Custom GPTs for Marketers: Creating a Virtual Team

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. Today, I'm going to be joined by Russ Summers, and we're going to talk about something really fascinating. We're going to talk about how to create custom GPTs that can almost be like virtual employees or virtual advisors for your company. And we're going to talk about all sorts of different ways to do it that I think is going to open your mind in a pretty dramatic way. By the way, if you're new to this show, follow us on your favorite podcast app by simply clicking that download or follow button so you don't miss any of our future episodes. Let's transition over to this week's interview with Russ Summers.

Helping you simplify your AI journey. Here is this week's expert guide.

Today, I am very excited to be joined by Russ Summers. If you don't know who Russ is, he is the VP of marketing at at Quantified AI, an AI simulator for salespeople in life sciences, insurance, financial, and technology industries. He's also a songwriter and doing some great things in the world of AI. Russ, welcome to the show. How you doing today?

I'm great, Michael. It's good to see you, and thank you for having me.

I'm super excited that you're here Today, Russ and I are going to explore how to use custom GPTs in business and marketing to do some really, really cool stuff. So, Russ, before we get down the fascinating rabbit trail of all the great things that you're doing with AI, I'd love to hear a little bit of your story. Start wherever you want to start. Like, how the heck did you get into ultimately working in the AI industry and doing all this crazy AI stuff that you're doing right now?

Yeah, I'll just start with that. I had an interesting experience a few years ago. I had been at a really cool company that I was enjoying. We did some great work. We did a rebrand. It was going, you know, we, we did some great stuff. But we came to a crossing point where it made sense for me to leave the company because we were not growing quite as fast as we wanted. So bluntly, I got laid off and I kind of engineered my own layoff a little bit.

What were you doing for those guys? Were you in marketing or what was your role back then?

I was in marketing. That's pretty much what I've done for the past 20 years, really. But like a lot of marketers, jobs come and go, and that time on the beach between jobs, I had been getting increasingly fascinated by AI, Michael, because I've seen a number of disruptions in my lifetime. I, I'm old enough that the first one that I remember is going from black and white TV to color. But then the disruptions of the Internet emerging, of mobile, of social, each one of those created this time of massive possibility and chaos, which were always my favorite times in my career. And I was aware there was one going on with AI, and I'm like, I need to take a little time to explore this. And so I dove in, started talking to tons of people, started researching and playing with what tools were available pre chatgpt, which were not many.

Oh, interesting.

But I also during that time connected with the founder of Quantified because he had been for the past 10 years building an AI driven company, which is, yes, there was AI before ChatGPT, kids. There really was. And so I learned a lot from him working with Noah and consulting a little bit with Quantum Quantified and ultimately discovered as Chat GPT emerged that this was this incredibly powerful tool that could amplify me as a marketer. And so as I was doing marketing consulting, I started to ask the question, rather than doing what I'd usually do, which is, you know, sort of build a little team and outsource, how much could I do by myself if I was amplified by custom GPTs? And that really became the challenge is how much can I amplify myself?

Okay, so I'm fascinated by a couple things here. You started dabbling with AI stuff before ChatGPT. I'm assuming I'm thinking of Jasper. That's the one tool that I know a lot of marketers were using. I don't know what other tools people were using, but was it after November of 2022ish when ChatGPT rolled out and you started messing with this interface, this simple chat interface that Dot started connecting to you, or was it a little bit later, once they started rolling out custom GPTs? I'm just curious, what was the aha moment for you?

I mean, life is a life well lived is a series of aha moments, isn't it?

Yeah.

And so I had early aha moments with Jasper and some of the early tools. Then I had a big aha moment when ChatGPT emerged. That felt like the first time I plugged my computer in and connected it to the Internet. I vividly remember installing the network card myself and plugging it in, going, there is a whole world open here. But then, certainly the emergence of custom GPTs was another big aha moment. So it was really, I think, a series of three.

So U.S. gray hairs, metaphorically, or no hairs. That's exactly. I mean, I know you qualify as a gray hair even though you have no hair. You know, we've been around for a little while and we've seen a lot of innovations. Like the Internet, obviously, was one of the really big ones in the 90s. And then, of course, social media was another really big one. What did you see when you started noticing what was like, what was going on inside that brain of yours? Because how did you see this as an opportunity? Like, I just want to explore that.

