Building Custom GPTs: Personalizing an AI Assistant

Happy 2026. Hey, how's this for a New Year's resolution? How about you Master AI marketing? The AI Business Society gives you live training with experts who are getting real results. No hype, no fluff, just proven workflows that work. And you've got access to an incredible library of training from many of the incredible guests that you've heard on this show. So secure your AI transformation by visiting socialmediaexaminer.com AI.

Welcome to the AI Explored podcast, helping you put AI to work. 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 use AI. Folks, I love that I get to do this show and today. I think you're going to find today's show absolutely fascinating. I'm going to interview Brian Piper and we're going to explore how to build custom GPTs, which are basically personalized assistants that can do all sorts of things that are going to blow your mind. You are definitely going to want to listen to all of today's episode because we get into examples and we also get into, like, here's exactly how to do it. I think you're going to love it. By the way, I'm at Stelzner on Facebook and at Mike underscore Stelzner on X slash Twitter if you want to connect with me or share it with friends and tag me. And then also, if you're new to listening to this podcast, be sure to follow this show on your app, whatever app you're listening to me on, by clicking that follow button so you don't miss any of the great content we've got coming your way. Let's transition over this week's interview with.

Brian Piper, helping you simplify your AI journey. Here is this week's expert guide.

Today, I'm very excited to be joined by Brian Piper. If you don't know who Brian is, he is the director of Content Strategy at the University of Rochester. He's also the co author of Epic Content Marketing. He's also the author of Epic Content Marketing for Higher Ed. Brian, welcome to the show. How you doing today?

Doing great, Mike. Thanks so much for having me on.

I'm super excited to have you here. Today, Brian and I are going to explore how to create custom GPTs that will perform advanced tasks to help you in your work and also in Your life now. Brian, before we get into this, I would love to hear a little bit of your backstory. How'd you get into AI? I know that there's a journey there. So tell us, how in the world did you get into this new frontier?

So I started off as a website developer back in 1996, creating online learning courses and websites. So I've been doing content optimization ever since then. I worked at a marketing agency for a while leading their web team. Worked for a defense contracting company doing training for them. And I've always been fascinated by how technology can help us solve problems and make the most out of our time. I read the first edition of Epic Content marketing back in 2013 and immediately knew that I wanted to get out of website development and get into content marketing. I went down to the marketing department. I told the VP there that she needed to hire me to be the digital marketing manager and she did that. And ever since then I've been just, you know, passionate about content marketing and I knew that AI was going to have some sort of impact on how we create content and how we we optimize content. So I've read all sorts of books, including several from your previous guests, Chris Penn and Paul Raitzer. And so always wondering how that technology was eventually going to get integrated. Back in early 2022, once mid journey and Dolly came out, I started playing around with those kind of text to image tools. And then as soon as ChatGPT came out in November 2022, I got an account and just started diving in and it's been drinking from a fire hose ever since then.

So you come from a background of content. We have known each other, we've come across each other at various different conferences. And in the content marketing world, you're very well known, I believe, and I'm just curious. Like the AI frontier. When you started to see the emergence of ChatGPT as a creator of content, was your first reaction one of the fear or uncertainty or doubt that this could be a challenge to content creators, or did you have a different perspective?

Yeah, I always saw it as opportunity. I mean, these tools are, you know, really no different than, you know, calculators to help you do math or, you know, they're just ways to make you more productive, to basically take away all the things that you don't like doing anyway. And that's the way that I look at, you know, AI. It helps me brainstorm and it helps me do a lot of the repetitive tasks and the data for focused tasks that I don't Want to spend my time doing all of those repetitive tasks. I want to do more of the strategy, more of the creative parts.

So I saw you speaking at an event recently, specifically on AI, and what I'm curious about is how did you get into what we're about to talk about today? Like, what was the journey from messing around with Chat GPT and different image generators to ultimately going really far down that rabbit hole?

Yeah, I mean, I'd been playing around with different ways to kind of know, use Chat GPT to, to help with brainstorming. And I, I was writing these prompts and they kept getting longer and longer and then, you know, I kept putting more and more information in there because really, the more content you put into these tools, the better outputs you get from them. So once I saw, you know, these custom GPTs come up and ChatGPT said that, you know, now you can create these kind of stylized, personalized mini models, I was like, this is exactly what I've been looking for. And I've been use them ever since. They're fantastic.

