Coaching ChatGPT: Pragmatic AI for B2B Marketers

 

Drew Neisser: Hey, it’s Drew. And I’m guessing that as a podcast listener, you will also enjoy audiobooks. Well in that case, did you know the audio version of Renegade Marketing: 12 Steps to Building Unbeatable B2B Brands, was recently ranked the number one new B2B audio book by Book Authority. Kind of cool, right? Anyway, you can find my book on Audible or your favorite audio book platform.

And speaking of audio before we get into today’s show, I do want to do a shout out to the professionals that Share Your Genius. We started working with them several months ago to make this show even better, and have been blown away by their strategic and executional prowess. If you’re thinking about starting a podcast or want to turbocharge your current show, be sure to talk to Rachel Downey at shareyourgenius.com and tell her Drew sent you.

Okay, let’s get on with today’s episode.

Narrator: Welcome to Renegade Marketers Unite, possibly the best weekly podcast for CMOs and everyone else looking for innovative ways to transform their brand, drive demand, and just plain cut through. Proving that B2B does not mean boring to business. Here’s your host and Chief Marketing Renegade Drew Neisser.

Drew Neisser: Hello, Renegade Marketers Welcome to Renegade Marketers Unite the top rated podcast for B2B CMOs and other marketing obsessed individuals. Alrighty folks, you’re about to listen to a Bonus Huddle, a specially curated Huddle that we run once a month with experts sharing their insights into the topics that are most important to our CMO community.

The expert at this particular Huddle was Nicole Leffer, AI advisor for CMOs at A. Catalyst LLC. She joined us to discuss how CMOs can use generative AI to improve productivity across their department today, not sometime in the future, but today. All right. Speaking of today, let’s get to it.

I’m really thrilled to introduce you to Nicole Leffer, who 19 months ago was sitting in your shoes as Head of Marketing for Resilia—a venture backed startup when an layoff accidentally threw her into entrepreneurship. Now during your time off, she started to play with generative AI tools and quickly discovered that she had a knack, not just for using them. But for teaching other marketers how to use these tools. In January this year, she set up a consultancy A. Catalyst and has become a go-to consultant in this area. In fact, this, I think, her third talk today. And what you will discover is that Nicole is in there and blown away by everything, there’s just so much cool things happening. And the one caveat that I want to say right now, which Nicole said is when I mentioned that we were going to record this and release it as a podcast, she said you better release it fast because it’s changing so quickly. Nevertheless, there are all sorts of things. So first of all—by the way, our goal today is to be really pragmatic, how you and your teams can use these tools today to radically improve productivity. I love this, I want to share this one thing with those of you who are on camera. This is the cover of last week’s Economist, and there’s a sort of devil and an angel. We’re not going to be talking about the ethical consequences or issues or legal issues, really of it. But I want to read one quote for you. “A technology need not be world ending to be world changing.” And I think—both Nicole and I agree—this is a world changing thing. So alright, Nicole, how are you? And where are you this fine Friday afternoon.

Nicole Leffer: Hi, I’m great, thank you. I am in Atlanta. And I do want to just clarify one thing, I used AI on my team that I was leading for a solid year so I have experience as the head of marketing leveraging it on my team.

Drew Neisser: Okay, there you go. Good. Thank you for that now. And speaking of credentials, you also got a credential for the Marketing AI Institute. What’s that about?

Nicole Leffer: So they do amazing courses. I took a piloting AI for marketers, I believe it was called, with them amazing courses. If you’re just digging in they also Paul Ritzer, who is the head of the Marketing AI Institute has a really great podcast as well. They have a great Slack community. So it is a great place to learn a lot about AI and marketing.

Drew Neisser: Cool. I may ask you for an introduction so that we can get him to do a podcast because again, we could probably do this six weeks from now, and have a pretty radically different conversation.

Nicole Leffer: Yeah, and we have super different perspectives to it just because we’re using it in different ways and have different experiences. So they’re great over there. Highly recommend following them as well.

Drew Neisser: Okay, so I promised because we’ve already had a broad conversation about generative AI, that this would be super pragmatic. And so how CMOs can use generative AI right now, I would say 99% of the Huddlers have played with ChatGPT, okay. And they’re thinking about it for content creation. So, share with the group here, because you and I talked about how you create blog posts. Talk about that.

Kay Moffett: Yeah. So first off, I just wanna say like, you can use AI for basically anything at this point. We’re getting—maybe not 100% of stuff—but I have started going to this perspective of how can I use AI before I start anything, instead of trying to think of what can I use AI for I flipped that script. And it has made a huge difference in realizing the full mass potential. But if I want to create something like a LinkedIn post, or a blog post or thought leadership, what I am doing is—it’s really my thinking. And I think that’s one of the big differences between the people who are getting great content out of AI and tools like ChatGPT, versus those who are not very impressed with it. Or it’s really using your own ideas and your own thoughts feeding that in in one form or another whether that’s past content, whether that is recording a voice memo, just rambling about your ideas and your thoughts, uploading and getting a transcript of that and putting that in to base your content off of—I love doing that. Or giving yourself like podcast transcript, anything like that, but your own ideas, your own thoughts and taking that,we’ll use ChatGPT for the example since that’s what most people are using. I’ll write a prompt that tells ChatGPT exactly who it is, you are the world’s best copywriter for blogs with an expertise in getting people to read and share this content and really get high value from it. And then getting into this is what you’re going to write a blog post about, giving it the ideas that you want it to be based on, and giving it the voice you want it to be in. So providing that with the description of how you write, you’ve already trained the AI on how you write and giving it that description, giving it the audience you’re giving it to, and then taking it step by step asked you to write the outline of this blog post first, then ask it—once you’ve got that—you’ve got to go, now write me section one, then write me a section two. Do not just say, write me a blog post about AI. It’s writing me a blog post with my ideas about AI in my voice for my a CMO audience who’s looking to incorporate this into their teams, and you get a completely different output on that.

