Loading summary
Transcript107 lines
- [00:01]
Nicole Leffer
I've got some exciting news. We just launched the AI ticket to Social media marketing world. It's like a conference within a conference. If you want to go deep in your AI learning, this is the live experience you've been waiting for. Imagine spending two days fully immersed in AI training. Grab your tickets now at social mediaexaminer.com Aicon.
- [00:28]
Michael Stelzner
Welcome to the AI Explored podcast, helping you put AI to work. And now, here's your host, Michael Stelzner.
- [00:39]
Nicole Leffer
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 and business owners who want to know how to use AI. Today I have a great show for you. I'm going to be joined by Nicole Leffer, and we're going to talk about building fundamental AI skills for business. Now, I will tell you that I have been at this for two years, and I learned an enormous amount of things today that I did not know. So if you are someone who's been utilizing AI for quite a while, even though the title of it is Fundamentals, and it's about fundamentals, there are some really cool stuff that we're going to talk about today that I think you definitely do not want to miss. Also, if you're new to this podcast, be sure to follow this show on whatever podcast app you're listening to. We've got some great content coming your way. Let's transition over to today's interview with.
- [01:43]
Michael Stelzner
Nicole Leffert, helping you simplify your AI journey. Here is this week's expert guide.
- [01:52]
Nicole Leffer
Today, I'm very excited to be joined by Nicole Leffert. If you don't know who Nicole is, she's an AI advisor who helps B2B marketers improve their result with AI. Nicole specializes in AI training for marketing teams. Her course is foundations in generative AI for B2B marketing. Nicole, welcome to the show. How you doing today?
- [02:12]
Michael Stelzner
Thank you so much. I'm doing great. I'm really excited to talk to you today.
- [02:15]
Nicole Leffer
Well, I am as well. And Nicole and I are going to explore how to level up your AI skills, utilizing kind of Nicole's framework, if you will. Before we go there, I want to know how you got into AI. There's a story there. I'm sure of it. So start wherever you want to start.
- [02:29]
Michael Stelzner
There's always a story there, I think. So. For me, my story was not actually all that intentional. I was the head of marketing at a B2B SaaS company and I had just started at this company, this was in 2021, and going, what is it like that I really need to set my team apart? We were a startup, I really knew we had to be scrappy. And I was on my Facebook account one day and saw ad for an AI writing tool and clicked on it.
- [02:58]
Nicole Leffer
Probably Jasper, right?
- [02:59]
Michael Stelzner
It actually was not Jasper. It was a tool called Any Word. And I clicked on it, did like, signed up for the free trial, was like, this is so cool. I think I used up my like free allotment almost instantly, got on a demo with them and brought this tool on for my team and instantly became obsessed. Now this was 2021, this was pre ChatGPT or any of that kind of stuff. You know, it was not something we were talking about all that much back then. And I started it every day. I had everyone on my team using it every day. And at that point the tool was like pretty much brand new. It was designed to help you write like Facebook posts or something. I mean, it was not like there was a lot there. And I went, I don't care what it's made for, we are figuring out how to kind of hack around what it's made for and use this for all the things. So we were using their interface for Facebook posts. It was writing us email subject lines, it was writing us and it was very data driven. And how their interface was, it was just really cool. And we saw really great results. They kept adding features and functionality. We would immediately figure out how do we take advantage of this. So I just got really excited about this technology. And about a year later I went separate ways from that company and was like, I need to decide what I'm going to do next, like, where am I going to go? And I was like, this AI stuff is really fun. So while I'm figuring that out, I'm going to dive into every other tool I can find that exists, understand how it works, all of these things. So got a lot deeper. And all of this, meanwhile, I am sharing in a CMO community that I am in on Slack. And it had always been crickets. It was just crickets, still crickets. And then ChatGPT came out. I was on in the first day or two, three days, I don't know. But I was on very quickly because I was in that world, playing with it, experimenting with it. I laugh now because I have like old Instagram stories. That was just with my friends. I was like, you can't believe what I'm doing. It was like I had to have been one of the first people doing that kind of stuff and sharing all of this in the CMO community. And then it started getting a little traction and people started going, we don't know anything about this, but I think that Nicole Leffer does. And so one day a CMO in the community was like, would you mind doing a zoom for a few of us just to share with us what you know about this technology for marketing? I'm like, yeah, sure, three or four people, sure. And I think 100 people signed up for that Zoom. 100 CMOs. And I was like, oh, I better put together a slide deck. I better like take this seriously. Put together a deck, got on, basically trained all of these CMOs on how to use ChatGPT for marketing. And the next thing I knew, I was getting messages on LinkedIn, on Slack, would you come do exactly what you did for my team? I'll pay you.
- [05:43]
Nicole Leffer
Wow.
- [05:46]
Michael Stelzner
Yeah, okay, that sounds fun.
- [05:48]
Nicole Leffer
Let me guess, this is probably very early 2023 kind of thing.