Okay, there's a lot of marketers and creators and entrepreneurial type people that listen to this show. Why do they need custom GPT? And maybe first define what the heck they are. Simply stated, we're going to get into a lot more depth, obviously, but why in the world should they basically pay attention to what we're going to talk about? Because I think a lot of them are like, I don't think I need anything like this, but they don't understand, so tell them what, what is it? And then let's talk about some of its advantages.

So basically they're like these personalized mini models that you can create within Chat GPT to help you deal with specific tasks or specific processes. So instead of having to, you know, put all your information requests into one huge prompt and then when you want to do that same thing again, you have to go like scrolling through your lists of prompts and try to find it and then, you know, pull that back up to the top. You can just train one of these models as a resource that you can come back to over and over to help you do some like recurring task or repetitive process.

And what does that make possible? What's the upside to that if you start using these things correctly?

So it just makes it easier to kind of iteratively go back and train it and customize it and form it to give you specific the output you want, the content you want in the format that you want that's most useful to you.

Okay, so what I'm hearing you say is that if everybody pays attention to what we're going to talk about today, they're going to be able to essentially customize at a very deep level a bunch of different variations of a ChatGPT that will allow them to get much better results. And when that happens, they spend less time. Right. They get less frustrated, they increase the likelihood that they look great because they've got these customized agents, if you will, that are doing all sorts of various things, and that just opens up a whole world of opportunities. I mean, that's me expanding a little bit about upon what you're saying. But is that kind of what we're talking about here? Is that the upside?

You nailed it. Absolutely. It's exactly right.

Okay, cool. So now we're going to start with some examples just to kind of open up everybody's minds of what a custom GPT can actually do. And you and I, in preparing for this, came up with a short list of different examples. So let's dig into these examples and feel free to go into as much depth as to what they do for you as possible, just so people can wrap their head around this.

You bet. So, you know, some of the best use cases I have and some of the most learning that I've done about how to leverage these tools have come from helping me solve problems in my daily life, so not just within my work life. So I've got one custom GPT. I call it the Culinary Creator, and I use it to help me with my weekly meal planning and my shopping. I've got another one that I use to help me with my fitness routines. I created one a few weeks ago to help me with list auction items on Facebook Marketplace and Craigslist. So all I have to do is just take a picture of the item and it goes out, it researches it, it writes the title, the description, it adds the price based on what's selling, and it can even create, like an Excel file that you can just upload, and it automatically creates the listing in Facebook Marketplace. So it saved hours of time. And then professionally, we have a whole slew of different custom GPTs, and we're doing all sorts of things with it. So we've created sort of interactive Personas for different audiences. So I love Personas from marketing, but, you know, typically you spend a lot of time creating them and researching and gathering all that data and all that information, and then you kind of give them to your team as a handout, and you hope that they use them to add, like, consistency to their content and address the questions and concerns that you outlined in there. But with these custom GPTs, we can actually use these interactively so we can have these different Personas look at our content and review our content and help us find gaps in the content or see what other questions they might have that we haven't thought of before. We can even have it. Do competitive analysis and look at our peer sites out there to see what they're doing with these tools. Moderna has 6,000 employees and they went deep into kind of the custom GPT landscape when they got everyone in their company on board. So within four months, they've created over 750 custom GPTs for specific workflows, problems that they can share within the organization. So there's a ton of opportunity within these. We can dive into any of these deeper.

Yeah. So, okay, first of all, I'm going to come back to the interactive Persona thing because I think that's really cool. Culinary creator or the meal planner, just explain exactly what that thing does that you cannot do on your own, just so people understand what that is.