A little bit, I think. I tend to see opportunities when there is chaos. So these things feel to me as a musician very much like the emergence of punk rock, where, you know, in high school, I'm listening to Led Zeppelin. Nobody can play guitar like Jimmy Page. You shouldn't really even try. Maybe you should learn to play the intro to Stairway. And that's about it, right? And then punk rock came along, and it's like, screw all that. Buy yourself a guitar, get some friends, write some songs. And I was part of that. I mean, I bought a bass guitar and had my first gig two weeks later. And you can only do that in those times of upheaval and possibility. So that's really what I saw was that sort of upheaval of, I don't know where this is going, but I know that for a little while there's not really a rule book and there's not really anyone that can tell me that I'm doing it wrong. So let's get out there and play and see what we can do.

So tell us a little bit about what you're doing now and a little bit about what Quantified does, and then we're going to get into some of the cool stuff, you know, just kind of like what your role is right now inside of Quantified and and maybe even who Quantified is for. Just give us a little bit of that background.

Sure. I mean, as I mentioned, Noah was one of my sensei's early on in showing me some of the possibilities of AI before it had really become mainstream. So he educated me a lot as I was consulting with him and what the company does, it's pretty cool really. It's a virtual simulator for salespeople that helps them up level their sales performance. Because with AI there are so many tools helping salespeople do more, send more emails, put more data into Salesforce, tailor your emails more, make more phone calls. And the problem is the audience you're reaching out to as a salesperson does not want more. And you can no longer achieve your goals through doing more. You got to do better. And so what I like about Quantified is using these simulated sales conversations to up level performance and help people get better. So it's not I make more sales calls, but I make better sales calls and get better results.

Does this mean that you're actually talking to what sounds like a human, but it's a virtual customer and it's coming up with real objections and stuff like that? Is that what we mean by simulated?

Exactly. So it feels as real as a zoom call, which is really pretty fun. The first time I saw it, I was like, holy smokes, how can I be part of that? And so to become part of that, I jumped in, started consulting with Noah initially and building out sort of the early go to market strategy for the company. And really what I am in charge of is all of marketing for the company, taking this new thing, this new idea, this new concept to market. And it's just me and my friend Becky, who I worked with at a past role that I recruited in here, and we both have a little army of GPTs that we use to do stuff and it's pretty fun. It's like we've got these virtual teams that augment and amplify us.

And this is one of the big unlocks. Everybody might have been like, what? Russ has virtual teams that does work for him. You got to be kidding me. And that's really what we're going to talk about today is how Russ is able to do things that what, five years ago would have required a massive team. So that's why I want you to share, like, why should marketers consider custom GPTs? What's the upside.

The upside is. That's a really good question. I mean, because a piece of it for me, you can tell that I'm passionate. And a piece of it for me is just, this is fun and I'm getting to do something new and I love to do new things. But if I get to the boring ROI part of the conversation, it's pretty compelling. It lets me do, do better output and it lets me do more output. So whether it's creating content for webinars, whether it is strategic planning, whether it is my least favorite thing, but a thing I gotta do, managing the marketing technology, stack, a lot of that, where typically I would tap a coworker on the shoulder or ask a team member, hey, could you do this for me? I can just get done much, much more quickly and frankly, in many ways more thoughtfully.

Do you find that this would have required like a staff substantially larger to accomplish the same things, or you would have been burning the candle at both.

Ends, burning the midnight oil at both ends, as my mother used to say. I mean, the informal number that I've come up with is it feels like I'm about to 3x able to about 3x my output. And that feels right. When I look at what we're doing at Quantified and realize that with two people, we're doing what I would probably expect out of a six person team. Still a small team, right? We're not the market marketing department for Coca Cola or whatever, but for our size, we punch pretty hard.

Well, and just to translate that for everybody who is listening, what this does for the company that Russ works for is it increases his value because Noah, your boss probably looks at you as exceptionally valuable now because you are able to produce so much more at a higher quality than what you've done before. And that's part of what I think everybody should pay attention to as we explore this discussion today is that you can become a more valued. If you work for as a consultant or work at an agency, you can be more valuable to your clients. Or if you work at a company, you can be far more valuable to your boss. Or if you work for yourself, you can get a lot more done without having to necessarily outsource a lot of work. And that's what I'm really excited about exploring with you today. So, Russ, when we were preparing for this, you identified three different levels of custom GPTs. Now let's just define for the newbies in the room what a custom GPT is and then start with kind of your first level.