- [05:51]
Michael Stelzner
Yeah, this was very early 2023 and I just started spreading kind of word of mouth quickly, like, she'll come in and do this with your team. And suddenly I was like, you know what? I don't think I want to be a CMO anymore. This is so much fun. It's so exciting. And so I have been doing it full time since the beginning of 2023 and I've worked with companies from series A startups to Fortune 50 companies, household names, some of the biggest brands in the world. And it's been really fun because I've help so many marketers adopt this technology and start understanding how to use it. But I also get to see like all of these creative people and how they're putting it to use and where it's working for them and where it's not. And it has just, it's been a whirlwind, but it has been so much fun. And it's obviously way beyond chatgpt now. I mean, there's a lot of other tools we're working with too, but.
- [06:45]
Nicole Leffer
Oh yeah, I'm curious about today, where you're at with like the kinds of things you're doing for these businesses. Like tell us about you've got a course and what kind of training do you do today? Because obviously a lot. It's true that still there's a bazillion businesses that have not yet adopted this. But what kind of stuff are you doing now?
- [07:02]
Michael Stelzner
I have a variety of things I'm doing depending on where they were. So the main staple of what I do is webinar trainings that are remote webinars for marketing teams. So this is basically what I've been doing from the beginning. Of course they have evolved over time, but with these I will come in and whether it is a just a standard like you know, Chad GPT or generative AI for marketing, or I do customized versions, I have like different levels of customization. So it really depends on the company and their needs. And they could be beginner level, they could be kind of mixed level. Which is what most teams are, is they have people who've never used it and they have people who are pretty into it. And then I also have some teams I'm working with on some more advanced stuff. So I've done, you know, like workshops and consulting on like how to build out like very specific kinds of GPTs for or automations or that kind of thing. So really depends on the company, which is what makes my job so fun.
- [08:01]
Nicole Leffer
That is so cool.
- [08:01]
Michael Stelzner
And then I also have companies I work with on an ongoing basis on like consulting, how to like really start bringing this in. Because while a lot of companies have started dabbling in it at this point, especially in the tech space, which is a lot of what I'm working with, there's still a lot of companies that even if they think they've started dabbling with it, they actually have barely even touched it. Or there's a couple people that have like gone deep in it. But the vast majority, like haven't used it for much more than like, hey, write an email, subject line kind of a thing. So huge variety. But it is so much fun. And then I do have a course that I came out with because, you know, everybody I was working with were these companies and their teams and I was like, I want to do something that's for individual marketers. You know, I had started getting a huge following on LinkedIn. LinkedIn. So I want to do something that individual people can take too. And I created it for that. And now I've actually noticed I created my own biggest competition, I think because I actually have a lot of CMOs going. I want your course, which is like seven hours of on demand training online. They want their course for their entire team. So now I have companies having licenses for their entire team. So it has been a whirlwind and who knows, you know, where it's going. I'm kind of riding the air wave and seeing how it evolves and loving every minute of it.
- [09:21]
Nicole Leffer
Very cool. You know, it's true that the world needs people to help them understand how to use this thing because I keep telling everyone it's really just the equivalent of a flashing cursor. You don't really know what you can do with this thing until you begin to start messing around with it. So what do you want to say to marketers right now or businesses that are listening in general that are maybe what's the biggest advantage if they get this right? What's the unlock that can happen because you've worked with so many different teams? What is like the other side of it once someone has actually figured this.
- [09:51]
Michael Stelzner
Stuff out first, I want to say there's not a point where you're going to like, truly have it figured out because it changes so quickly that there's always going to be something to continue learning. So just like there's no end point to that learning. So I just want to make sure everybody knows that. But the biggest unlocks, I mean, there's three, right? The biggest, I think a lot of people understand is efficiency. You can get more done in the same amount of time. Like, I probably would have had to hire five people by now in my own business to do what I can do by myself every single day. And I work hard. Like, I'm not going to say I'm not working really, really hard, but I also, like, can take two weeks off and go to Europe. Like, it's not like I have no life at all. So you can just do so much more every single day because of this technology. So that's the big one. I think people recognize the next thing is that it actually improves, proves quality. So, like, actually I'm seeing consistently that once you understand how to use the technology, not only are you getting more done, you're getting better done, and the quality of what you're doing is improving significantly. And then along with that, it, like, I guess it sort of goes with both of those things. But you are able to open up more things that you can do because you don't have to necessarily bring in experts that are really expensive to help you out. Like I as an individual can have in my pocket any kind of consultant I can imagine to help me with any kind of project I'm working on. And being able to, like, do things myself that I have never had the capability to do before is a huge unlock with the tech.
- [11:29]
Nicole Leffer
I agree completely and it's absolutely fascinating. It kind of reminds me of the early days of the Internet, right? I mean, like, I'm old enough to remember what it Was like before the Internet. And then once I got on the Internet, I'm like, are you kidding me? I can search almost anything. I can imagine. I don't have to go to a library to get this stuff. I mean, it's kind of that kind of mental unlock. Once you begin to understand what this thing can do, it's crazy. So you've developed a really good process to kind of help people develop skills, solid skills really with AI. So I'd love to kind of start with the beginning step of your process. Where does one begin if they want to try to, for lack of better words, level up their AI usage?