So my wife and I have six kids between the two of us, so we have a lot of food. I'm the cook in the house. So typically on a Sunday night, we'd be sitting around and I'd ask everyone, you know, what do you want for dinner for the week? And it was always, you know, whatever, we don't care, you know. So I would have to spend my Sunday night planning out the meals and coming up with the ideas, and we would get in these routines of kind making the same thing, and then I have to put the shopping list together. So as soon as custom GPTs came out, this was one of the first use cases. I said, this has got to work for this. So now on Sunday night I can go in and I can just say, give me some meal ideas for the family. And it already knows, like, what our food preferences are, what types of things we like to eat. One of our kids is a pescatarian. Some of them don't like spicy foods. So it can take all that information. It knows what appliances I have, what cooking methods I like to use, what types of foods. So it'll just give me, here's all these different, like, main dish ideas and all these side dish ideas. And it will just kind of brainstorm for a minute and then once I figure out, okay, you know, here's the list that I want these recipes for Monday night and these recipes for Tuesday night. It'll give me all the Recipes, and it'll give me an image of what, you know, each one of those should look like when I'm done cooking it. And then I'll say, okay, now give me the prep list. So it'll tell me, you know, on Monday, you need to chop up all these things, and you need to get them all ready for Tuesday night so that everything takes about half an hour to cook from start to finish. And then I'll say, great, give me my shopping list for the week. So then on Monday morning, I can go to the store and I can buy everything that I need. It gives me the list in order that I shop in my grocery store, so it's personalized to me. And I can do my entire week of meal planning in about five minutes on Sunday night, where in the past, it would take me, you know, an hour or so to put it all together.

So are you telling me you've literally told it which aisles have which things, and then it knows when you walk down the aisle exactly what thing in each aisle?

Yes, absolutely. So I was like, first, you know, this is the path I take. I start in produce, and then I go to the. To the deli, and then I. And I go through, and then, you know, row, and I take. I just took pictures. I just took pictures of markers at the top of the aisle, and I just uploaded them. And it knows Aisle 4A is where you go to get your spaghetti sauce.

Okay, this one example is blowing people's minds. Okay, now the fitness. What are you doing on the fitness front?

So I have the same thing I tell it here are, you know, it knows basically how many days of cardio I like to do, what sort of activities I like to do, but I don't like to do the same routine every time. So I'll be like, well, this week when I'm doing my weight training, I want to do these particular areas of focus. I want to do, you know, arms or chest or whatever, and it'll change up the different exercises that it gives me each week. And then it'll ask me at the beginning, it's like, do you have any injuries that you're, you know, nursing this week? So it'll stay away from those particular areas, or it'll say, you know, are you feeling energetic this week? Do you want to do more cardio this week? Would you rather. What's the weather look like? Would you rather bike this week? Or do you want to, you know, do the rowing machine? So it creates the custom plan for me based on the input that I Give it.

Wow, that's pretty cool. And then you already talked about the live auction stuff. Now, now to the interactive Persona. So help people understand what that thing is capable of doing. Interactive is like a hint, but what does that really mean?

So we can basically point it to different pages on our website and say, look at our admissions page. So we've created like, we have a potential pat, the undergrad. So it's set up to be a potential undergraduate student, so a high school student. And we have trained it based on, you know, the data models that we have about what our incoming student class looks like, what the makeup of that class is.

Real quick, when you say data model, that scares people. So what do you mean by data model that you're training on? Data model.

So from our admissions team, we have all sorts of information about where are different incoming first year students, where they're from, what ages they are, you know, what ethnicities they are, all sorts of different information about them as people.

Okay.

And we have also gone out and asked because we, we've got some Personas that we've already created the traditional way where we ask and we interview, we do surveys, we gather that qualitative and quantitative data. So we put all of that into the system as well. So it's this really fairly comprehensive model of what this audience looks like and what their makeup is. So then we can just say, go look at our admissions page and tell us, what do you think about the content there? Is there content that's missing? Are there questions that you have that aren't answered on our page? Okay, you can ask it to dive deeper into, you know, would you rather see course information sooner? What other concerns do you have that we're not addressing? We can do competitive analysis and say, go look at MIT's site and see if you know what their admissions page looks like and what content they have that we don't have. So you can do all sorts of information gathering about this. Is the layout right? Does the language seem authentic to you? And it does. It provides great feedback. Especially we did this for international students on our international admissions page. It was like, this content needs to be more bulletized. It needs to be shorter sentences. You need to remove the jargon. We're like, oh, yeah, all right, that, that all makes sense when you think about it from the perspective of that audience. But when me, you know, as an admissions content creator, is trying to think of those, I can't always get in the mind of those students. But I've got this tool that can do exactly that it can get in the minds of the students and it can look at the page from their perspective and give me that feedback and without me having to go out and like capture these people from across the campus and have them come in and do case studies and do research groups and those sorts of things.