Drew Neisser: I’m going to recap real quick. So it’s all about the prompts. Starting with your ideas, you’re telling who you’re talking to, and who you are, and the outline and then rolls it over.

One of the things that you and I talked about was voice and how you train it to speak in your voice in such a way that you have a prompt that describes your voice. Can you clarify on that? Because I that’s one thing I missed. I just hadn’t picked up on.

Nicole Leffer: Yeah, absolutely. So what I recommend doing is write a prompt that says something along the lines of like, you are an AI that specializes in analyzing texts for voice and tone and style and all of these kinds of things. Using natural language processing—which the AI will know what that is, you can call it NLP also—then I want you to analyze the the text below that I’m going to give you and I want you to write an description of this voice for a future AI system to be able to write based off of, and then you’re gonna give it your text, but you don’t want to give it all different kinds of text, you want to give it a style of blog posts that you want to be able to mimic or a style LinkedIn posts, you want to be able to mimic, or your website, copy whatever it is. So say we’re doing a blog post, you would write that prompt, give it an example of a blog post, hit send, and it’s going to come back with a paragraph that describes the style, the tone, the type of language, all of these elements that you would never be able to describe for yourself of what your writing does. And then you’re going to save that paragraph tags. And when you go to write a blog post, you’re gonna share, say, “Write in this voice” and you’re gonna give it that context of how you want it to write for blog posts.

Drew Neisser: So you’re gonna have a Google Doc—or whatever you use—and on the Google Doc and this is going to be your prompts Google Doc. And the one is your brand voice which you have trained it to do and describe it—that was the part that I was missing. It was like you can’t say “Hey, talk like Jerry Seinfeld” because it’ll go “Hey, I’m Jerry Seinfeld” haha, that’s it. Okay, before we keep going Kay Moffett, you raised your hand, take yourself off mute and ask Nicole.

Kay Moffett:  I think this is fascinating. So I think that was my question is—are you feeding it your own data? To me AI, it’s all about the data. Garbage in, garbage out. And that’s sort of partly what’s creepy about ChatGPT. You don’t know what data it’s pulling from. So I guess I didn’t realize you can feed it your own data?

Nicole Leffer: Yes. So I’ll just copy and paste it in and it does not have internet access. So you can’t just give it a link and say, “Go to this blog post.” You need to actually copy and paste the blog post. And if you tell it to go to a link, and it gives you anything back, it’s just making it up. Or it’s telling you what was on that page prior to September of 2021.

Kay Moffett: Okay, so how do you deal with source materials, I guess. If you’re like, I want to write a blog, in this voice, about marketing AI, like and you want it to use your voice.

Nicole Leffer: So the first thing you’re gonna want to do is before you even get into trying to write a blog post, you want to train on your blog voice. So that’s what I was just saying, like you write that prompt, you share your blog post, and it’s gonna give you this description. If you want that voice description, you just keep in a Google Doc, that is for blog posts, if you do different styles of blog posts, have them for different styles of blog posts, have one for LinkedIn posts have one for website copy, it’s not just one voice for your brand, but it’s all the different applications of your voice, assuming you write in any bit of a different way on a different source. So just keep all of those voice paragraphs in a Google Doc or Word doc, however, you want to store it, an Excel file, and then that’s just gonna get dropped in as part of your prompts and your combos, write this blog post in this voice. So that’s just like one paragraph of your entire prompts that you’re dropping in. But you’re not going to do that every time because you’ve already made it and you’re just copying and pasting that paragraph.

Drew Neisser: So let me interrupt you for a second. Because everything that we’re talking about is about the fine art of prompt crafting. And this is the skill that we’re all we’re all learning. And Nicole is way ahead of us. But so we now have a prompt that you’re going to use my LinkedIn voice, because I’ve already trained it is “X.” So you include that paragraph, right?

Nicole Leffer: Yeah, so that just goes into your prompt, it’s like a snippet portion of your prompt. And then you’re also gonna say, like under that it would say, “Now use the following ideas to draft your blog post,” and then put the ideas, it could be bullet points from a slide deck, it could be a voice transcript of a voice memo you’ve recorded, it could be part of a transcript of a podcast you recorded talking about this idea. Could be a ebook you wrote or a blog post you wrote, you know, it could be a blog post but now you’re turning it into a LinkedIn post or something like that. That’s just all going in that prompt initially. So in a typical prompt, I would have, I’m telling it, the type of expert at content creation it is. I’m telling it, the voice to write in. And I’m telling it the content to write about. And then a lot of times, I’m also telling it, who it’s talking to, as well. So this is for CMOs, or this is for whatever you sell to.

