Supercharging Marketing Productivity with AI: A Deep Dive into ChatGPT
Show Notes
Episode 2: A discussion about the application of AI, specifically OpenAI's ChatGPT, in the field of marketing.
The hosts discuss how AI can significantly accelerate productivity by assisting in tasks such as content creation, ideation, and data analysis. They also discuss the potential concerns related to AI, including plagiarism and the accuracy of AI-generated content. The hosts highlight the importance of treating AI like a marketing intern, providing it with detailed instructions to improve the results. They also mention the use of third-party plugins, such as "Ask Your PDF,” to enhance the capabilities of AI.
Key Takeaways:
- AI, specifically ChatGPT, can be a powerful tool in marketing, helping to accelerate productivity.
- AI can assist in tasks such as content creation, ideation, and data analysis.
- It's important to treat AI like a marketing intern, providing it with detailed instructions to improve the results.
- There are potential concerns related to AI, including plagiarism and the accuracy of AI-generated content.
- Third-party plugins, such as "Ask Your PDF,” can be used to enhance the capabilities of AI.
Key Moments in the Podcast:
Discussion on the application of AI in marketing
Discussion on the potential concerns related to AI
Discussion on treating AI like a marketing intern
Discussion on the use of third-party plugins to enhance the capabilities of AI
Resources and Links Mentioned:
https://openai.com/
https://askyourpdf.com/
https://www.actualtechmedia.com/
Transcript
Intro (00:09):
Welcome to Tech Marketer Live, helping you create and capture demand in the enterprise technology market. In this episode, we'll uncover how marketers can get 10 times the productivity and reduce their workloads with the help of ai. Now, here's your host, Geordie Carswell.
Geordie (00:34):
All right. Hey everybody. Geordie here. I'm excited for this one. This is going to be awesome. We've got James Green here, who's my co-founder at ActualTech Media. Our resident nerd James has some amazing ability to, namely the ability to understand marketing as well as the technical side, having been an IT architect in the past, and he's going to help us walk through what I think is one of the more fun topics that we're going to discuss for a long, long time. How you can use AI to supercharge your marketing capabilities, both for your organization, for yourself personally.
(01:19):
I was blown away when we started putting this together, what it's going to mean for everybody and how just some of those things that come along in life once in a while that are just an iPhone moment kind of thing. So this is going to be really cool. This is probably just part one. There's so much to cover. We're going to talk through today, why you would want to do this, and then what you can do and examples of what you can do and get ready to have your mind blown. This is going to be fantastic. So welcome James. Happy to have you here.
James (01:54):
Thanks for having me. I'm really excited to talk about this and we don't have nearly enough time to cover all the things we could talk about. I think you, it's overused to call something big and revolutionary, an iPhone moment, but I mean, this is probably bigger. It's going to be insane. Can't even we, I've been sitting here staring at it for a couple days thinking about having this conversation, and I can't wrap my head around how far reaching the ramifications of the AI wave is going to be. It's crazy, but if you're a marketer, we're going to talk about some things today that you can use immediately to do amazing things really fast.
Geordie (02:36):
All right, let's just jump right in. So let's set the table here. Can you set the table here a little bit for us, James?
James (02:43):
Yeah, for sure. So if you're a marketer, you're staying up to speed on trends and AI obviously is one of them. I'm sure that probably everybody listening is aware that this is a big deal and probably most of you have checked it out, but if it's not something you're looking at all the time, I think there's probably three or four categories of people that could be listening. One is like, this is a big deal, but you've never checked it out because you are a really good experienced marketer, not necessarily a tech person. And the idea of playing with artificial intelligence just sounds intimidating to you, but I want you to know is there are things that you can use right away that are very approachable, and we're going to talk those through and you could hop in and start using it when you're done listening to this show,
Geordie (03:28):
Even if you're busy.