What I love about this interactive Persona concept is it's very different than any of the other things we've talked about with the other guests on the show. Because in this case, you're essentially asking the AI to act as if it is your prospective customer and you are training it up with as much information you can and then you are querying this quote, unquote, AI customer to ask whether or not this makes sense so that you can increase the likelihood that your messaging resonates to the real prospective customer. Which I think is absolutely brilliant. Okay, I'm glad we went down this little rabbit trail and talked about all these applications because I think it was important for everyone in their mind to understand the different kind of use cases. And I'm sure that people are listening right now and saying, okay, there's all sorts of different creative things I could come up with. So when it comes to building a custom GPT, what are some of the first things we need to understand? Like what's the tech we're using to do this?

Let's start with that. Yeah, so you do need the paid version of ChatGPT. So you can only build custom GPTs within the paid version. I mean, you can do this within Gemini or within other like open source tools, but you have to have a higher level of technical skills and you have to kind of go into the back end and set those environments up. ChatGPT makes it super easy and the environment is already built for you there.

Now just to clarify, you only need the paid GPT to build it, but anybody with a free GPT could potentially access it, or is that not true?

I think anyone with the new. Now, since they updated with the chat GPT4O, I think those 4.0 users can use any of the public facing GPTs. But I think if you create like institutionally as we create these, we don't make these publicly available.

So you'd have to have multiple licenses to be able to get into it. Okay, got it. Okay.

Right.

Okay, so let's say we've got a paid version of Chat GPT, which I know I do. Where do we even find this, by the way? Is there a certain place we have to go to get started?

Yeah, I mean, right within, like in the upper left menu where you go to create a new chat. Once you get the paid version, it'll just say explore GPTs. And then when you click on that over on the other side, you'll see my GPTs or create GPTs. And you can just go in and, and build them after that. And the first time you create one of these, it'll ask you questions. It'll say, you know, what do you want to do? What are you trying to do with this GPT? Who's the audience that you're trying to talk to? I don't spend a whole lot of time on those initial questions because it doesn't always pay really strict attention to the information that you put in there. But once you kind of get through those, then you can get to this, like, configure setting. And once you can click on that configure button, then you can open up the instructions area, and that's where you can really get specific and you can really start figuring out, here's the workflow that I want you to use. Here's the data that I want you to access to have all this information to make your decisions with and to reference as you're going through and solving this problem for me.

So what are some of the things we need to be gathering up, if you will, when we start to create our own custom GPT?

You really need to have a firm understanding of the workflow. Like, what's the process you go through when you're trying to do this? So for, you know, for instance, with, with my culinary creator, I knew that, you know, first thing I want to do is just come up with ideas. I want to figure, I want to brainstorm some ideas. So you just go in and you tell the model that. And you, you know, I use Chris Penn's race model, so you give it a role first. I'm like, you're a gourmet chef. And then you tell it what actions you're going to take. I'm going to ask you to brainstorm ideas with me. I'm going to ask you to come up with recipes for me. I'm going to ask you to create a preparation schedule for me and then give me a shopping list. And then you go into that context, you give it all the information. Here's what our family likes to eat, here's the types of foods that we like, here's what we don't like. Here are the accessories that I have, the cooking implements that I like to do and the preparation methods that I like. Kind of Go through all of that process. If there are particular like chefs that you like, you can send it to, to the email or to the URL for their websites where it can go get recipe ideas and you know, get inspired by those different things. If you have a PDF of a recipe, like a family recipe book, that's your favorite, you can upload that and it'll use that to help, you know, gather more ideas. And then so you've got this whole kind of data set together. It's and you've outlined the exact specific process that you're going to have it use with you.