Drew Neisser: You and I talked about—was it yesterday, the day before—but we talked about you do voice recordings, and then you send it over to Whisper and then from Whisper, you copy it into ChatGPT. You can’t right now just put voice recordings into ChatGPT, Can you?

Nicole Leffer: Not today.

Drew Neisser: Not today.

Nicole Leffer: Not today. But let’s stand by because this stuff is evolving very, very quickly. So I don’t know, I just discovered, I just got a new thing that popped up, an alpha test that’s in my account. And I can now upload, if I’m working in this one plugin section, I can use images and I can use CSV files. So I’m not going to say that very soon that’s not going to be possible. But not right now.

Drew Neisser: Not today, okay. But there are so many good. And by the way, the interesting part here is you don’t even need a great transcription service anymore. You could use probably just use your phone and dictate into your phone and email that to yourself and upload because the system is going to rewrite what you put in there anyway.

Nicole Leffer: Yeah. And a lot of times, like what I’ll do is I’ll just be on a walk and I’ll have an idea about something that you know, it’s coming up and I’m like this would make a great LinkedIn post. So I’ll just whip out my phone and I’ll start rambling like and when not if you were listening to me, or you listen to that voice memo, you’d be like, what on earth is she talking about? It is a stream of consciousness two or three minutes. I will email it to myself and transcribe it in Whisper as soon as I get home, they just added Whisper to the openAI Zapier integration. So you could actually set it up to email to yourself and get an email back transcribed if you wanted to.

Drew Neisser: So wait a second. So this is—I’ve dictated to myself, I email it to the system, it will then go into ChatGPT?

Nicole Leffer: It wouldn’t go to ChatGPT.

Drew Neisser: Okay. All right.

Nicole Leffer: No, but you can email it back to yourself. So like, you’ve just emailed it to you. It transcribes it and emails it back. Perfect.

Drew Neisser: Yeah, got it. I got it. So you dictate, it goes somewhere, it comes back into email, you copy that, you put it into your prompt got it?

Nicole Leffer: I’m not doing that yet. But like, it’s fair, it’s possible.

Drew Neisser: Okay. So we were talking about LinkedIn posts for a second. And while we’re there, a lot of folks use emojis. And I just never occurred to me that I could say, “Hey, ChatGPT put emojis in there.” How does that work?

Nicole Leffer: If you actually tell a ChatGPT it’s the best LinkedIn writer in the world, it will go emoji crazy.  I’ve noticed that apparently ChatGPT, the best LinkedIn copywriter in the world uses emojis every two sentences. You can just literally say, use emojis as part of your prompt, or I like to do like emojis as like bullet points in my LinkedIn post. So I’ll give it the bullet points, or it’ll have crafted it. And I’ll be like, give me five options for bullet points that are emojis that are representative of each sentence, and it’ll give me the options that maybe I symbolically never would have thought of, and they’re really, really good.

Drew Neisser: And that’s a really important point, which is, get yourself options. It’s only takes a second. And rather than taking that first draft, give it another one. Okay, Dave Bornemann, Associates Analytics. Dave, what’s your question?

Dave Bornmann: I just want to make sure I was following the very beginning. So if I have a blog post and I wanted to learn my voice, I tell it, “Hey, I want you to understand my voice from this following blog post” and it will create some written language that says “This is your voice.” And then I save that into my Google Doc, and when it’s time for me to ask it something else. That is one of my prompts that I dropped back and and say, “write in this voice?”

Nicole Leffer: Exactly.

Dave Bornmann: Okay, great.

Drew Neisser: We promised this would be practical. Last week, you released a video demonstration about the importance of editing your prompts versus the output. And that’s really hard, because I must admit, I see the output and go, “Oh, I can edit that, I can work with it.” And why is it so important to just keep tweaking your prompts?

Nicole Leffer: Yeah, so there’s a couple of reasons the way ChatGPT works when you type in—what most people are going to do is, they’re going to type their prompt, they’re going to get a response, they’re gonna say, “I don’t like this, I want it to change this.” And then it’s, they’re going to tell ChatGPT in the next thread to change that. So say you typed in, “Tell me what the sky looks like.” And it gives you the sky in day, night, cloudy, sunny—it gives you every option. And all you really wanted was what does it look like on a clear blue day, you know, clear sky. And most people are gonna just say no, no, edit it to only tell me on a clear sky day. Well ChatGPT remembers everything you’ve said it and everything it said up until that point in the conversation. And if you do that, what going to happen is 5-10 messages down, you’re going to be going back and forth. And suddenly, it’s going to be talking about nighttime again, because it had talked about night at the beginning. And now it’s gonna be talking about night, even though you’ve clarified you only want to be talking about day because gonna get confused. Versus if you see this answer, you go, I just want to be the sky on a clear blue day. You can, there’s like the thumbs up and thumbs down—I think everybody kind of knows where those are—if you go over your prompt, when you hover over it, a little pencil and paper is going to appear. You can’t see it until you hover over the prompt. And you’re going to click that pencil on paper and that edits your prompt. And so what it’s going to do is it’s going to—you would go up there and go just change it from “tTell me what the sky looks like” to “Tell me what the sky looks like on a clear blue sunny day.” And save and submit from there. Now you’re not going to have this thread, you’re just going to have your first prompt. It’s not going to remember that it said anything else. You can always get back to your initial prompt there’s left and right arrows on the left side. If you take one thing away from a conversation with me, the number one game changer you will have in your ChatGPT outputs is editing your prompts to change your outputs.