Cool. Cool. So if you think about what a custom GPT is, and my definition is going to be inexact, but also at least work for me and probably will work for you. If you think about anytime you seriously prompt chat GP to do something, you start by telling it who it is. You know, ChatGPT, you are a writer who is exceptionally skilled in B2B copywriting and knows life sciences and insurance and technology cold and, and you have this writing style and et cetera. You do that sort of almost preamble to the prompt to set up what you want ChatGPT to be, because without it, it doesn't have context and you get generic answers. So typically you start with that. What a custom GPT does is it lets you create kind of this little separate instance or whatever of ChatGPT that already has that. So if I want somebody that's a super skilled recipe writer, for example, I'm able to do that once in the process of building the custom GPT. But then when I go back and say, hey, I want a recipe for leek and potato soup, I don't have to reprompt. Hey, remember you're this amazing recipe writer who's read all these cookbooks and knows all this stuff that's already there. So I can just go in, hey, recipe writer, GPT, give me a recipe for leek and potato soup without all of the preamble and pre work that I'd normally have to do.

Now, I know a lot of people today are probably just doing this inside of different conversation threads inside of ChatGPT. I've not created a custom GPT yet. My understanding is it lives somewhere else inside of ChatGPT and it also can be private or public or shared with the team. Is that correct?

That's correct. And there's sort of three levels. In order to build them, you've got to be on at least the lowest paid level, which is 20 bucks a month. I think they call it pro or whatever. The next level up is teams, where you can actually share them with each other. And then there's a level that is enterprise, which you see some really big guys doing. Moderna does that. For example, they have enterprise license and they've done some phenomenal things with custom GPTs for their team.

Okay, so let's start with your level one of custom GPTs. Go ahead and just kind of explain what it is and then let's dig into some examples.

Cool. Well, I mean, the number one reason and what most people start with and frankly stop with is they're given a task and they don't want to do it. So level one, I call, I don't want to do it. It's like, I need. I'm going to do a webinar. A webinar is going to require three invite emails and landing page copy and thank you and sorry we missed you emails and all of that. And, you know, having been a B2B marketer for 20 years, I can do that in my sleep. But anytime you reach a point where you can do it in your sleep, you also go, can't I just hand this off to the intern? Can't I hand it off to somebody? Do I really gotta do this? And so really the first one I built was for specifically that purpose. Her name's Wendy Webinar. I do tend to name them like team members just because, frankly, it's fun and gives them personality.

Real quick, before we go into Wendy Webinar, I just want to explore this concept of the I don't want a list. There's a lot of people listening right now that they have to do a lot of things they don't want to do. Right. So any tips on how to, like, identify what the things are that I don't want to do that are good prospects to create a custom GPT? Do you understand where I'm going with that?

I think I do. And you can steer me if I go off course on this, Michael. But in terms of identifying good use cases, there's probably two criteria to start, and I would advise people, by the way, to start by building one for fun, something they care about. My first one was a guitar tutor because I obviously care about that. I've got about six guitars behind me here. And so by doing something that's fun, you actually learn to do it, you know, because you'll just dig in and you'll care about the quality. But beyond that, the use case ought to be something initially that you know how to do really well that you can kind of do in your sleep. And an acid test for that is think about if you had a smart intern start in your office and, and the intern was smart and ambitious and had absolutely no business context or training, how would you teach that person to do a thing, take a thing that you could teach that person to do rather than we're going to develop cold fusion. I can't really teach somebody how to do that very well, but I can sure as heck teach somebody. Here's what good webinar content looks like, or here's what a good book of recipes looks like. Or whatever.

I love this point. You're starting with something you already know a lot about. And that seems to be important because that means you can train the custom GPT and tell them when it's right and tell it when it's wrong and refine it. Is that kind of what I'm hearing you say?

Yeah. And as a matter of fact, as part of that refinement, what would you do with that smart intern if you gave them the assignment and you said, here is the landing page I need you to write for the webinar, bring it back to me when you're done. And they bring it back and it's just kind of off. You wouldn't say, this is great, let's publish it. If you're a bad boss, you would say, give it to me, I'll do it myself. But what a good boss would do is say, okay, Bob, you know, great job on writing this. Really appreciate you turning it around. There's some things we need to change here, and I'm going to tell you why. And in the same way with building a custom GPT, you can take that output, feed it back in and say, okay, so the landing page you wrote me, here's the good things about it, here's the bad things about it. I want you to take another crack at it. And if you do editing, if you take it and you like, change a few words, like it uses classic chatgpt words like let's dive into or whatever, which, you know, just seem very AI written. If you change those things, feed it back your final edit and tell it why and say, hey, great job. I changed these words and here's why. Next time you write, I want you to keep that in mind.