Let's talk about filling the gap between AI potential and AI results. I'm sure you've tried ChatGPT. Are you seeing real ROI? Most marketers are barely scratching the surface of what's possible with AI and it's because they're missing the training and the frameworks that will allow them to be very successful in their careers and in their future. AI business world allows you to get that training. When you attend, you get two full days and 20 expert practitioners, many of the people that you've heard on this very show, that will teach you exactly what you need to know to build the workflows and to succeed as an AI enhanced marketer. It takes place April 29th and 30th in Anaheim, right alongside social media marketing World. And here's the good news, you can save $300 when you register by January 16th. Head to AIbusinessWorld live right now. This discount ends soon. So some of the questions that come into my mind have to do with particularly the process. So so far what I've heard you say is like, hey, what's the role that you want this custom GPT to take? In your case, it's a chef, right? Or some sort of culinary advisor or something like that. And when you get to the next step, which is what's the process you want to go through? A lot of people might be thinking in a very linear, strict way, like you would think you would have to do when you're programming something or when you're like providing really detailed instructions. But it seems Chat GPT is pretty darn smart. Does it really need very detailed step one and then all the. Do you have to figure this all out yourself? I guess is what I'm asking, you know what I mean? Like how much detail does it need in order to be able to do what we're asking it to do? Because it's interacting with you presumably, right? It's asking you questions as you Go, right, yeah, absolutely.

And, like, I have another GPT that I've created to help me write prompts. And, you know, it will help me walk through that process and I'll say, you know, now ask me what other questions that I should be answering to help you create a better prompt for me. But it's all about, you need to have a basis to start with. So you give it some, like, guidelines to go from, and then you start just using the tool. And as you use the tool, you're going to figure out, no, I really wanted to do this this way. It'll work best for me if I get the information in this format. So then you'll go back and you'll be like, okay, now here's how I want to change the process, or here's what I want you to do at each, you know, at each step of the way. So you can be very free form with it and just start playing with it. But as we know, with any prompts with ChatGPT or any of these tools, if you want to get back specifically what you're looking for and what you need, you need to be specific with it up front.

Okay, so again, having never built one of these myself, and assuming most of the audience has never done this, once you've built the custom GPT, is it smart enough to modify itself when you tell it to ask different questions, or do you have to go in and edit it?

You can tell it within one prompt of that tool. Like, one iteration of that. You can tell it and it'll remember for that particular iteration. But if you wanted to do it consistently every time you use the tool, like, one of the things I was doing with my culinary creator is after I got the recipe, I wanted an image of that recipe. So it would give me the recipe and I'd be like, okay, now create an image. And I did that three or four times. And I'm like, well, I'm just going to go back and add that to the instructions. Anytime you give me a recipe, give me an image of that recipe. So now I don't have to ask that question anymore.

Very intriguing. Okay, so when we're setting this up, what we've talked about is, first of all, we've talked about what is the role of the GPT. And then we've talked about the process. Is part of the process to specifically tell it? This is the output that I want. I want it to output image, I want it to output a table, I want it to output text. Or is that something else? I'm Just curious.

No, that's absolutely right. Yeah. The more details you can give it. So, for instance, on my Auction Creator tool, at the end of that, I ask it, I'll do all of the. You know, I basically, I just take a picture of the item, and it generates the listing, and then I'll take a picture of the next item, it'll generate that listing, and then I'll get like five or six items together. And then one of the things I had uploaded as an example was an Excel file from Facebook Marketplace, where it gives you kind of their outline for being able to upload multiple listings at a time. And so then I would say create all the listings together. And so I knew in my instructions that I was like, when I say this, I want you to go use that Excel file and I want you to create a duplicate of that in the same format and add all of the items that I had generated in this prompt up to this point and put those all in this file.

Awesome. You mentioned that the magic kind of happens when you go underneath the configuration and you open up these instructions. Are we to do everything in the customization configuration category here? Is that where it's just. Is it just a blank open field?

So when you first create it and it asks you those, like, initial questions, it will populate some of that into the instructions. So it'll kind of gather what it figures out based on what answers you gave it to those questions. And it'll take those answers and say, you know, here's the audience that we're talking to, here's how we want to kind of address it, and here's the content that you want, you know, the output that you want. But then you can go in and you can either take all of that out and put it in specifically the way that you want it, or you can use that information in there because it asks you, what's the tone that you want this GPT to have with you? Do you want it to be professional? Do you want it to be casual?

Is this under configuration or is this in the main settings?

Yep, it's under config. So it'll ask you that initially in the upfront, but then once you finish that kind of upfront, it will put that into the configure instructions section. But then you can go in and customize that and change that. If you're like, I don't really like the voice or the tone that it's using, you can go in and change that within the instructions.