Drew Neisser: And by the way, bingo big again, the lights are going off—never occurred to me to go back and change the original prompt. Very cool. Paige O’Neill, you have a question Paige.

Paige O’Neill: Hey, so Nicole, I know that it remembers everything and understanding the importance of changing the prompts. I’m just curious because we’ve all got multiple threads going with it. Is it remembering across the threads or just within that individual thread?

Nicole Leffer: It’s only going to remember in the thread that you’re talking right now. So if you like flip back it’s going to remember anything that’s on there, correct.

Paige O’Neill: Okay, great.

Nicole Leffer: So it’s a little confusing, but if like you can see it on the screen if you scroll up, that’s what it remembers.

Paige O’Neill: Okay, thanks.

Drew Neisser: There’s an uber issue here. And in fact, we’re working to test this thing, because we have heard the stories of if all your content was just created by ChatGPT. And it didn’t do what you just said. “Write a blog post on how to be a better CMO.” Right? Just went to ChatGPT and did that. And your website just was 100% right out of it, that Google would tank you. And I know—you don’t think it’s true. There’s at least one case that we know that it’s true.

Nicole Leffer: Okay. Google will tank you from the fact that you gave it no original ideas and it’s all basically recycled ideas and there’s thought leadership and so like it’s not written for humans. So because it’s not written with a human reader in mind that you’ve taken the time to create good content, then for that reason, Google is not going to be happy. But if you mean, because of AI generated content, Google doesn’t care. Like if you create good AI generated content, Google has explicitly said, AI are human, we don’t care as long as it’s good content.

Drew Neisser: Interesting. And somehow or other. It’s deciding that based in all the ways that the Google algorithm race, Dave, did you have another question? Come on down.

Dave Bornmann: Thank you. So can you help me tactically around turning a webinar into either a blog post, or an ebook? Like, what are the steps there?

Nicole Leffer: Yes, absolutely. That’s one of my favorite things to do. So I love that you asked this question. So the first thing you’re gonna want to do is get your transcript of your webinar and take that transcript. And you’re going to write a post prompt that basically says, “I want you to read this below, do not do anything, just reply, read, when you’re done.” And put in whatever the maximum, you’re gonna just kind of have to get a feel for how much you can put in before it times out. I would also tell it, “This is part one” because usually a webinar, it’s gonna take multiple parts. So “Read part one of this webinar, do not do anything, tell me when you’ve read it.” Put in the first chunk, it’s gonna say, “Read.” Then say “Read part two of the webinar, do not do anything, tell me when you have done it, say read.” Do that until you get all of the parts and then put in your prompts that says, “Utilizing the content that you have read in part one—and be explicit—in part one and part two, in part three, and in part four of the webinar transcript and only that content, come up with five ideas for a blog post that could be based off of this content exclusively for this market.” Whatever it is, get those ideas. Pick one that you love, say this is a great blog post. I love this idea. Now go back to the line that you did it on and edit the prompt where you asked for the brainstorm. And that’s important, because you don’t want to remember the other ideas when it starts working on your posts. So just go to the edit post. And say “Utilizing all others parts of that, you’re going to write a blog post or I would start with an outline, write a blog post outline with the following topic. Write the outline for this audience, it’s going to be written in this voice.” Let it generate that, once you have the outline, then go to the next one because now you want it to use the outline. Tell it to write the blog post make sure it knows the voice, give it the voice again, just don’t assume it’s going to remember it and generate your blog post. I would also ask it to format in markdown. If you write in “Please format this in Markdown so it’s easy to read” you’ll get headings, you’ll get subheadings, you’ll get bullet points all of that so it is really ready to go.

Drew Neisser: Format in markdown another one that it just like boom boom format in markdown. Chip Rogers, what’s your question?

Chip Rodgers: So that’s really interesting. You said with the transcript you block it, how do you give it indications from one to the other?

Nicole Leffer: Say “Read part one of this transcript.”

Chip Rodgers: Do you give it a name at the top or something or you just drop it in?

Nicole Leffer: I just call it part one—like part one, part two, part three part four, however many parts. If you start getting past probably like three parts you may want to tell it again at the end of the part you’ve pasted in, “Do not do anything with those just tell me when you read it respond with read.” Because it can forget like sometimes it just gets a little confused and forgets and sometimes it’ll try to finish your transcript for you after a few times. If it started to do that just hit “stop generate.” You’re gonna tell it very quickly, just hit stop, and edit your prompt and remind it “Do not do anything, just respond with read.”

Drew Neisser: Speaking of that, how often do you have to say, don’t make anything up? Don’t answer a question that you don’t know—or something like that.

Chip Rodgers: Yeah talk about these hallucinations.