I love it.

Just like you would do with a smart intern.

Love it. So tell us a little bit about Wendy Webinar. And I know you've got another one we're going to talk about as well, but what does Wendy Webinar do for you exactly?

Wendy's great because, you know, again, if I were in a larger team, if we were in this, you know, six person team that Becky and I would otherwise be, I would have somebody that specialized in writing that copy and I'd say, hey, we're launching a new webinar and they would just know. A webinar involves a landing page. It involves three invite emails, it involves a thank you email and a sorry I missed you email. It involves, and I need to add this because I haven't. It involves content for the BDR team to Use in outreach.

What does that stand for?

BDR Business Development Representative.

Got it.

Those are the people that follow up afterwards and, you know, salespeople reach out to the audience. Yeah, salespeople.

So you train Wendy to do the landing page, to do the emails, the follow up emails. And is Wendy also trained in recommending the content for the webinar as well, or mostly just the marketing of it?

It's a good question. She's got a lot of opinions on things because I trained her to be pretty opinionated. So I will. She'll have opinions about what the subject line should be, what the headlines should be, all of those things. And if I said, hey, we're planning to do a webinar with Michael, what do you think the topic should be? She would probably opine on that as well because at this point she's read a lot of webinar abstracts. She understands the layout, she understands what's compelling and what's not. So she becomes truly like your team member. After a while, she rises from smart intern to experienced team member. That's where she is today.

I want to dig in on the. I trained her to be opinionated. Talk to me a little bit about how you did that.

One of the problems with ChatGPT is it's like a Labrador retriever. It's very eager to please and will do whatever it thinks you want. So in building a custom GPT, you write your instruction sets and you give it the, you know, insight into what I always want three emails, I always want an abstract. But in that instruction set, I include, I want you to have strong opinions. You're not going to like webinars that are not likely to call people to action. You're not going to like a bunch of business buzzwords. You're going to like simpler writing. And if I don't give you enough information, you will push back on me. So if I say, hey, Wendy, we're doing a webinar about frogs, I need you to give me all the content. She'll say, well, before I can give you the content, you need to tell me more. I need to see a full abstract. I need to understand who the speaker is. I need to understand what the topic is, what the call to action is. So if you don't very explicitly put that stuff in your instruction set, like the smart intern with no context, she'll go off and write webinar content about frogs that is not useful to you.

Now, I know you've also created Blake Blogger and Social Celeste. I don't think we're going to have time to dig into them very much. But if there was anything particular about those two that was different than the others that you want to share, feel free.

They are both pretty similar in terms of understanding the format of the content. I guess the two things they got to understand the format of the content I want them to to output and they've got to understand the requirements to create it.

Got it. Okay, cool. So in summary, this first level of custom GPTs are things that you don't want to do, that you know how to do, that you are pretty gifted at doing. And it's at this level that we have a lot of our own personal insights that we're training these GPTs. And this is probably where you advise people start when they're creating custom GPTs. But there's a second level which is really fascinating. So tell us about the second level of custom GPTs.

Well, one of the problems with being a one person marketing team, Michael, you are probably, based on what you do for a living, aware that marketing is increasingly a complex discipline. For sure, whether it's managing paid media, whether it's, you know, managing the tech stack, whether it's brand, nobody's good at all of it. And I came up with this second level when I was struggling with the thing that I'm not particularly good at, which is the revenue operations work of ensuring that the tech stack is working appropriately. I value that. But I've always had a really good person on the team that I could go tap on the shoulder and ask those questions. And after a day of digging through, and by the way, I love HubSpot, it's a wonderful product. But after a day of digging through HubSpot documentation to solve a simple problem, I found myself saying, like a person in an infomercial, there's got to be a better way. And it occurred to me, what if I build a custom GPT and point it to the documentation for all of the systems I use? I'm like, well, that would be interesting, but maybe it would hallucinate and make things up, I don't know. So I tried. I built one. His name is Roger Revops and he's turned out to be other than Becky, probably my favorite team member. And as I built him, I said, I need you to read all of HubSpot's documentation. For our specific instance where HubSpot has been customized and we or our agency have documented that I fed that information as well. I need you to read this and understand how we're set up. And I did the same for other systems we use such as sales loft and whatnot. And I also had to be pretty prescriptive. I learned on what not to read because if it goes out and reads all their marketing content, it'll give you these very general answers. Oh, HubSpot can do a million things that doesn't help me. What I'm looking for is, oh, to connect HubSpot to Zoom info. What you need to do is this, this, and this. So getting very specific is what made the difference. But that one actually exceeded my expectations in terms of taking the place of that team member that I tap on the shoulder and say, I know you know HubSpot really well. Can you help me?