Okay, crazy question speaking. I know that as we're recording this In June of 2024, audio has been announced for ChatGPT4. Oh, not 4.0, but for O, the letter. Oh. Does voice work with these custom GPTs? Have you started messing around with that yet?

Yes. So I have another custom GPT that acts as a job interviewer. So I can go in and upload a job description and then turn on the voice mode and I can, you know, simulate a job interview with it asking me questions and me giving it feedback. And then at the end, it'll give me a summary of how I did during the interview. It'll tell me what other sorts of information might be good to include.

Okay, crazy question. Can you record an interview and upload the audio file and analyze it that way? I mean, is this something yet that they've allowed the addition of audio in there? Because I know you can speak to it. I just don't know if you can upload like an MP3 file. You probably haven't tried that yet, have you?

I have not tried that. I know you could get a transcription of that and upload that. I've done that before.

Yeah. And then you could use that to help you analyze whether or not there's something you might have missed when you're interviewing Candidate X.

Right, Exactly.

Fascinating. Okay, let's talk about how you put together this virtual Persona a little bit, because that's a little more complicated in some regards and a little less complicated in other regards because you're not dealing with photographs and you're not dealing with aisles. But. But it is equally complicated. So if someone wants to put a virtual Persona together, what are some of the things they need to be thinking about in order to increase the likelihood that they get a Persona that acts and behaves in the way that they would hope.

Yeah, I mean, so. And it's really the same as building like a traditional Persona. You want to try to gather as much information as you can from the GPT, basically. And one of the things I like to do, because we know these different models, these different tools will give you, you know, different qualities of output. So when I'm building the initial kind of Persona, I'll open up Claude Gemini Hyper. Right. And chat GPT. So I got four different LLMs open, and I'll say to each one of them, you're going to be a potential undergraduate student looking at this university, and I'm going to want to create a Persona for you. And so then it'll, you know, spit out information about kind of a base level Persona. Age demographics, where they're from, kind of some back history, and then you just ask it the same questions that you would ask students if you were interviewing them. What are your priority initiatives? What are your decision criteria? What roadblocks are you facing about making a decision about higher ed? What content journey would you like? How would you like to get your content? What channels are most interesting to you? Who do you want to be delivering these messages? Do you want to be hearing from some counselor or another student? Or another student? Absolutely. So you ask all these questions and you get all this feedback, all this information, and that's what you would take and put into your custom GPT.

Okay, why are you using multiple LLMs, large language models? Why are you using Claude, ChatGPT, Gemini? Help everybody understand. Like, because that seems like a lot of potential redundant data, but what are you looking for when you're doing something like that?

Yeah, and basically I'm looking back and forth between all the tools and saying, okay, you know, they all cover this. So, you know, I'll just take any one of the four and use that chunk of data. But, oh, Claude is giving me more detail on what the user journey might be for this particular piece, and hyperrite's giving me more information about what their pain points are and what they're really concerned, what their fears are about going off to college for the first time. So we can kind of pick and choose between the four different tools and figure out which one's given us the most valuable content that seems the most authentic with our users.

Hyperrite is something that I've not yet heard about. So what is that? How is that different than the other tools that you're using?

Yeah, it's just another LLM and it basically has different sorts of modalities, so you can go in and be like the reporter. Hyper. Right, Variation. So it kind of has its own little GPTs within it that specialize in different areas. So you can click on one of those and then get more of a, you know, kind of a reporter feel for a particular piece of content that you're creating.

Do you think hyperrite is powered on the back end by some of the larger models like ChatGPT and. Or Claude, but it's just got those special training things? Or do you feel like this is its own unique thing?

I don't know. I use the tool without really understanding where the model is.

So back to the building out the Persona. What depth are you going in? Because we're both writers and you know that you could have a Persona that's like 50 pages long or one page long. So how much depth. Depth are we looking at when we're creating a Persona here?