James (03:29):
Yeah, I mean, it's not only is it easy, but it's fast. Awesome. Another category of people is you checked it out before, but it was a little clunky, hard to use, didn't produce the results you wanted, and you haven't checked it out for a while. And what I need you to know is things are changing super fast. It's even more approachable than it was before. And more importantly, we're going to talk a lot today about ChatGPT, and forgive me in advance if I conflate AI and ChatGPT in this conversation. Most of the time what we're talking about is ChatGPT. Sure. ChatGPT now has in beta the ability to search the internet. So prior to let's say a month ago, it was capable of responding based on the dataset that it was trained on, which ended in late 2021. So you could do anything that the internet knew about really prior to 2021, but anything real time was not possible. Now it's possible. Additionally, OpenAI is rolling out support for plugins, like third party plugins that ChatGPT can interface directly with. And so the reach and knowledge that ChatGPT has now is not what it was if you check this out six months ago. That's my point.
Geordie (04:48):
So they let it out of its cage
James (04:51):
And this is really going to make what you can do with it 10 x what it was before. And then the other category of people that might be listening like this better than me. And if that's the case, let's talk because I'm enjoying learning about this.
Geordie (05:06):
Right on. So why do we want care? What do we want to do here?
James (05:11):
Okay, so for me, when I think about if I'm in a marketing role and my job is to come up with lots of good marketing ideas or actually create lots of really good marketing content, the biggest struggle is my capacity for creativity because I only have so much just willpower in a day. My brain has so much energy for being creative, and once it's spent, it's spent
Geordie (05:36):
And I have five cells. If the brain testing, yes, that's it,
James (05:40):
Three to five, depending on how good of a day,
(05:43):
What's awesome about ChatGPT or other tools that you could use specifically for marketing tasks is that it doesn't run out of energy. It just keeps going. And so if you take an example like brainstorming, Hey, here's a bit of background information. I'm looking to create 10 blog posts on this topic. And then you ask, you get 10 ideas, and if you don't like those, you ask for 10 more. And if you don't like those, you ask for 10 more. By the time I came up with 30 halfway decent ideas, I'd be spent, but that took ChatGPT 30 seconds and the 30th idea is just as good as the first. So the reason you care is, or for the reason I care, I don't want to speak for everybody else. The reason I care is it can augment my creativity and my capacity to be creative, but also it's just way faster than me.
(06:33):
So if I want to write a rough draft of a blog post, I might block out an hour to do that. Well, ChatGPT could literally do that for me in 20 to 30 seconds. So especially if you're pretty early stage and your goal is just to get a bunch of stuff going and then refine it later. Yep. I think I said to you yesterday, Geordie, you could scaffold a microsite in an afternoon. And by scaffold mean, I mean you could populate it and have it full of pretty decent content with a really good foundation idea, bit of background info and an afternoon to hang out with ChatGPT.
Geordie (07:09):
Yeah, and I mean, if you're a startup, you could build out your resources section in time. Yeah.
James (07:14):
And then one last thing that is really important to me is the human brain is a pattern recognition engine. We're really good at that. However, one thing that we're not great at is parsing huge chunks of data. So if you were to take, say 200 survey results and try and create meaning out of that, it's doable, but we're not that great at that. Computers are really good at that because again, with the energy thing, it could look at all 200 with the same lens as the first one and the last one, I can't do that. So to be able to look at a big volume of data, whatever that is, and then make some meaning out of it, is something that I'm seeing myself using AI for a lot.
Geordie (07:59):
That's awesome. So what's it not for
James (08:05):
I, I'm hesitating because I don't know. I don't. I think it's could be for just about everything. I think there are some concerns that people still have that are valid and we're needing to figure out how to work through those things. One is, and this applies to text-based stuff or coding or art image generation applies to all of it. There is concern about plagiarism because the ai, whichever one, it is informed by stuff that already exists and trained on stuff that already exists. And so the risk is somewhat high that either literally legally you are actually plagiarizing or at the very least you're kind of like morally plagiarizing. And so that's a concern for some people, and that's for each company and each individual to sort through as far as what level of risk you're willing to tolerate and where your boundaries are. I think you got to be especially careful if, let's say you're in an industry that's highly regulated and compliance matters, you got to be careful what you let AI write for you, right? So that's one. Another concern that as a community we're sorting through right now, is AI generated content going to be penalized by search engines, for example?