Very cool. And I know you've got a couple other examples I'd love to hear as well.

Well, you know, mentioning that no marketer is good at everything. I'm not particularly a paid media person and our go to market is not primarily driven by paid media, even though far too many people think marketing and advertising are the same. And I had worked with a very gifted paid media specialist at another company. He happened to be Egyptian and so I built my Egyptian ad.

Oh cool.

GPT is Amon ad man and he understands Google pay per click, he understands LinkedIn ads, he understands the other platforms we use. But I also told him all of those platforms and their documentation are, if you will pardon the observation, geared to get marketers to spend a lot more money. And so in his programming is you're very skeptical of increasing spend. You're very skeptical of this and that. Here are the few metrics you really care about, notably return on ad spend. And here are all the things you don't care about as much. So I programmed him even more than the others to be super opinionated. Because otherwise, if you ask an ad platform what you should do to get more results, it will always say, well, how about you spend more money? And I wanted somebody a little more discerning than that.

Quick question. How many podcast episodes have you recorded? How many webinars have you hosted? How many client calls or team meetings have happened that are now just sitting there in your files? Here's the big unlock. Those aren't just old recordings. They're full of insights about your audience, your brand voice, and your customers actual objections. You just need to know how to extract them. That's exactly what AI educator Erica Stanley is teaching AI Business Society members on January 8th. By the end of this training, you'll have a repeatable workflow that turns every past recording into a Content and data powerhouse, one that keeps paying dividends long after the mic's been turned off. This training is exclusive to AI Business Society members. But here's the good news. You can join us right now by going to socialmediaexaminer.com AI and start transforming your old content into new marketing gold. Socialmediaexaminer.com AI to join right now. I also know you built SEO Serena and Alex Life science. And what's really fascinating about Amon Admin and Roger Revops and all these is that you're training them in areas where you do not have domain expertise is essentially what I'm hearing you say, Right? So what are some of the tips on how to actually train something when you don't actually know what you're training it on? Do you understand? Like, how do you do that? I mean, give us a little bit of advice on that.

Well, let's look at, you know, Alex Life Science as an example. We sell into the life sciences industry, which is a new industry for me. I don't know the lingo, I don't know how people act. I don't know really. Actually, I know a fair amount now, but took me a while to learn. Right. So the dirty secret about training it to know something that you don't know is you at least have to know where to get the information. So I talk to customers, I talk to people in our space to go, where are credible sources of information about life sciences and life sciences marketing. And they said, oh, well, there's L10 life sciences trainers and educators network. There's a set of other resources. They told me what was credible and what was not. And then when I built Alex, I trained him on those sources. I said, you're going to focus a lot on these sources. You're going to ignore these other sources. Same with SEO, Serena. I'm pretty opinionated about SEO, not an expert, but I care about it. And we all know an awful lot of SEO is SEO stands for someone else's opinion, right? And so I picked a few blogs, a few writers, a few people that I trust in the space. And I said, these are the people you're going to take seriously and take their advice and use as reference points to determine what you believe.

This is really intriguing. And I would imagine if someone listening wants to create a, for lack of better words, Persona, custom GPT for an area that they don't have any expertise in and they don't have access to experts like you do, I would think, Russ, that you could start by asking ChatGPT or Claude, hey, who are the leading sources for information on X, Y and Z? And then you could go to those sources and see whether or not they seem to have the kind of information that you would want to ultimately use to train. Have you ever done anything like that? I'm just curious.

I have. I've used those to generate information sources. And I'm glad you mentioned Claude, because I'm talking about ChatGPT a lot, just because I use it a lot. But for things like that, for sort of broader analysis, I sometimes find Claude a lot more useful. And to sort of reframe that prompt you suggested just a little bit, I might go, what are the primary issues and controversies in this space? And who are people? What are the different viewpoints essentially? And who are people or resources that sort of espouse and represent those different viewpoints points so you start to understand the industry dynamic.

So once you've created these level one and level two custom GPTs. And again, level one is things that you don't want to do that you know how to do, and level two are the things that you don't know how to do, but that you'd like some advice on. Essentially. How often are you using these and what kinds of tasks are you doing with these? I'm curious.