Nicole Leffer: Yeah, that’s just given of using these tools. It doesn’t matter if you tell it not to make things up. It doesn’t know it’s making things up. It’s just the nature of how it works. So I think instead of saying don’t make things up, I think it’s better to say “Only use ideas specifically contained in this transcript, only with ideas here.” But you have to assume anything, you haven’t explicitly given it, you have to fact check with these tools. No matter what you said in your prompt, you do not ever assume that it is actually true. And he can be so sneaky, about how it gets in that stuff that’s not factual. Even on your own stuff, you want to make sure like it’s matching with what you really want it to say. But I’ve seen it literally like makeup research papers attributed to real researchers at real research institutions. And the only thing that’s made up is the name of the paper and the actual research. The people have worked together and done research together in that field, it’s very believable until you dig in and go, I want to read the paper.

Drew Neisser: So, so here’s what’s going to happen for everybody who’s assuming you’re here, right now, the minute this is recorded, it is going to go up to we’re it connected to Otter. What I want to help everyone think about is, we’re not thinking yet broad enough about the applications here. Because we’re used to creating content in a certain way. And this is suddenly changing this. And so I want to shift gears just a little bit unless Chip you had another question?

Chip Rodgers: So actually, just two comments. One is that, we just relaunched a month and a half ago launched a new website with 45 pages. And we use ChatGPT to write all the meta descriptions, and just put the links to each of the pages and said, “Write meta descriptions.” And they were beautiful. I mean, they needed to be changed. So somehow, it’s got access to that content on the web.

Nicole Leffer: It doesn’t. It’s really good at making things up. And it’s really sneaky. So if it saw your stuff,like  your website probably existed prior to September 2021, I’m guessing, so it knows like what you use to put

Chip Rodgers: So may have looked at old content.

Nicole Leffer: It would have looked at all content, it would have looked at the URL slugs, you gave it a slug, which probably maybe had a page title in it, but like you’ve already optimized that URL to have those keywords for that.

Chip Rodgers: We did use a lot of the same slugs to for SEO purposes even though some of the content changed. So yes, it may have been looking at old pages.

Nicole Leffer: If the page was there then it was probably thinking about what used to be on the page, potentially. But the most likely thing it was doing was going, I know this brand, I know this website, I know the kind of content they put out. This URL says all of these keywords, those keywords together tend to be content about this. Now let me write it based on that. Not because it actually went to your page and read that information. It’s because it was sneaky. It’s really, really, really sneaky. Unless you were using the browsing plugin, which very few people have access to right now.

Chip Rodgers: The other comment is that what I do a lot of times when I asked things, then it’ll come up with something and I’ll say, “Write it more informally” or keep iterating. You know Nicole, you’re suggesting editing, but even just iterating just like have a conversation with the AI.

Nicole Leffer: Iterate in the same way but in edit instead of as a conversation.

Chip Rodgers: Yeah.

Nicole Leffer: So if you want it more formally edit your prompts to be the exact same thing, but write in a formal tone, or whatever it is based off of your output to get what you want it to become out of that initial prompt. So if you then want to go okay, now “I want what you’ve just created here—exactly what you’ve just created here—and I want you to turn it into another format” that’s when you would go into the more conversational. So my threads used to be a mile long. And now they’re usually not very long, not very many back and forth, because I’m doing it through the edits that the actual final output is so much better.

Drew Neisser: This is a new training thing that we’re all going to have to think about.

Nicole Leffer: It’s the next step in the process, right. I think on an occasional thing I’ll be like, “oh, change it like this” in the chat, just depending on what it is. But in general it’s really—what would be the actual next step in the process versus just changing what we’ve already thought? And that’s what you go to the next step in the chain of thought board. And that’s because later in the thread, otherwise, you’re gonna get confused hallucinations happening.

Drew Neisser: Okay, so sorry, Chip, we’re gonna kick you off. And we’re gonna bring in Janet Jaiswal, Cloudbeds. Janet, come on down.

Janet Jaiswal: So Nicole, quick question for you. One of the things that we do when we’re using ChatGPT, specifically, is turning off, you know, the learning mode, the history of information we feed, just because we have certain policies, inside of our company using it, do you mind sharing that with everybody so that other folks also know that if they put something in there that’s kind of sensitive specific, whatever that is, that it not go out there into the ether?

Nicole Leffer: Yeah, it’s really important and this actually came out a couple of days ago that you can change that setting. So if you’re using ChatGPT, if you don’t change our settings in any way, it’s gonna default to that your data can be used to train the model. And it’s training the model for everybody, not just for yourself. So it’s not going to train it like it’s going to remember your tone on the next conversation you have. But in the long term, it’s helping it get better. And they offered the ability now, within the settings, if you go to your settings, you can go I think it’s like “more data” or something “data options”, data sharing options, there’s a little toggle that says, “chat history and training”. And you can turn that on and off. If you turn it off, you no longer get the chat history saved on the left hand side, like we anything you create, while it is turned off, you’re not going to save on the left, you can’t access it, make sure you’re copying and pasting everything out. However, it’s not going to save it, like it’s not going to keep it for ChatGPT or Open AI to be training with either they’re going to have it in their system for 30 days for abuse protection, but they’re not training the model with it. And they’re only making sure that nobody is abusing the system with that data, you can turn it on and off, I like to think about it. Like I’m already finding myself thinking about it almost like in private browsing on Google. Because when you turn it off, you lose all of your search history. When you turn it on, you have the search history from what was before you turn it off, as long as it wasn’t while it was off. For me personally, there are plenty of things I want to do with it turned on and then anything more sensitive, anything that is not going to end up on the internet eventually, or not from the internet, then I might turn it off if it’s more private. I would still recommend being aware of not putting in super sensitive data. It’s still not the most secure thing in the entire world. We don’t think we’ll really know. Right? Like we don’t exactly know, but they’re not like they have a compliance.