Yeah, I mean, I tend to start off fairly high level and then I'll test the. The nice thing about the GPTs is you can always test them. So I'll. I'll go high level. I'll test it. I'll see if it feels like the information is giving me what I want. If it's not, then I'll go back and I'll just add more details. I'll. I'll go dive deeper into like we had to do that with our international student. We really had to think more about what the different obstacles they face are and what's what their decision process is because they're, you know, making. It's a very different decision to travel to another country and study there versus, you know, traveling to institution within your state if you're, you know, a local to that. So it's the same thing. It's that iterative process and it's like all of these tools, using them, you're really kind of pushing and poking as you're querying until you figure out what the right structure is to get the best output.

Okay. Speaking of the output, generally speaking, most of the output is text, but as you mentioned with the meal planner, you can actually output images. Now, I'm assuming these are DALL E3 images that are generally photorealistic and sometimes it can make crazy mistakes and stuff like that. So when you're asking it to create an image, do you also ask it to create images of the. What the Persona might look like? I mean, have you done that or no?

We do. Yeah, absolutely. So it'll use the base model Persona and then we will ask it to create an image. We've also tried having it create like an entire laid out Persona, you know, takeaway, basically that you would give to your marketing teams. That's what we typically, when we think of Personas, that's what we think of is this nicely laid out, very well designed, you know, with the image and all the details and everything like that. And it does a pretty decent job of giving you like at least a starting point for using those Personas. It doesn't always do a great job with integrating the text into the Persona, but then you can give that to your graphic designers and say, here's what we think we want the layout to look like. And now you can go and really implement this.

So in this particular case, if you're going to create a PDF out of this, for example, for the marketing department, you're asking your custom GPT to essentially write the copy and recommend where the images would go. Is that what I'm hearing you say? And then you give that off to your creative department and they'll build the thing.

Absolutely. Yep.

That's really cool. Okay, let's talk about. You work for a business, and there's obviously a higher education business, and you're not just using this GPT yourself. So kind of help everybody understand that's working in business that has multiple employees. How in the world do they let this custom GPT out into the wild with all their other staff? Because I would imagine that's going to require A, some training or B, some rules and guidelines. Like, how are you guys using that inside of the university?

Yeah. And so right now we're doing that within our marcom AI committee. So we've got this committee together that are. It's kind of spread about across marketing people across the institution. So we've got about 20 of us, and we have GPT teams account for all of us. So we can individually create these custom GPTs. So I can take the potential Pat, the undergrad, and we have a Google sheet where we have the URLs to these different GPTs. So anyone within our team can go click on those URLs and they can go in and access the GPTs and use them to do their own research. So, and, you know, and we're also sharing the instructions back and forth. So I can reach out to someone who's in our admissions team and say, hey, go look at this, you know, potential undergraduate student GPT that we have and see if that's helpful for you to use for, you know, admissions. And yes, absolutely, it is. But then you go to advancement. It's not really practical for them. So then we can say, all right, well, here's the instruction block that we did, here's the prompts that we use to create this. So now you can go and create a donor profile for, you know, basically using these same prompts. And now we've got another GPT that we can use as an alumni. So there's different teams that can leverage the same kind of techniques, but just customizing the data to fit their particular needs. And we do. Anytime before somebody starts creating their own GPTs or using any of these AI tools, we start off with, here are the risks of using generative AI. Here are the things you want to be aware of. Copyright, hallucinations, all these different ethical issues that we have to think about. Here are the prompt guidelines that we're using. Here's the model that we're using for those. So we have some basic kind of understanding about how you create the GPTs, what the process is. So they have to go through those kind of education segments before they can actually start building their own GPTs.

Okay, I got a couple questions here. First of all, on the TEAMS account. Well, how does that look different than the individual account? What do you get with the TEAMS account? I would imagine it's slightly more expensive too, isn't it? Right. Per user?

Yeah, it's $30 per month per user. Do have the, the new GPT EDU version, which is $15 per month per user. But you have to have like a minimum number, like 300, I think, or.

350 something for educational institutions.

For educational institutions only. Right.

Are there feature sets that you get with the TEAMS account? Like is it the sharing of the information across the different GPTs and is there some level of extra level of privacy that might be involved as well? If so, tell me a little bit about that.

Yep, that's exactly right. So you can share information back and forth with your team. You can share not only these custom GPTs, but you can also share like individual prompts so you can show other users what your prompt was and what the output was for those prompts. And it does have information that you put into these teams, the prompts you use, these custom GPTs. None of that information goes back into ChatGPT system. It's not used to train their models at all. So you have that level of security. If you're trying to put intellectual, you know, IP information or personal information into these models, you've got a little extra security built in there.