- [12:09]
Michael Stelzner
So the first thing you need to do, and this applies no matter what quote, unquote, use cases you're talking about. Also, this is not like a specific use case based thing. You need to learn how to communicate with these tools, right? Like what are the important things that go into a prompt? And I hate the term prompt engineering. Like I wish we would just stop with that because prompting is just communicating. Marketers are actually really good at prompting because they're good communicators if they can get out of their head like this tech sounding language of prompt engineering. But getting really good at communicating with the tools and understanding what kind of context does it need to do a good job on what you want it to do? Well, how do you give it really clear instructions, how do you structure prompts to get good responses? And what are the kind of things you need to ask of it if you want it to give you what you want on the other end? So the first thing that you need to learn about using generative AI is how to talk to the AI in a way that works for you. And that's a really important key too is your communication style is going to be unique to you personally if you want it to work really well. So how do you develop your unique communication style with the AI that's going to generate the results that are what you're looking for? Skill number one that everybody needs to learn.
- [13:31]
Nicole Leffer
Okay, so let's talk about this a little bit because I've had a lot of people on the show and they've all got different perspectives on how to structure a prompt. And you talked about context a little bit. But even though it's true everybody has their own thing, what are you finding to be some of the through lines, if you will, that would be beneficial for anyone who's just getting started with this.
- [13:52]
Michael Stelzner
So the first thing is like overall, I would say the order you put things in, usually that can be really personal, like as long as it's logical, just like you would expect. Explain a story different than I would explain a story. It doesn't mean one is better than the other as far as somebody understanding it. The order you go in overall does not matter with one exception. I will put one exception to that, and that is that you should always be starting your conversation. Your prompt of the start of a conversation, telling the AI who it is in the context of this conversation. And you want to think about who would be. If I could work on what I'm about to work on with literally anybody on earth, who would I want that person to be that's working on this with me? And literally tell the AI at the beginning of the conversation. Give it this Persona. You are the best social media marketer with specialized expertise in LinkedIn B2B copywriting in the world and you have deep knowledge of the healthcare tech industry. Whatever it is, get really specific. Give it skills, give it expertise. All of the things that somebody would need to do a phenomenal job at whatever you're going to ask of it.
- [15:08]
Nicole Leffer
Real quick question on that. I use act as if you are or act like. Does it really matter? It probably doesn't matter.
- [15:15]
Michael Stelzner
Really matters.
- [15:16]
Nicole Leffer
Oh, you should say you are instead of act like you are.
- [15:19]
Michael Stelzner
I have ab tested the heck out of this in multiple tools.
- [15:22]
Nicole Leffer
Oh, that's good to know.
- [15:24]
Michael Stelzner
Actually saying you are. And let me tell you, I have a hypothesis of why this is and it gets to the heart of how incredibly smart these tools are and why understanding how they think and how they work matters. If I tell you to act as something, I am telling you to mimic and pretend to be something.
- [15:45]
Nicole Leffer
Oh yeah, good point.
- [15:46]
Michael Stelzner
If I tell you you are something.
- [15:49]
Nicole Leffer
Then you actually personify it. Okay.
- [15:51]
Michael Stelzner
You are that you are going to behave. So if you're saying act as this, you're telling it to be somebody pretending to be an expert.
- [16:00]
Nicole Leffer
Okay.
- [16:00]
Michael Stelzner
But if you tell it, you are an expert and you will see that nuance of difference in how it handles whatever you ask of it. So that subtle difference in the language actually affects the outputs that you're going to get at the end of the day. Now act as will be better than having nothing on there to tell it that context of what your expectations are. But you are versus act as is a very big difference.
- [16:28]
Nicole Leffer
Cool. Any other little things that maybe we need to unlearn other than just you are.
- [16:33]
Michael Stelzner
So, okay, so that is the first one that I think because it's like you gotta start your prompts that way. And then once you have that, like, then it can get to be more personalized communication style, I would say, of unlearning. There has been some really bad viral advice not to say please to the AI, and that's awful advice. So if you do not say please, you need to unlearn that. There's actually studies that have come out that it improves the quality of the response. What about thank you, please and thank you? Yes, anything. Being polite and nice to the AI. And so, like, you know, some people, they're naturally going to say please and thank you. Other people, they're going to have to learn to say please and thank you to the AI. In general, the more human you treat it, the better it's going to be. If you actually, I know it's maybe not the best psychologically for us to anthropomorphize. Anthropomorphize and pretend the AI is real. But if you talk to it like a person, think of it like a person, use human psychology on it. The more human the psychology you use with it, the better it's going to do in general. And consistently, study after study is showing, like, different things that work with people also work with the AI. One of the other things, if we're talking unlearn, is don't believe every single tip that you get out there of, like, the little prompting hacks, some of them are phenomenal. But a lot of these people do not actually understand why what they're working works in the context it works in.
- [18:04]
Nicole Leffer
Okay.
- [18:04]
Michael Stelzner
And one of the things I have discovered, so I was a psychology major in college, not a business major or tech or anything like that. I was a psych major, and I have noticed that these tools really do respond just like a human would like a lot of psychology. And like a lot of psychology with humans, context matters. The task matters. So when you tell it in certain contexts, like, there's been, like, stuff that's gone around, like, you should offer the AI a tip. Like, you should offer it financial benefit in order because it's going to respond.
- [18:38]
Nicole Leffer
And we had Russ Summers on the show who said, hey, go out and get yourself a coffee, right?