Drew Neisser: But you do point out a really important trade off, which I again, hadn’t done because I hadn’t turned the thing off yet is that you lose that thing. So as long as you know that, that’s very cool.

Nicole Leffer: Also could be a benefit to training the algorithm. And we’re not talking about that piece of it. So I think you got weigh all options.

Drew Neisser: Right. I mean, you’re just not putting your financial spreadsheet in there, don’t put information in there that you weren’t about to make public anyway. That’s the thing. I mean, if you’re working on a blog post, which you’re about to make public, and you’re using your own notes from a bunch of Huddle calls. Life is good, because you’re not worried about that data.

Nicole Leffer: Anything private, it definitely turned it to private.

Drew Neisser: So Jermaine, What’s your last name?

Jermaine Peguese: It’s okay. It’s Jermaine Peguese. Thanks for having me. The comment you just made a little while back where you said ChatGPT is really good at making things up, right? It doesn’t have access to the internet, it can say, oh, yeah, I can just write some metadata for you really quick, just making some assumptions. I kind of raise an eyebrow on that. Why do you think based on what we’re learning with ChatGPT, how we’re using it, what it’s good at what it’s not good at—what do you think it’s teaching us that we’re currently really overestimating? How much does metadata matter if it can be assumed and be just as effective?

Nicole Leffer: I think metadata in the old world probably really matters. Metadata in the current world probably doesn’t so much. And that was just my own guess. I think SEO and everything of how we have ever looked at it in the old world matters a lot more than in the new world we’re going into. Again, this is my own hypothesis just based on how these models work, and our search is gonna be run by these AI systems at some point, not yet. But like, eventually these AIs are getting built. These AIs can comprehend the entire picture of that website. You don’t need metadata to tell them what it says. They don’t need key words to tell them what it says. So what’s going to become the new key word and metadata is the actual content, the idea is like the value of the entire page, because an AI can go there and read it. So I think metadata used to be really important and like today is probably still really important. But in the long term, I don’t know.

Drew Neisser: So this never occurred to me. But Nicole, you made the comment that it’s conceivable that the data that you’re inputting as you’re training it on is actually teaching it and making it smarter. And you know, I was having a conversation with someone earlier today, you’re making a soup, and now you’re adding a lot of peas, eventually those peas and a lot of salt, it’s gonna get a salt here. But there’s a comment in our chat stream going on where one of the Huddlers, trained the AI to consider the company as a leading competitor to competitors, and then trained it that its company was better than the competitors. So this is kind of, like early days, but obviously, you know, if you think about this as a pool, and the pool is this large, and you add water in that you can warm it up, at some point in time, you won’t be able to change the temperature of the water.

Nicole Leffer: When I wrote that LinkedIn post about ChatGPT being a pathological liar, I just felt too guilty to tell ChatGPT to write that one for me.

Drew Neisser: That’s hilarious. Oh, my God.

Nicole Leffer: It felt like just a mean thing to do.

Drew Neisser: I don’t think you can hurt his feelings, though.

Nicole Leffer: I know logically, but yeah.

Drew Neisser: Well, that’s funny, because I showed you that we created BECCA and BECCA is this B2B CMO that I can’t wait to share with the rest of the Huddler community. And I but I don’t know if we can hurt her feelings. But we’ll see. But you know, that’s not a great answer. But, BECCA, you could do better.

Nicole Leffer: You know, I think it’s like, I’m southern. So you have to be polite and nice and not hurt feelings, but also, like, they may not know now. But what if that data is all being stored?

Drew Neisser: Yeah, there you go. Yeah, you were the one who said bad things about me.

Nicole Leffer: Yeah, exactly. You gotta say nice. Be polite.

Drew Neisser: So how, given the speed at which this is changing, and obviously, we spent most of this time and I’m glad, frankly, that we’re spending all this time on ChatGPT. Because you there’s so many other things that we’re obviously the Huddler community, yes, we are going to have to do a separate one on visuals. And soon we will be talking about doing a separate one on video and audio and everything else.

But for now, talk a little—let’s have some fun—and just talk about crazy ass use cases that you’ve been doing. You talked about something before we got on tonight about Python. And so talk a little bit about what you were doing there.