I bounce back and forth between most of them during the day. Sometimes it's simple, you know, content creation, whether ideating content or finishing content. And by the way, just to be super clear, a lot of people are like, AI, written content is the devil. It's terrible. And yeah, a lot of it is pretty crappy. But by the way, a lot of human writing is pretty crappy too. So no writer is good without an editor. So either chat GPT is the ideator and then I edit, or it's the editor as well as the creator. But the point is, it's iterative. If you think I'm going to plug in a prompt and get a blog post, you'll get a terrible blog post. So you got to iterate a lot. But I'm doing it to create content like that. I'm doing it to fix problems, as in dealing with Roger. I'm doing it to identify opportunities, as in with Serena. Hey, we see some SEO upside. How do we make the most of it bouncing back and forth between them all multiple times a day on tasks like that, just as we were would if we were in a little office and there were six of us bouncing ideas around and learning from each other.

Now, the importance of feedback is something we talked about earlier and I'd like to like dig on this a little bit more. What type of feedback do you provide? What we know so far is when it provides written output and you ultimately edit it. You basically said, hey, I tend to put that back into ChatGPT and say, this is how I've improved it. I would imagine you could say, do you see what I've done differently? If you wanted to. And it will tell you, right? Have you asked that?

I haven't asked that, but I need to because. Because it's again that it gets to the Labrador retriever thing. It goes, oh, thank you. I really appreciate that. And asking it, testing for understanding like you would for the smart intern. Do you see why I did that? That'd be a really good idea. I'm going to try that.

What other kind of feedback are you giving the GPT beyond just I fixed your output. Is there any other kind of feedback that you're providing into the GPTs? Do you let it know when you like its results? Do you say great job or do you give any of that kind of feedback?

I always say great job because that works with human team members and I believe it works with custom GPTs as well. And in fact, there was a study done a while back that said that you offered to tip chatgpt it would do better work.

Really?

Yeah, I saw that and it blew my mind. Michael. I don't want to have a running tab with Skynet just in case the singularity happens. Right. So all of my custom GPTs have a favorite coffee and I'll reward them with a coffee when they do great work. A virtual coffee. I'll say, oh, Serena, you like Earl Grey tea? Go get yourself a cup or whatever.

That is so cool. Do you ever tell it to take its time? Because I've experimented with this a little bit and I don't know why, but for whatever reason, sometimes if you tell it take your time, it, for whatever reason, outputs slightly better content. Have you noticed that at all? Have you tried that?

I have. The phrase that I was taught by a friend to use was do this step by step, which is a slight variation on that. But in either case, it does seem to cause it to, like with a person, slow down and just not slam output out. Just in the thinking fast and slow. Daniel Kahneman about how we think differently when we take our time. And it does seem to somehow unlock that. But I'm going to try the take your time with this because I haven't and I'm curious to see what that will yield as you are building.

And we're going to get to this third level here in just a minute. But as you're building out these custom GPTs, do you find yourself modifying kind of the instruction set to the custom GPT a lot?

I do, I do. And I probably should do it more when I notice things that annoy me, you know, and a lot of times it's just simple words. Please don't use this word over and over. That is a chatgpt word. But I'll open them up, I'll refocus them, I'll sharpen them. I had an existential session like that with SEO Serena, because month and a half, two months ago. Are you aware of the Google leak?

I'm not sure. Tell me more.

Okay. Rand Fishkin, who's an influencer in the space, and Mike King, who's a great, great SEO that I follow. 8 hundreds of pages of documentation from Google were leaked that they believed pointed to how the core algorithm worked.

I do remember that. Yes.

Yeah. And, and it contradicted some things that we're told that we've been told about domain authority and things like that, whether they mattered or not. So it was pretty controversial in the SEO world. And it, when it came out, I had this weird experience. I like, opened up Serena and I was going to add that to her instruction, said, I want you to review this and take this into account. And I'm like, this is controversial. This is blowing up the SEO world. Is she ready for this? And I almost felt like if I was sitting with a junior team member and going, serena, I need you to sit down. There have been some very big changes and some of them might be kind of upsetting to you because this is your world, but I want you to know I'm here for you to work. And I got over myself and I'm like, look, it's just a GPT. She not going to freak out. But I did really think about whether I wanted to train her on that new information, especially because it's controversial.

Well, and that's a fascinating point because it could forever alter her output. And I wonder whether in that situation it would be wise. I don't even know if you can copy. Can you copy a GPT, or is that not an option I've never tried? Wouldn't that be an interesting experiment? Like, you make a copy of the GPT and you call it, like Serena 2.0 or whatever, and then you give it that document and you see whether or not the output is fundamentally changed as A it result of. Result of that input. I wonder.