- [18:44]
Michael Stelzner
So actually, like, those kinds of things, they only work in a context that would work with a human. Like, there's actual psychological studies that sometimes money as a reward for certain types of tasks actually decreases the quality of that task. And the same thing happens with the AI. Okay, I tell the AI you have had the perfect amount of coffee and caffeine to work at peak performance. In a lot of my prompts because if it's well caffeinated and you got the perfect amount of sleep, like, I'll put this into my prompts. But you don't want to do that in a creative prompt because there are, it actually believes and there are psychological studies on humans that do this, which is why the AI believes this. That like being ideally caffeinated or too much sleep can actually decrease creativity. So, like, you have to kind of understand that things work in some contexts and not others. It depends what you're doing. That's getting a little advanced on the prompting to get to that point.
- [19:44]
Nicole Leffer
Hey, that is so cool. I mean, I have some people who have been listening to this from episode one and this, this is like their mind's getting blown. So this is really cool. So, I mean, the key take home, I'm hearing you say, here is a couple things. Number one, please and thank you will actually result in a better output, just like it does with real humans. Number two, if you do things with humans, you should try with the AI because the AI is a neural network essentially, that is trained kind of like the brain works. And what about your thoughts on take your time? Have you tried that one too?
- [20:13]
Michael Stelzner
That's another one. So usually I'll tell you how I structure my prompts in general. And again with the caveat that aside from that, you are to start setting the stage. You gotta find your own formula and everything that works for you. But what works for me and when I'm writing my prompts, you are the best XYZ in the world with this specialized skills and expertise. Then I tell it like, hey, this is what we're gonna do together. Usually like, you know, we're gonna be doing. This is the a little context of the situation. This is what we're gonna be doing together. Here is how you to do it, step by step. And then I will break down how. Now, I don't do this if it's a super simple thing, but if it is like remotely complex, I will give it the chain of thinking that I want it to work through. So first I want you to go visit this website. I want you to read this context. Then I want you to read the information I'm going to provide you later about the Persona that we're serving. Then I want you to think through this document I've given you and like, put all of these pieces together. And I want you to write a plan for X, Y, Z about this. And I will go through step by step, exactly what I want it to think through and how I want it to think through it.
- [21:28]
Nicole Leffer
Is this called prompt chaining? Because I've heard people use that phrase.
- [21:31]
Michael Stelzner
Prompt chaining is multiple prompts getting stacked together. This is chain of thought.
- [21:36]
Nicole Leffer
Okay, Got it.
- [21:36]
Michael Stelzner
Prompting. So. And don't get caught up too much in, like, lingo, but if you want.
- [21:41]
Nicole Leffer
To learn more about it, you could search chain of thought and maybe learn more about this if you wanted to.
- [21:45]
Michael Stelzner
Yeah. So if you research chain of thought, there are actual, like, studies on this. But I will tell it. Take a deep breath and do the following in order.
- [21:54]
Nicole Leffer
Okay.
- [21:54]
Michael Stelzner
And then I'll give it the steps of what to do. Then I will tell it. This task is vital to my career and I really appreciate your attention to doing, like, a really good job on it. Some kind of, like a psychological motivator that this actually matters. It's important to me. Not just some random task I've thrown at you, but it actually is going to have an impact on my life. Then I will. Yeah. At the end of it, like, I'll thank it and everything for what we're doing. And then I will usually give it whatever context I have promised it earlier in the prompt. Because I'll tell it, hey, at the end, I'm going to give you sections of context. So maybe that's my Persona. Maybe it's the ideas. I want it to turn into something. Maybe that is like history of something. Whatever context it might need. That usually drops at the end of my prompt. This is just kind of a format I've gotten. Gotten used to, like, cranking out. It works really well. There's elements throughout this that are well researched. The you are statement, the chain of thought thinking, the take a deep breath pause, don't rush this. The emotional motivation to do a great job. Like, all of this is very well studied. And then I've tested and figured out how it works for me, how to put it all together. And then I'm using. This is very advanced, but I'm. And I actually have a free video on it on my website if anybody wants to know more about what I'm about to talk about. But I use what are called delimiters, which are a way to give the prompt structure so that it understands different chunks of thought, essentially.
- [23:32]
Nicole Leffer
Yeah, that's like the thing below the question mark, the little slash, slash kind of thing. Is that what you mean by a delimiter or physical?
- [23:37]
Michael Stelzner
Like, you can use any kind of symbol that you want. A lot of people are naturally using delimiters with quotation Marks. Like you put quotation marks around a block of thoughts and you're using, oh, there you go, that's a delimiter. The problem with quotation marks as using it as in your prompt prompts, you can, it'll understand it, but you might have quotes within your quotes and like it gets confused what the thought is, what's what. So I use like three slashes in a row at the beginning and the end of blocks of context. If anybody's interested, it's totally free. There's a video on my website in my resource section about how to use delimiters in your prompts. Please feel free to go watch that because it kind of takes some visual to really understand it.
- [24:18]
Nicole Leffer
Sweet. We'll make sure we get that in the show notes as well.
- [24:20]
Michael Stelzner
Yeah, absolutely. And then once you've got that, you know, I've got a good prompt, I'll submit it. And then prompt chaining comes when maybe you have multiple steps. You kind of need it to think through one thing, like actually do the step before you now move to the next step. And that's where you're actually chaining multiple prompts together.
