a lot of people in my audience already have some sense of who their ideal customer is. Right. So are we just giving AI like a little bit of guidance? For example, I know for me, for social media marketing world, my average customer is likely a woman 35 years and over who works at a small business in likely a marketing role who has X number of years, you know, maybe a decade's experience with marketing and is relatively new to AI. Would I like give that to that and then would it make it more enhanced? Is that the idea?
- [42:09]
Natalie Lambert
Yes.
- [42:09]
Michael Stelzner
Okay.
- [42:09]
Natalie Lambert
And that's when you can add the like, what keeps you up at night? What are your goals? What's the water cooler talk? Where do you go to get learn information? I mean like ask all the questions.
- [42:19]
Michael Stelzner
So the AI will just make it up kind of based on what it knows about someone. Given what little it's given you, it.
- [42:24]
Natalie Lambert
Will make it up based on whatever piece of information. So the more it will be more specific. I see as just an interesting aside, I have done this exercise in many of my trainings and one of the trainings that was most interesting was it was in the public sector. And it was a big training that include kind of marketers within a big public company, but also their agencies. And a lot of the agencies they employed had people who used to be in these public sector jobs to really help make that content more relevant. And so we did this exercise of building Personas and I just sat there and watched some of these folks go. This literally reads like my back, my resume. Like this was the role that I did in my background when I came in. So it is amazing at how close they can get. But yes, in short, based on whatever information it will give, it'll make it more robust.
- [43:16]
Michael Stelzner
And are we using the deep research to kind of make it more robust when we ask it to do it like you know how you have that little option or does the AI smart enough, it doesn't even need the deep research to kind of come up with that more descriptive Persona.
- [43:27]
Natalie Lambert
I would use the more detailed one personally, the deep research. With deep research, I will tell you, Perplexity here is my preferred tool. I think that the deep research capabilities in ChatGPT and in Google Gemini are fantastic for doing genuine. When I want 20 pages of market research on a topic or to go look at competitors across all of these different spaces. Perplexity, it just has a perfect level of information.
- [43:56]
Michael Stelzner
It's just concise. Huh?
- [43:57]
Natalie Lambert
Okay, it's concise but complete.
- [43:59]
Michael Stelzner
I love it.
- [44:00]
Natalie Lambert
I don't need a 40 page. You know, maybe if you're doing something really specific, you care about that and in which case use the 40 page upload that as, as knowledge. But I find the perplexity length is always on target.
- [44:14]
Michael Stelzner
So we take this output that we feel is really accurate description of this Persona. We're creating a custom GPT with this and we're, we're telling it to act as if it is this person. Then how do we connect that with the other tasks, maybe to make the magic sauce stuff happen?
- [44:31]
Natalie Lambert
So then what you do is remember, we got here because we said, okay, now our content is complete. And I said, this is what a good marketer can do. AI can do it faster. That last piece of text is your blog, is your social post, is your email, whatever it is. And now just like we called in different specialists on the content side, we can now call in our Persona specialist. So I can call in Genevieve and say, please review this blog, social post, email, what have you and tell me what you don't agree with. Tell me with what's confusing. Tell me, you know, where you'd want to learn more information, like ask all those questions. And I'll always kind of end with a, you're not going to hurt my feelings. Be harsh. Like, tell me, be honest, because I do find AI wants to make you happy. And so it'll then go through and tell you what it likes and the areas that aren't clear or, you know, disagree with all of those things that I said. And I find I tend to take about 70%, not 100, but it will find like acronyms I didn't explain. It'll say like, this would be a lot stronger if you had an example because it's not clear. It'll say things like, you know, you're using, I want to say jargon, but I don't mean that in the traditional sense, like you're using words. Too much of it simplified. Yes, exactly. And you. It will go through and tell you all that. And then one of my favorite questions that I always ask it to do is, you know, what do you wish you knew more after reading this? And let's say it gets you three or five ideas. Those are your next blog post ideas or LinkedIn post idea, whatever. Like, let it help you create that what's next content too, because it has read it and can provide feedback on kind of what it wants to learn more about.
- [46:22]
Michael Stelzner
Do you go back to your AI team, if you will, and ask them to make the fixes based on the feedback you get from the Persona, or do you tend to do that yourself.
- [46:30]
Natalie Lambert
As 1 million percent?
- [46:33]
Michael Stelzner
Okay, cool.
- [46:34]
Natalie Lambert
A lot of times I'll say, let's go through this one by one. And I will at the end say like, I like 1, 3, 4, 5, 7. Like do those because I do not take it off.
- [46:45]
Michael Stelzner
I love that. And then it sounds like you could go back to your Persona again and say, what do you think now? I mean, if you really wanted to.
- [46:51]
Natalie Lambert
Yeah. To that end, though, it's really crazy is I have built a lot of content going through this process many times, and I was actually doing a training session and it was one of the ones where I. It's probably the 15th time I've done it and I. I'm constantly kind of growing it on ones that have already been done by the Persona. And I'm in this training and it was basically like, this document's great. Here's one piece of feedback. And I had to tell people, I promise it usually gives a lot more feedback. It has just provided feedback like 15 times. I think it's kind of perfect now. But yes, it'll constantly kind of get better and better and provide more and more and help get you in the right direction.
- [47:28]
Michael Stelzner
Natalie, this has been absolutely amazing. If people want to connect with you on the socials, where do you want to send them? And if they want to learn more about your business, where do you want to send them?
- [47:35]
Natalie Lambert
Awesome. Well, first and foremost, find me on LinkedIn. I'm sure Michael will post my profile there. I offer a newsletter there, so please don't hesitate to take a look there. Gen Edge Dot is my website and I also offer a maven course that is focused on building your AI team. So I'd love to see you on any one of those places.
- [47:54]
Michael Stelzner
Natalie, thank you so much for joining us today.
- [47:56]
Natalie Lambert
Thank you.
- [47:57]
Michael Stelzner
Hey, if you missed anything, we took all the notes for you over@social mediaexaminer.com a52 and do follow this show on your favorite podcast app. And if you've been a longtime listener, would you give us a review and or let your friends know about the show? Do check out our other shows, the Social Media Marketing Podcast and the Social Media Marketing Talk Show. This brings us to the end of the AI Explored Podcast. I'm your host, Michael Stelzner. I'll be back with you next week. I hope you make the best out of your day and may AI help you to become more successful.
- [48:31]
Natalie Lambert
The AI Explored Podcast is a production of Social Media Examiner.
- [48:38]
Michael Stelzner
Just a quick reminder before you go. If you're ready to become indispensable in the age of AI, the AI Business Society is your solution. But spring Enrollment ends Friday, May 16th. Join now and secure your discounted membership by visiting social mediaexaminer.com AI I can't wait to see you inside the AI Business Society.