It's a really good idea. A thing I don't do as much as I should is save my instruction sets as I write them. Because if you go into a custom GPT, there's a place where it will show you the instruction set, but it's heavily summarized from what you initially put in there. My initial instruction set may be 10 pages and it summarizes down to a page or less. And so probably a way I'd love to do that is save the instructions set and then create a new. But then give the new one that version. But there may be a way to copy. I just haven't ever looked for it.

So what I'm hearing you say is that even if you edit it, it's only editing the abstract of the original instruction set. It's not like the full detailed instructions that you put into it.

Well, the instruction set that it will output to you when I say, hey, Serena, I need to see your instruction set is going to be shorter than the instruction set I created it with. With.

Got it.

Theoretically, it's the same information, but it's not the same words or phrasing.

Okay, cool. So let's get into the third part of your process, which is really intriguing. We've been talking about things I don't know how to do in this level two of the custom GPTs. What's your third level that you've been experimenting with?

Well, I mean, I think every marketer sort of says three things to themselves, and we've covered two, right? They say, I don't wanna, and they say, I don't know how to do it. But the third one that all marketers say to themselves, when we have our dark nights of the SOU goal or our strategy session or whatever, we find ourselves saying more often than we might like, I don't know what to do. And to that end, custom GPTs can be phenomenal. Sparring partners and almost mentors. And so the. The ones that I've built here, you know, there's probably two mentor, Maya and Piper, the product marketer. I've built them to give me advice and guidance. So like with Maya, I told her, here are the marketing thought leaders that I want you to study, that I want you to learn about, and here are the marketing principles, disciplines, frameworks that I want you to listen and to read and care about. And I told her that she is a very successful cmo. She's taken a number of companies to exit. She's one of the biggest Thought leaders in the world. She follows all these other people, but she has a real soft spot because her focus is early stage companies. I did that selfishly for me because I don't want her to give me guidance that would work for the CMO of Coca Cola, but not for me in my situation. And then when I have strategic options to work through, I just kind of spar them out with her. I go, so Maya, we're. Let me tell you about where we are. Here's where we are in terms of market presence, in terms of lead count, in terms of differentiation, whatever things I might be focused on. And I'm struggling to make a decision about where we should go strategically. What options do you see? And more than any of the others, the iteration here really matters because again, if I just go to chat GPT or even a custom GPT like Maya and say give me a marketing strategy, I'll get generic output. But if I say think through and I'm going to try your, take your time with this because this might really help with that one. Given this situation, what are some options of strategy? And then you can take each one and kind of drill down. Okay, so if you think we should go really, really heavy in events and not have a website at all, which would probably be a terrible recommendation, but I imagine one could make it, what would be the upsides and downsides of that? How would we execute that? What do we think the cost might be? What do we think the time to value might be? How would we expect a competitive market to respond? And I can think of all of those things and I can call a friend and spar out all of those things. But to have somebody on demand to spar those things out with can be so, so valuable in clarifying my thinking because I'm not cheating by having her give me ideas so much as I'm using as a tool to work through those ideas systematically and flesh out all the possibilities.

This is really interesting. This is next level of level two because it's deeper work. I mean, what is it about level three? You're coming at it with zero knowledge. But I mean, do you really have. With the Alex Life Science, you probably don't have a lot of knowledge either, either. So what is it about this level three that differentiates it from level two, I guess is what I'm asking.

It's a good question. And they are in ways flavors of the same thing. The difference though is when I'm leaning on an Alex Life Science or a Roger Rev Ops, I know what I want them to do. I just don't know how to do it.

I see.

Versus in level three, I may not know what to do. So it's less about how do I execute this marketing strategy and how do I work towards a marketing strategy that will be effective for us.

So you're almost using them as a coach, right? Or an advisor.

Yeah. And if I can tell you a funny story about this, that may expose a little bit of the bias that is in all human communication and therefore is in LLMs. Noah, my boss, built himself a CEO coach and her name is Sophia. And although he named her Sophia, when he created an avatar image, it kept showing him a guy, an older guy, and he had to remind it, actually Sophia's a woman. So, you know, I just, I think that doesn't speak well of the bias of everything we've trained it on that it assumes, oh, CEO coach must be a guy. But so be it.