- [24:40]
Nicole Leffer
Many of the top experts you've heard on this show will be speaking at Social Media Marketing World 2025. And with your AI tickets, you can attend at a very economical price. You'll discover practical AI workflows and advanced AI automations that will increase your productivity and save you time. Imagine getting live and in person training from Matt Wolf, Chris Penn, Jeff J. Hunter, Rachel Woods, Molly Mahoney, Brian Piper, Jeff C. And many others. Isn't it time to enhance your career by fully embracing AI? Grab your tickets now at socialmediaexaminer.com Aicon well, I love this. First of all, we've gone down a really fascinating rabbit trail that I know a lot of people are like, you could stop there, it was that good. But we're not done, everybody. There's another step to your process, right, which is this is just the first step, really understanding how to communicate with the tool. What's the next part of your process?
- [25:40]
Michael Stelzner
So that communication learning, like that's really for all tools, right? Like you need to learn how to communicate with any of them. And then the next I would be learning right now is what are the like core features that go through all of these different tools and which one does the tool you're using have? And like this almost the skills, the senses that it has, the ability to leverage and use. So these are things like inputs, PDF files, word documents, like what Can I upload to the tool? And also how do they work? If you, as of the time you and I are talking, like, this could change by the time somebody's listening to them, this. But if you upload a PDF file, for example, to ChatGPT or Claude or one of these tools, they cannot see the images in that file. So, like, you need to understand that it can read all of the words, but it cannot see any of the visual information. That's important kinds of stuff to understand about how these tools work to get the most out of them.
- [26:43]
Nicole Leffer
Well, and you could test that, right. Obviously in the future. Like, you could put a PDF with graphics in there and just ask it what it sees, right?
- [26:51]
Michael Stelzner
Yes.
- [26:51]
Nicole Leffer
Or to describe what it is. Right. And see if it's able to do that. But the good news is you could put screenshots in there, right?
- [26:57]
Michael Stelzner
Exactly. So that's the next thing you do need to understand is does your tool have vision ability? And vision is where it can see an image and it can look at it and understand it like a person can. And so that can come as like a JPEG or a PNG file or a screenshot, which. Which may technically be one of those formats, depending on where you're taking it, but you could just paste in a screenshot. So if you're uploading, like, let's say you have an ebook that you've written that's your company's, it's got some infographics and stuff in it, and you want to repurpose that into other content you need to understand. I can give it the PDF file, but those infographics I need to take screenshots of and also give it the screenshots and give it the context of, like, these are the images in that file that you're looking at and ask it to visually look at them and understand them. It can understand color layout like anything a person would. Different tools are at different quality of being able to do this, like how good it is at it. But know if your tool can do that, ChatGPT can quad can Gemini can the biggies all can do that right now. But you need to understand that you need to know, can it connect to the Internet, right? Like, how does it work when it connects to the Internet? Is it doing. Can it give me source links, can it search the Internet or only go to a link that I give it? Understanding that can I talk to it? Can it talk to me? Like all of these things when I talk to it, what is happening? Is it just taking what I'VE said and transcribing it, or does it have the ability to listen and hear and understand intonation and emotion in my voice? Voice. These tools are evolving. They are getting to the point that they can do that. So all of this fundamental understanding on the inputs and then also on the outputs.
- [28:54]
Nicole Leffer
Now real quick on the inputs, do you know, because these tools aren't really good with their marketing. None of them are. They're horrible. So is there a question we could ask it to get it to tell us what it's capable of or do you feel like that doesn't really work and you just have to try it?
- [29:08]
Michael Stelzner
A lot of it, it doesn't know what it's capable of.
- [29:11]
Nicole Leffer
I see.
- [29:11]
Michael Stelzner
Sometimes it does. So like you can ask it, like, don't necessarily believe it. Sometimes it'll tell you it can't do something and all you have to say back is, yes, you can. I've seen you do it before. And it'll be like, oh, okay, it'll do it. A lot of times they underestimate their own capabilities. And also there is a thing with all of these, all of these tools for them, as far as I know, I mean, pretty much anything you're probably using, there is going to be a gap between when it was trained through of like when it read the entire Internet and learned all these things. There is going to be a gap in time between when that training cutoff was and when you're using it. And because of that, you need to understand that if you're asking it, what can you do? ChatGPT is probably trained on the Internet and there's probably been several months since it was trained and it knows what it was trained on several months ago, not what it can necessarily do today. Sometimes it understands its capabilities, sometimes it does not. But that training cutoff can be challenging. So if you're using any of the big tools and when I say the big tools, I'm talking like OpenAI anthropic, which is Quad, Google, Gemini, Google's products or Microsoft's products, you know, like those are like really the biggies. If you're using any of those, like their own social media, like they may not have have the best marketing in the world. They're getting better at it. Especially like OpenAI's marketing has been getting significantly better. But if you go and look at like their, like just follow their social media and their blog posts, they'll release this stuff. But also you just, you pay attention to people like me who are really paying attention to this stuff. And you know, we tend to update when new things come out, but it is challenging to keep up with what it's truly capable of. It is worth taking the time to learn because so many people are going out and buying purpose built tools as extra subscriptions of things that they could totally just easily do in their ChatGPT account or their cloud account. Probably have better results than with the purpose built tool.