(09:32):
To some extent maybe that's been hinted at, but at the same time, a lot of what we're going to talk about is really using AI to bootstrap your work. It's not just having AI do all the work. And I think there is definitely a line there where when you're just using it to make you faster and better and more creative, that's not necessarily going to be a problem.
Geordie (09:55):
Awesome. Well, I think, well, so
James (09:56):
There's, lemme mention one more thing. People experience with G P T especially know that it has a propensity to hallucinate, and that means make up things that are not true and state them in a very factual way
Geordie (10:13):
Like a human.
James (10:14):
Yeah. I also hallucinate from time to time. When I was looking in the mirror this morning, I was like, you're so handsome hallucination, but if you're writing content that is meant to be authoritative, you got to be really careful that you know are expert enough to fact check AI because it will say things that seem really believable and true that are totally patently false. So another thing to be cautious of,
Geordie (10:41):
And I think that's a good point. So these are all the disclaimers and run stuff that you're doing that you don't know if it's accurate or not accurate past your SMEs. So they can at least not be staring at a blank piece of paper being asked to write something, but you can give them a starting point and then they can work from there. So I think this is where I come in with the way I look at it. AI is an accelerant, it's not a replacement. So basically it's like putting your capabilities on steroids. You can 10 x your productivity as a marketer and it can help you ideate, which is 90 when we're tired and overworked and everything else. Half the department's been let go in workforce reduction, and now I've got their job and their job and their job to look after. Now I can scale myself and I think my team much better than maybe I did, was able to before.
(11:41):
I can also, so I've got help with ideation. I've also got help finding holes in my work that where things need to be short up or improved. And I think for me, using g PT day to day to help get through the stuff that bogs me down, it's not, it's important, but it's not critical, but it can tie up a lot of my time and keep me through more important things. So basically it's not just about asking it questions and then spitting back the answers is about getting it to think like you do. And if you can get it to think like you do, you'll see the results improve and you'll know, excuse me, that it's on the right track and it's helping you. So a big part of that is the prompts, we'll get into all of that. A marketer, I believe that a marketer who can master the prompts and how to get the right output from these tools will have a massive leg up in terms of productivity and the ability of their team to scale in 2023 and beyond.
(12:49):
So just as an example, I just did an experiment. I asked Chachi PT to do this for me in about 30 minutes. Here's what it cranked off, just to give you a sense of scale. Create a content calendar for a month, create a bunch of article outlines, write the articles with SEO friendly optimization, turn the articles into a series of Twitter threads with emojis and hashtags. Create a 16 email marketing sequence, a drip outline, create outlines for each email including graphics, titles, call to action, write the actual emails in sequence in the proper tone. Create a script for a 15 minute podcast on a very complex technical topic. Brainstorm ideas for incorporating AI into disaster recovery planning, a detailed description of a business idea or something like that to play around with and to let me know what the steps would be to launch it, that now there's trial and error, there's tweaking, but that's in 30 minutes.
James (13:49):
L, look at that for just a minute and take a guess at how long it would take Geordie to do that, all that stuff.
Geordie (13:55):
Would Geordie be in a fetal position on the ground?
James (13:59):
But I mean, I'm thinking probably two or three days
Geordie (14:04):
At least
James (14:05):
And at least, and it did all of that in about 30 minutes.
Geordie (14:09):
Yeah.
James (14:10):
And what is your estimate of how ready to use those things were?
Geordie (14:17):
It would need, let's say another 30 to 30 minutes to an hour to get all those things tweaked and ready for. Yeah, so
James (14:25):
I mean realistically you're talking about two or three days worth of work with your sort of augmented robo marketer, Geordie. You could get all that done before lunch on day one.