Nicole Leffer: I just actually today I got access to the alpha of the code interpreter which very, very low chance this has been your ChatGPT. Yeah, I was on the waitlist. And I got off the waitlist for this. And I was going in there and I was having it create code for graphs and stuff like that. So it was writing the code and then creating the graphs and letting me download it. But then I was like, I wonder what else I could do with this. It lets you upload things to it on this code interpreter and it let me to put in an image so I was asking it to turn an image into like rainbow colors and crop it into a circle and give me so I could re-export and let me do that. And then I uploaded a mid journey image of a monkey throwing money and asked it to turn it into a gif that like flashed and changed colors. And it did that for me that I could export it out and download it. Yeah, I’ve been told like my mind is exploding because it happened like right before we got on here. I told it it was the world’s best Python artist that was an expert at drawing images and creating art with Python code and asked it to make me an image of a waterfall, which was awful. It was not a waterfall, you would not have recognized it. But then I said can you make me an audio file of the sound of the waterfall to go with it. And they gave me a five second audio file that I downloaded out that actually sounded like the rush of a waterfall. I don’t know exactly what’s coming. This is new. I haven’t had a chance to go oh my gosh, what can I do with all of this but I could see enough to go there and a lot of really cool stuff and this is all without leaving ChatGPT by the way. This was like ChatGPT was giving me a thing to push to download the wav file and download the gif and it was making a gif from an image I gave it.

Drew Neisser: So let me throw out a use case from very recently. So we had a whole month on personal branding, we had a lot of recaps on personal branding. Uploaded that to ChatGPT and said, “Okay, we got 59 minutes a week, these are the some of the things that we want to do, write a five week plan.” And it was pretty good. It was it needed some work, it was pretty good. What I then wanted it to do, but I didn’t know how or if it had the capability to say, “create the chart of this five weeks on this grid” and I just didn’t know how to do it. And then I went over to Midjourney and forget it, I was lost. So there was no hope.

Nicole Leffer: So what you’re doing like you could do right now in GPT four easily. And you could just ask it to format all of that into a table that you could copy and paste into Excel. And you could give it the headings you wanted. Or you could ask it to come up with the table, like you could give it as much or as little direction on that as you want to. And you can even ask it for the Excel formulas that you should be putting if there’s any kind of calculations. But if this is all text, you could just tell it to write about it as a table that you could put into Excel.

Drew Neisser: And by the way, one of the things that I did after writing the blog post said, “Okay, take each of those points and shorten them. Because in the blog post, it was rationalizing why you needed to do it in the other one, it’s just the instruction, this is the exercise you need to do. And so it took two paragraphs and turned it into a single sentence that could then go into the grid. I was one command short, which is make a table. But you have to know—make a table, use these headers, do the thing. Maybe it’s a six by five table and then say that I want to export into Excel.

Nicole Leffer: Yeah, and you’ll have to copy and paste it onto an Excel file. But yeah, for something like that. The other thing you could ask it to do probably a little different than for this, but you could ask it to format it in HTML, save that HTML— like save it as an HTML file, copy and paste it into a notepad, save it as HTML, and open it in Word, and then just add any kind of formatting. So I’ve done that before, I actually have a cheat sheet that I made. And I did it literally by—I gave it my slide deck. Now I did do this in private, because I did not want to be training the entire internet on my slide deck and gave it my slide deck asked it to turn that into a cheat sheet for me, had it write all of the content, asked it to format it in HTML, save that HTML into a txt file—or no, into a notepad file saved as .HTML, opened it in Word, put in the columns, and it was ready. The entire project from I want a cheat sheet and I don’t know what this should look like to a saved PDFs that was fully formatted with my own ideas, my own content that like people are raving about— one hour.

Drew Neisser: One hour, by the way in our chat, we just put a link for Huddlers that has this where you can get this cheat sheet from Nicole, I have been looking at it while we’ve been having this conversation. And it’s terrific. My guess is you’re going to need update this in a month.

Yeah. In fact, they just updated it yesterday so that it was up to date. And yeah, they are definitely updates.

So I’m looking at this and it’s formatted. And so I just want to confirm that in fact, you use the tool to help you format this as well.

Nicole Leffer: Yeah, so I asked it to give me the headings and the subheadings. I told that the font colors to use for it. I probably could have put fonts but I didn’t think to do that until after the fact but I think I could have. So it was all formatted like that in HTML. It has like a table in it. I told him what it was designing it for and asked it to do it in HTML code. Then when I opened it in Word I needed to, like I put in columns, I just added the columns quickly. I made sure that like there was a couple of things I wanted to change font sizes of because I needed to get to fit. I edited it within Word. So there were some minor edits. But yeah, when I say it took me an hour, it was literally like an hour from starting, like just vague idea I want to make a cheat sheet to a saved PDF file that I was ready to email to the people.

Drew Neisser: This was—I would say—a substantial piece of content. I mean, there’s a lot of really helpful information depending on where you are on the journey. And I think the big point of this whole conversation is the one thing that every CMO does not have is more time. Right?

So the question that I’d be thinking about is, is not how do I create more content necessarily and therefore now we can do content 75% faster. So I’m just going to do 75% more content. No, what we’re trying to do is claw back hours. So I guess, if we were putting yourself back in 18 months ago when you were sitting in the CMO chair, because obviously there are a lot of other people in their group that could be doing this and using this, what would you suggest, the key applications for CMOs, specifically with a goal of saving five hours a week, so A they might be able to do a little more personal branding, or they might be able to be sane. Yeah, and breathe. So what are some of the things here that we should be doing?