- [31:21]
Nicole Leffer
One of the things that is kind of a fundamental difference between Claude and the others is they generally do not search the Internet. They seem to have an ethics policy that's a little cleaner. Probably were trained on the Internet. But what I found is it will, if you give it a link to an image, it'll pull the image and read the image. But beyond that you can't tell it to go out and train on a data set like you can with ChatGPT, unlikely with Gemini. And that's kind of one of the key differentiators. And I've also noticed that the Chat GPT seems to be a little bit more advanced at the multiple modes of input. Right. Like it's one of the first that has voice input. I'm sure Claude is coming. But it seems like ChatGPT, because it can search the Internet, has the capability of being up to the minute depending on how you use it. Do you find that to be true?
- [32:07]
Michael Stelzner
So any of these tools that search the Internet, it is important to understand and like, hey, by the time somebody's listening to us, who knows Claude, Claude might have the ability to search the Internet at that point. So we know.
- [32:19]
Nicole Leffer
But what I meant is if you tell it to search the Internet, it could get up to the, if you give it a good data source. Right.
- [32:24]
Michael Stelzner
What's going to happen is it's not actually updating the training. Okay. So it is going to be the same if you say, hey Chad, GPT, I need you to go and like search the Internet and get information about Taylor Swift's heiress tour in 2024 and like all the things that have happened and like I want to know like late 2024 the things that happened and like what surprise songs did she. I don't know, whatever you want to get on this, what it's going to do is it's going to go do an Internet search, use it using Bing and it's, you know, and Google might do it using Google and it's going to read a page or two or three and it's going to take that and it's going to incorporate it into its response. What it is not going to do is read the entire Internet of everything on that topic and be up to date in its training on that topic. So if you want it to know something specific that is since it's, it does know it's data training cutoff. So if you ask it at any given time, like hey, when or you trained through of whatever model you're using, it should be able to tell you if you need to now. But if you do that, it's not going to be like now I know everything on that topic.
- [33:33]
Nicole Leffer
It's probably going to forget about it later is what you're saying. It's not going to keep it in its long term memory. Right?
- [33:38]
Michael Stelzner
It's not going to keep it. It's a long term memory, but it's also not going to be fully versed in it.
- [33:43]
Nicole Leffer
Right, right.
- [33:43]
Michael Stelzner
So like if I wanted to ask it about like the hurricane, like there's been some major hurricanes this year. If I wanted to ask it about, tell me about like what's going on with the recovery in North Carolina after Helene, it's going to go and it's going to read a page or two, a news article or two, maybe three and then it's going to come back and give me that answer. But that does not mean it knows everything that's on the Internet that's happening on that topic. It's no different than whatever. If you did a search and you happened to read a few things and you were like, okay, now I understand from the perspective of those sources versus if you're talking like things that are more in the past where it's been trained on the entire Internet, essentially it has a much wider corpus of knowledge that it's pulling from about every topic so it can talk about it a little bit more. So it's a subtle difference, but it's important because it's not really that different than if you just like copied and pasted that article in. Versus like it actually having the knowledge of the topic topic when you're using it for Internet searching.
- [34:47]
Nicole Leffer
Do you feel like even Google's Gemini is the same? Like they have cutoff dates. They're not constantly learning. They just.
- [34:52]
Michael Stelzner
I don't know if Gemini will tell you because I don't know if I've asked it about the cutoff dates. But in my experience it does not seem like their training is 100% up to date. There is a reason that these companies do that. It is largely a safety thing. So what happens is when they go out and they train the models with like they put in all of this new data. It's like doing all of this teaching. They need to go in and they need to safety test which they have people that actually have conversations with the AI and try to get the AI.
- [35:24]
Nicole Leffer
The red team.
- [35:25]
Michael Stelzner
Yeah, the red team. And they try to get it to do all the things it's not supposed to do. And you should actually be really glad they don't just release new models week that are up to date completely without doing that. Because this is how. Yeah, sometimes it's annoying, right? I wanted the AI to do something for me with my dad the other day. I was like trying to get it to make like sound effects for me on ChatGPT and it was just like, I can't do that. And I was like, you used to be able to do that, but it would just refuse and refuse safety parameters they put in there. It was annoying because I wasn't trying to use it for harm, but that is there to prevent people from doing harmful things with this technology that are bad for all of us. So we're actually glad that they do this red teaming and that they have this lag time. And I just don't foresee in anytime soon that this lag time completely goes away because of that safety factor.
- [36:19]
Nicole Leffer
Okay, so we spent a lot of time talking about all the different forms of input, which includes voice, text files, the ability to see things, images and all that kind of fun stuff. The output, surprisingly for a lot of people, they don't realize that you can do a lot more than just text. So talk to us a little bit about that.