Geordie (14:36):
Yeah, for sure. And a big way to get these tools to do what you want them to do so that you spend less time editing and more time deploying what you're getting back or working with the ideas that it's giving you. And somebody had this comment a while back, the results from ChatGPT and similar tools improve significantly when you treat AI like a marketing intern. So the same in instructions, the same level of detail that you would give to an intern in real life. You give to ChatGPT and you invest in its training just like you would with an intern. It'll take more time upfront, but once it learns it's go time, it'll pay off. If you hired an intern tomorrow, you wouldn't just give them one or two sentences of direction. You'd create checklists for them, explain why we do things a certain way, ask if they have questions, all of that kind of thing. It seems crazy, but that's the conversation that you can have with Chachi bt to help it figure out what it is you want. And once it's figured that out, like we say, it's go time.
(15:43):
So there's things you can do there to help it along in that path. One of them is to create constraints for it. So if you give it vague information, it'll give you vague results back. So basically nail down very concise questions for it just like you would for the broader instructions you give to an intern, the less likely you'll get back what you're looking for. So it's the same thing there. If you think about it in this framework, you do the thinking and tell AI to do the doing. So you can't expect AI to entirely think for you. Sure it can ideate and things like that. But basically you need to help it understand why you want what you want. So I want to make a podcast transcript that talks about X, Y, and Z. Tell it what you need, please make it 15 minutes, blah, blah, blah, blah, blah. And how you want it delivered? Deliver it with timestamps or whatever the need might be. So if you get very specific and people just ask it, why is it sky blue? That's what they tested it with. What they didn't realize was the power came from telling it exactly what to do based on exactly what you want.
James (16:53):
I'm having a thought right now that I hadn't thought about this before. What are the chances that people coming up in marketing who are a little bit more junior, they use AI quite a bit to get their work done and as a result they end up being better leaders because they're better at giving direction and setting expectations and tempering their own expectations. It might sort of as a byproduct, train people to be better managers and leaders
Geordie (17:18):
Could be. I mean in two you think when you ordered content from a content marketing agency, they gave you a brief, they said, tell us what you want to see, and you would give them back, here's the tone, here's the subheadings, here's what I'm trying to capture, the key points, et cetera, et cetera. Then you have this huge long word doc that was the brief. You might as well have just written the thing sometimes. But the point is give the brief to chat PT and have, of course it's more complicated than that, but give the brief to chatt and it create that for you. And it's the same model. It's using AI instead of a content marketing agency.
(17:58):
It'll get confused. You can ask it, are you confused? Tell it. If you didn't like what you saw, say do that again, chase, don't have it. Get too far off the beaten path, keep it tight, help it. Remember you're hand holding it through to get to a result that you want and then afterwards you can save your work and be able to save time next time as it learns. And we'll talk about how to do that. Okay, so let's jump into something cool. So one of the first things that we talk about chat GBT a lot, but one of the more powerful tools that marketers may not be aware of is called, so synthia is basically AI, video and audio with avatars that will speak and look like real humans. And you can feed the script. It will even AI generate the script for you. Now if you give it the high level points. So rather than you writing the script, it will and including all the inflection points and the voice and everything that will make it turn out well. So what I did is I took, and I think we'll get this teed up, Derek, I think I took the megacast that we have coming up around ransomware. It's a online summit type of thing. And we took the abstract or part of it and gave it to Synthia, turn into a little promo video. And here's the result.
Video (19:36):
During this megacast, you'll see and hear about both innovative and time-tested solutions to help stop the scourge of ransomware. Solutions will range from human-centric tools for training users or flagging potential phishing messages to technologies for rapidly recovering when a ransomware attack does get through it.
Geordie (19:54):
Unbelievable. Like the inflection. That's crazy
James (19:56):
Man.
Geordie (19:57):
The inflection is perfect. The voice is natural. You can choose any voice you want. They've got like a hundred and something voices, any accent, they've got dozens of avatars that you can use. You can even upload yourself.
James (20:12):
I was just going to ask, am I remembering that you can also train one to talk
Geordie (20:15):
You? Yeah. And the other, so you can train it on your own voice or how do you know that This isn't my avatar right now, but I'm going to
James (20:24):
Train one to talk Scott Lowe.