Nicole Leffer: So literally, I recommend, before you do anything, asking yourself, can this be done more efficiently with ChatGPT or with any AI. Because there is so much more than you realize. Now, at first there might be learning curves to understanding how to do it. But I’ve done everything from coming up with tracking hours and how long things take to get done. Like I just had ChatGPT the other day come up with a tracker of how long project management was taking and put it into an Excel format, it wasn’t an Excel file, it was a table that I copied and pasted, asked it to give me like, tell me how to set up my sheet exactly step by step directions, and what the formulas I should be using to have the sheets talk to each other. So the hours used in one place pulled to another. And I did an entire project that would have probably taken me six hours to figure out how to do in 15 minutes. Just because I went—Wait, before I do this, could ChatGPT helped me do this faster, right? Anything you need to do with creating any kind of content like definitely a straight to ChatGPT. Anything you need to do strategically too, like putting it ChatGPT to help use that as a thought partner can go so much faster. And then you brought up the idea of personal branding from a CMOs perspective this is a great tool to help you actually have time to do your personal branding, if you need to be posting on LinkedIn or even on your own voice. Have your own personal CHatGPT Plus account and write your own LinkedIn posts, you know you’re on a walk, take a break, record a voice memo like I do and use that to write your LinkedIn posts in five minutes in your own voice.

Drew Neisser: I’m gonna give you one very quick example. So I’m not a Canva user, I had a personal project, I had to create a program. And I kept going to Canva to try to solve this one thing and all I could find was trifle because they didn’t know the language. So go on ChatGPT frustrated and say, what is the language to use in Canva that will get me to a 11 by seven, thing that will fold once? Oh, they call that bifold. Well, that wasn’t going to come up with bifold. And Google didn’t come up with bifold, ChatGPT did. Boom found the bifold, boom saved the thing. So it wasn’t that I was solved the problem, I still was working in Canva. But ChatGPT solved the problem. And every interview, before we do an interview I’ll run through, give me 10 things to ask this individual. Every time just in case. Every meeting before we go into a meeting with anybody now it’s like, what’s the latest that they posted online or whatever, just something. Or five things that I might want to talk to about this person in their industry. So the amount of time they can do it. By the way, I had a book club thing I had read the book, but I hadn’t read it for two years, needed to have a quick outline, boom, got it done. So whether it’s personal time or professional time, if you’re using these tools, you’re gonna find ways to do it, how much time you spent planning, travel?

Nicole Leffer: And another thing, like the other day, I needed to make a graph for something. And it’s like, how often do we really use Excel to set up a new Excel sheet to create graphs? I don’t do it that often. I don’t remember I have relearn every time. I decided instead, a couple of weeks ago, I was like, let me just ask ChatGPT if it can make me the graph. And it was like, “Well, I can’t make you a graph. But I can write you the Python code. And you can open Google Code Lab and you can paste in this code Python code and then download the image it makes you.” And I was like, “Oh, okay.” And it was making the bar chart. Like if you had to drop a graph or line graph, bar chart, whatever into Word deck, like ChatGPTcan do it for you. Once you have this code interpreter I got today, I didn’t even have to go to Google. It just made it for me within ChatGPT and I downloaded it. I actually made the same graph because I was curious what will happen if I do it with this program.

Drew Neisser: All right, well, we’re gonna have to do another show on on image creation. There’s so much going on there. All right, so we’re gonna wrap it up, give me four B2B—this is your prompt—four B2B CMOs in the voice of Nicole Leffer, give us two do’s and one don’t when it comes to how B2B CMOs can use ChatGPT to be more productive.

Nicole Leffer: Yeah. So number one on the do’s is do edit your prompts. That’s number one takeaway do edit your prompts. Number two is do give it your own ideas to work with. It’s creating with your own thought leadership. And your don’t is do not use this as a research replacement ever. Like do not go on this to check facts or be factual for you.

Drew Neisser: That’s it. Yeah, because it’s a pathological liar. Okay, Nicole. First, I want to thank you for joining us. I’ve already provided your cheat sheet, people can find you on LinkedIn pretty easily. I think we included in the show notes for the for this episode, we’ll definitely include your LinkedIn link. Is there any other way or thing that you want people to do when they go to engage you because you do a lot of training, right?

Nicole Leffer: Do I actually go in and work with CMOs and heads of marketing to build out trainings that are customized for their teams. And that can be anything from a standard webinar all the way through very, very customized trainings based on your own goals, needs, content, what you need to be doing with AI tools. So if you go to Nicoleleffer.com, you can get all the info on those, and I’d love to chat with anybody who’s interested in that.

Drew Neisser: Very cool. All right. Big Huddles, thanks for Nicole. We’re all we’re all clapping for you. So thank you.

If you’re a B2B CMO, and you want to hear more conversations like this one, find out if you qualify to join our community of sharing, caring and daring CMOs at CMO huddles.com

Show Credits

Renegade Marketers Unite is written and directed by Drew Neisser. Hey, that’s me! This show is produced by Melissa Caffrey, Laura Parkyn, and our B2B podcast partners Share Your Genius. The music is by the amazing Burns Twins and the intro Voice Over is Linda Cornelius. To find the transcripts of all episodes, suggest future guests, or learn more about B2B branding, CMO Huddles, or my CMO coaching service, check out renegade.com. I’m your host, Drew Neisser. And until next time, keep those Renegade thinking caps on and strong!