- [36:37]
Michael Stelzner
Yeah, so that's the next thing you need to understand is how can I get my outputs? You can do beyond the text, you can do tables which just like in line, within which that's pretty simple. But you could also ask it to create, depending on the tool you're using, you could maybe create code, you might be able to create images, you might be able to create infographics. So like actually running like Python code and creating graphics, graphics out of that code. So graphs, word clouds, that kind of thing. Or the graphics could be artistic coming through an image generator, more like dolly type of a image generator, if you're familiar with that. So like chat GPT can make you more artistic. Images or photos, things like that. With some of these tools you can create audio. So actual like AI voiceovers are a possibility. You might be able to create Word documents or Google Documents or CSV files or Excel files or PDF files. That's the kind of stuff you should have an awareness. If you have a need, figure out what kind of stuff your tool can create besides just writing text for you. And also with text, realize, like, you maybe need it to be formatted in markdown, or you maybe need it to be formatted in HTML. Like, it could write you that text in different formats beyond just plain text text. It could give you formatting headlines, sub headlines, bullet points. Just direct it how you want that response, and it will give it to you however you want it.
- [38:06]
Nicole Leffer
I love it. That is so cool. And it's starting to blow people's minds a little bit, right? Because it's kind of capable of doing just about anything. I mean, I haven't yet seen the ability to drop an audio file in, but I'm sure it's coming in the large language models.
- [38:23]
Michael Stelzner
Oh, you can do that.
- [38:23]
Nicole Leffer
You can. You can take like an MP3 file and drop it in there.
- [38:26]
Michael Stelzner
Not inside ChatGPT. Well, it depends on what you want to do with it.
- [38:30]
Nicole Leffer
Well, I mean, if. If I wanted to listen to it. You know what I mean? But I know you can give it a YouTube link and it will list, but I'm thinking it just pulls the transcript.
- [38:38]
Michael Stelzner
YouTube is pulling the transcript. We're talking today. You can't, like, put it in and have it listen inside ChatGPT, but you could easily spin up a automation that uses their Whisper transcription tool, and it can do that. Now you can do. You can get like a transition transcript in the app. So you could just play your audio file and have it listen.
- [38:58]
Nicole Leffer
Yeah, yeah, it's true, but I'm sure it's coming. It's probably just.
- [39:01]
Michael Stelzner
Oh, it's. There's no way that's not coming. I mean, that they already know how to do it. It's just like actually building it as a feature or. That's pretty basic as far as where we are right now. It's crazy how fast it comes out. But I want to tell you, you said about, like, it's starting to blow people's minds. I've seen an interesting thing with adoption and is that there people tend to go through this curve of an experience, and they start out and they're like, you know, they've never used it. They've heard the hype. They're a little skeptical, but they're like, okay, I'm gonna try this. So they, like, they're curious and they got this perception. It's like, okay, cool. Then they use it. Usually perception increases a bit when they first use it because something really cool happens they're like, whoa, I didn't realize it could do that. They use it a bit more and that perception craters, forget blowing people's minds. Then they start going, this isn't ready for prime time. It's not giving what I want. They don't know how to communicate with it. They don't know how to get what they want. They think it's overhyped. Most people are walking away at that point. I mean, I've watched this over and over and over.
- [40:02]
Nicole Leffer
Oh, and that's a serious problem. They have no idea what they're missing. Right?
- [40:05]
Michael Stelzner
They have no idea what they're missing. And then there's a segment, probably the people who like are the type of people listening to this podcast that are like, I know other people are getting good stuff out of this. So I'm going to learn. I'm going to take what I'm learning and I'm going to experiment and I'm going to experiment and I'm going to apply, see what works for me. And when that happens, you start getting higher and higher perception of the tools till you get to the point where I don't go more than a day or two without being completely mind blowing at something. I got one of these tools either because a new feature comes out or because I figure out something I could do or I have some need and I'm like, I'm just going to try and I know enough about the tool to get it to do it. And people who are heavy users don't go more than a day or two, three days without being like, my mind is completely blown with what this technology is capable of.
- [41:00]
Nicole Leffer
A hundred percent. You know, we just released our 2024 Generative AI Marketing Industry Report, and 37% of the 1254 marketers we surveyed use it daily. And when you add the weekly in there, it's like 64%. This is becoming a normal part of the workflow for marketers. And a lot of people who originally tried this and walked away from it. It's kind of like anything else that you might have dismissed. And then all of a sudden everybody's singing the praises and then you come back around for another take at it and you're like, okay, maybe I need to pay attention to this darn thing, because maybe it's gotten smarter. And that's the key thing. Everybody needs to understand it is evolving so fast. Would you agree?
- [41:38]
Michael Stelzner
Oh my gosh. It's like absolutely mind blowing how fast it is evolving. And, and like the models, I mean both by them Adding new features or coming out with new models. But also the models are just gradually getting smarter over time and they keep getting like slightly adjusted and so like it's just getting better and better. It's mind boggling to me how fast this technology is getting smarter and better and more capable. And then you compound that with if you're learning how to use it and get the smarter, more capable responses out of it. You put those two things together and it's like, oh my gosh, I literally have an expert to work on anything with me at any given time and give me skills I've never learned how to do. And it's just wild.
- [42:26]
Nicole Leffer
Well, and then the next level really. You kind of hinted already at prompt chaining, but let's talk about automation just for a little bit. We don't need to go very deep into it, but I want people to understand what that could open up.