When you are training these, especially at the level 3 GPT, are you mostly linking to websites? Are you letting the GPT figure out where to find its information? Are you uploading PDFs? I'm just curious, like, especially at this third level when you don't have any really wisdom here and you're looking for wisdom, the training would seem to be very important, right?

It is. And Michael, I'm going to ask you a question. When you think of people that you follow as thought leaders or inspiration that you take guidance and ideas from, you probably have a list of those, don't you? In your mind?

Of course. Yeah, yeah.

So for me, I started there because if I went to see a coach and the coach was like Peter Drucker, who's that? Seth Godin.

Right.

Who's that? Right. I probably wouldn't take them very seriously. So I started with, you're familiar with the works of Drucker, of Godin, of. I don't have the list, but there were a bunch that I had and.

Maybe Robert Cialdini or something like that. Exactly.

Oh, Cialdini is great. Right. And he is in that list. And so I'm like, I need you to really understand these people's points of view and what they believe about marketing and what sort of advice and guidance they might give.

So basically you're just giving the GPT. These are people that have a work of knowledge that's out there, that's published and that are experts in their various industries by giving it that guidance and asking it specifically to study their work. Is that what I'm hearing you say? That's enough, typically for it to be able to have that education so that it can give you proper advice and guidance.

Seemed to work. But I put in a pretty long list of things, you know, I didn't put in PDFs of books I could, but I didn't happen to have them. But. But yeah, I mean, if you think about how you gain and you use the word wisdom, I didn't. I don't know that we have wisdom yet in AI. I know that we have intelligence of a sort, which is different. But if you think about anybody you view as wise, they bring together a number of perspectives and synthesize, don't they? And that's really what I was going for, is I could make a GPT that's just a Robert Cialdini, you know, GPT if I wanted, but somebody that synthesizes his perspectives with a number of other perspectives. Perspectives I felt like would give me what I would want from an experienced wise coach in terms of. Here's a number of different ways to think about it.

Okay, Russ, I want you to look into the future. I want you to imagine it's 18 months to two years from now. Where do you expect these custom GPTs to go? What kind of things do you expect them to be doing maybe that they cannot do today for you?

That's a really good question. And I got to admit, I haven't thought much about it. But one of the more exciting things to me, and you mentioned this to me when we last talked, is that there's now an analog with Claude in terms of projects. And I'm enough of a capitalist to believe that competition tends to make things better. Obviously, within reason. We could go there or not. But the point would be now that OpenAI is not the only thing with a concept like a custom GPT. I expect to see the competition increase the capabilities, and I don't see a lot that they can't do in terms of really tactical things. It would be nice if they could all have their own voices and such when I'm using the voice interface.

But, you know, I think. I think about what your business that you're working for is doing, and I start to think about the future. I can very easily see businesses, companies that train GPTs on really advanced selling techniques or really advanced marketing techniques. And then you could hypothetically go pay for these people and have them be really inexpensive virtual staff that can do first drafts of almost everything for you. I think that's coming faster than we think. What are your thoughts on that?

I think so as well. I mean, what we've learned at Quantified is the process of creating an intelligent video conversation partner, you know, that looks and acts like a person. Getting 80% of the way there is not that difficult, but the last mile of making conversations realistic and within realistic boundaries and all of those things and making them compelling and valuable. That part is still pretty hard. But I do see things going there to where that's going to become easier for people and where potentially it becomes less like I'm typing into a chatbot and more like I'm having a conversation.

Russ Summers, this has been a really fascinating exploration of the possibilities of what people can do in business and in marketing, specifically with custom GPTs. If people want to learn more about the company you work for and or you. Where do you want to send them? And if you're active on any particular social platform, is there anywhere that they could connect with you?

Yeah, I'm Most active on LinkedIn. I'd love to talk to anybody that's interesting interested in this, help them if I can. They're interested in Quantified. I'm. I'd love to talk to them about that, too. I really like to talk to them about that. And of course, Quantified AI is right there, you know, as a resource. But I'm happy to talk with anybody about this stuff. You can kind of tell that I love it. Michael.

Well, and for those that are listening, sommers is spelled S O M E R S. Russ, thank you so much for coming on the show and sharing your insights with us today.

Michael, my pleasure. This has been sucking a fun conversation and I very much appreciate it.

Hey, if you missed anything, we took all the notes for you over@social mediaexaminer.com a16 and be sure to follow this show on your favorite podcast app. And would you share us with a friend? You can tag me if you want. I'm Elsner on Facebook, Ike Stelzner 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 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 AIBusinessWorld live today.