Okay. Most people that are active inside of CHAT GPT understand that there's different variations. You've got ChatGPT4, you've got Chat GPT4O. Do you get to choose when you set up the custom GPT, which model you wanted to use or is that a user selected feature? When they're actually interfacing with the custom.

GPT, I think by default they use the latest version. So I think the custom GPT is like I created custom GPTs back with GPT4 that now I have had to update to 4.0 because they work differently.

Does that mean you went in and you went ahead and you actually had to say use the 4.0 model or are you saying that's not necessarily required?

I actually recreated. I, I just copied the instructions and put it into a new model.

Did you notice the difference?

Yeah, I did notice a difference, but I don't know if that was, you know, just because the, I mean, even the, the four model is, you know, these models all evolve somehow over time.

Right.

Or if it had just, you know, gotten more used to. Because now these, you know, the chat GPT has some more internal memory as well. So I don't know if there was really a difference or not.

Well, I know you also said you use Claude, and I also know that Claude has teams accounts, but I don't know if Claude does custom GPTs. Have you discovered anything inside of Cloud that allows you to do this? Because my feeling is Claude's a better writer. Do you find that to be true? Generally speaking, yeah.

Yeah, for sure. And that's one of the reasons that we use it. When we're doing, you know, the research part and the, and the content generation part, we'll use Claude and then we'll paste that into chat GPT. And we do have a teams Claude account set up for our committee, but there's, there's no like custom GPT capability yet.

Okay. Looking into the future, I want you to put your futurist cap on. You've been messing around with this at a level that most of my guests have not yet done, especially inside of a entity enterprise, you know, kind of level. Where do you see this all going 18 months to a couple years down the road?

I think AI agents are going to be the next big thing. I think these custom GPTs are like training wheels for AI agents. They're giving us a glimpse of what things might look like. So instead of just, you know, asking for ideas and having it help create like written, visual or audio content for you, I think the AI agents are actually going to be taking actions and solving problems for you directly. They'll be using other tools, they'll be going into your browser, they'll be going to website accounts and things like that.

Will they be more autonomous? Is that kind of what you're implying?

I think you are going to be able to give them access to certain things like certain accounts, and then they're going to be able to take actions within those accounts.

So what I think I'm hearing you say is right now the current AI, quote unquote, custom GPTs we're creating are very limited because they require us to provide it all sorts of information and require us to do the, the steps. Right. And what I think you're saying is that in the future, your culinary creator is going to go order all the food and have it shipped to your house. Is that really where you're going with this? Is that what you mean by that?

Yeah, absolutely. And once robotics gets to a certain point, hopefully it'll be able to cook the meal for me, too.

Okay.

Wow.

Okay. And what do you think that's going to do for, I don't know, just the space in which we operate today. How's that going to change things?

I think it's going to make us much more efficient with our time. It's going to give us a lot of time back to do the things that we like to do. I think it's going to create a lot more decisions on our part because, you know, these AI tools can do a lot of things and it's up to us to decide what they should do, where we actually do need human interaction within the workflows and within the processes that we perform every day, and where we can automate things that we don't need to be involved with.

Brian, this has been absolutely fascinating exploration. If people want to discover more about you, do you have a preferred social channel? And if they want to check out your newsletter, where do you want to send them?

Yes. So you can find me on LinkedIn at Brian WPiper. Also my website, BrianWPiper.com that's where you can sign up for. My weekly newsletter features one content marketing tip each week.

Brian, thank you so much for coming on the show and sharing your insights with us. We're so much better because of it.

Well, thank you, Mike. I love all the content you're creating and what you're doing with the show. It's fantastic.

Hey, if you missed anything, we took all the notes for you over@socialmediaexaminer.com a11. Also, be sure to follow this show on your favorite podcast app. And would you share this show with a friend if you found value in today's episode? You can tag me if you want. I'm Stelzner on Facebook and ikestelsner on X. And 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 become more successful.

The AI AI Explored podcast is a production of Social Media Examiner.

AI Business World early bird. Pricing ends on January 16th. You can save $300 on your tickets. Register at AI BusinessWorld live today.