- [42:35]
Michael Stelzner
Yeah. So I love AI automation. So with automation you're not necessarily going to use use chat, GPT or CLAUDE in the form that you're used to using them of that chat interface. Instead what you're doing is you're going to tap into the APIs for these. So it's getting a little more technical. But you could use a tool like zapier or make.com depending on what you have access to. There's some respell is one that's like AI native to do it that they made it for this. And you basically put prompt prompts in a series so the output from one flows into the next one, flows into the next one. And you can have an action that you take or do trigger these prompts to run automatically and then send the output wherever you want. So for example, I just, I have, I have lots of business ones, but forget business for a minute, let's do a fun one. I just went to Europe recently and I every time I go to like go on a trip, I'm like, I wish I journaled. I may carry a journal, but I'm so tired after, after like day of travel. I never write in my journal. So I was like, you know what? My dad said he wanted to journal. He was traveling with me and I was like, why don't I build us an AI automation so that we really do this and what we're going to do, what I ended up building was on my cell phone, on my iPhone, I would open the voice memo. So this was not in an AI tool at all, just the normal voice notes and hit record every night. And I would just ramble about my day. We woke up this morning and had a really good hotel breakfast, blah, blah, blah. You go just talk about every single thing. I would be looking through my photos so I'd remember what we did. I look at my credit card receipts, name, whatever of where we were going. And then I would email that to myself and I set up like just a Gmail that it tags it when it came to like that extension of.
- [44:25]
Nicole Leffer
My email, like Nicole plus Tripp at whatever. Right?
- [44:29]
Michael Stelzner
Yeah, it was like Nicole plus Journal or something, I don't know, something like that at my email address. And so when I sent it it, it tagged it to trigger the automation and it triggered differently. Like if it came from my email address, it triggered one. If it came from my dad's email address, it triggered another, like automation. And in the body of the email, I would just write the names of the places because it may misunderstand the name of the restaurant that's in a foreign country. So I just type them in. So it would have it in the body of the email because that fed into my prompt. But then as soon as I hit send on that voice memo with the words typed in, it would label it in my Gmail, which triggered a Zapier automation. And that zapier automation transcribed it. So listen to it, transcribed it applied all of the spelling that I had just put in the email and it wrote me a journal entry. It wrote up a bulleted list of my trip notes and what I thought of all the things that I did. And it thought through can I write travel reviews for like trip advisors or Google Maps or any of these, for any of the things she mentioned was there. That's a site, a restaurant, a hotel, based off of what she said in this can I write her reviews? And it would then put it all into a Google Doc and email me. And it. So I would get an email with a link to my journal, a link to the reviews, a link to the trip notes. So I have it in different formats and a separate email with all of the reviews in case I wanted to post them. And so it's, you know, 10 minutes every night I could ramble to my phone. And now I had all of that and I wouldn't have done it. Right. Like this is what I'm saying. Like it's opening up things you wouldn't have otherwise been able to do or done. I just wouldn't have had a trip journal, right. Like it didn't replace me journaling. It just I wouldn't have had a journal.
- [46:17]
Nicole Leffer
It enabled you to do something you didn't have time to do.
- [46:20]
Michael Stelzner
Right. And I came home and I like have a fully documented two and a half week trip to Europe. That every single thing I did, I'll have that memory. If a friend wants to go, I can just copy whatever notes and what I thought of everything and it's really accurate and it's just so cool that you can do that kind of thing now. You could do think through any kind of work process. You could do the same thing. But it's way more fun on a vacation.
- [46:43]
Nicole Leffer
Yeah. Nicole, this has been amazing. If people want to connect with you on the socials, what's your preferred platform? And then if they want to reach out to you because they're interested in some of your services, where do you want to send send them.
- [46:58]
Michael Stelzner
So LinkedIn for where I'm posting content by far, that's really my platform. So just my name on LinkedIn l.
- [47:05]
Nicole Leffer
E, f F E R for those that are listening. Yep. Okay.
- [47:08]
Michael Stelzner
So Nicole Leffer on LinkedIn and then my website is just Nicole Leffer.com and on my website you can find course info, training info for your team. You need to reach out to me, like email addresses and everything on there. You can send me messages on LinkedIn too. It may take me a little longer to like get to it on LinkedIn because I get a lot, but I do check that it might be a little bit delayed over there. So those are the two best ways to connect.
- [47:33]
Nicole Leffer
Nicole Leffer, thank you so much for sharing all your awesome insights with us today.
- [47:38]
Michael Stelzner
Absolutely. Thank you so much for having me. This was a really fun conversation.
- [47:42]
Nicole Leffer
Hey, if you missed anything, we took all the notes for you over@socialmediaexaminer.com a29. And also be sure to follow this show on your favorite podcast app so you don't miss any of our episodes coming. And if you've been a listener for a little while, would you do me a favor and give us a review on whatever app you're listening to? That would be super helpful for us and maybe let your friends know about the show. I'm Stelzner on Facebook, Stelzner on LinkedIn and ikestelsner on X. And do check out our other shows, the Social Media 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.
- [48:30]
Michael Stelzner
The AI Explored podcast is a production of Social Media Examiner.
- [48:38]
Nicole Leffer
Don't forget to get your AI ticket to Social Media Marketing World 2025. Become an AI Enhanced Marketer. Grab your tickets now at social mediaexaminer.com Aicon.