Geordie (20:26):
Yeah, that'd be good too. And I think we can also upload, if you shoot yourself against a green screen for a certain number of dollars, I think it's like, I dunno, a couple grand or something like that, you can upload yourself and you turn yourself into an avatar. So celebrities do this to send out personalized show notes for money to, it's unbelievable. But the cool thing is too, you can have it do it in other languages in a snap. So let's bring up the other one.
Video (21:02):
Ransomware.
Geordie (21:17):
So boom, there it is. Ready. Localized for other markets.
James (21:21):
You didn't touch anything else except said, now do it in Spanish. Well,
Geordie (21:24):
We took it and ran it through a translator and then fed the script. I see back to it. Oh, I see. Yeah. And then boom Spanish, avatar Spanish, wow. Accent language. Everything just helps you to start to picture what the options are in terms of how you scale your campaigns, providing if you're doing video content, things like that. And you crank it out. And by the way, we're going to send it out localized to all of our field marketing teams and latam or EA or what the potential is. Unbelievable. Anyway, that I just had to call out. I know we're talking a lot about chat g bt, but there's other tools out there too that just will blow your mind.
James (22:02):
That is a
Geordie (22:02):
Lot. 30 bucks a month, 30 bucks a month.
James (22:05):
I've overseen localization work. And to create those two videos might take in a more conventional process like a month in a really good case scenario
Geordie (22:16):
For sure. And I mean if you want to, the translation quality and all of that kind of thing matters and there's services that are better than Google translate for that kind of stuff. But the point is like you just load the script and go and the potential's amazing. Anyway, let's jump into a bunch of examples here about what chat j p t specifically can do to help accelerate your work as a marketer. James, I know you dug up a ton of examples here. Let's let's roll.
James (22:44):
Yeah. And I'm realizing now that I should have asked chat j p t to give me more because it probably would've, but next time. Yeah, I thought of a few really cool things and then just to make sure it was a good idea, I actually made it do them. So we're going to talk through a couple things that we've actually tried and went really well. So first one was we just recently have been doing some case study work interviewing customers, sort of parsing that content and creating case studies from it. And I thought, what if we could have ChatGPT, read the transcript from the interview or from multiple interviews and create a marketing persona of the person that we talked to. So I fed it a transcript from one of our favorite and best clients and it read over the interview that we had done, created me a persona with a catchy alliteration kind of name.
(23:37):
You do bulleted out a bunch of behaviors that this person likely has, challenges that they face, things they're interested in, probably better than I would've done on the first pass. It's really good. And so if I were to try to do this process again, that's like a couple hours for sure to read over the transcript, I mean is an hour long interview. So to read over it a couple times and try to wrap my head around the sentiment and the things that were important to her and then put this together easily would've taken me half a day. This was a five minute exercise with Chad and this
Geordie (24:18):
Included her behaviors from a business perspective for challenges or interests, everything.
James (24:27):
So we'll talk about this more a little bit later on, but the catch to getting this one was just being able to feed ChatGPT, enough information when I've got a transcript of an hour long interview that's more than you can put in a single prompt. And so I had to get creative with feeding it everything it needed to do this, but that one went great. So that's something you could play with. Is it anything that you can supply to ChatGPT about your client and then ask it, summarize this. Tell me in two sentences who my prospect is and what they care about.
Geordie (24:59):
Can you feed it like a word doc? How does this work?
James (25:03):
So not directly. Two, there's kind of two ways I can think of off the top of my head to handle that. One is something I'll talk about a little bit more later where you can chunk it up into multiple prompt submissions. So you basically say, I'm going to send you more than you can take at once. Just keep responding. I got it acknowledged until I say I've sent it all and then we'll move on to the next task. And so then you send a big block chat, G p D says, got it, keep going. You send another big block. Got it. Keep going. And when I finally have fed it all, then I say, I'm done. Now here's the task. That's
Geordie (25:34):
One. And there's a lesson. There's a lesson right there in terms of how you, like you're saying here, I'm going to give you a little bit and then a little bit more and just acknowledge me as you go in natural language just saying, Hey, yep, this is what's going to happen. Be ready. And it does it
James (25:48):
The other way, and I'll talk a little bit more about this later too, is you could take that word doc, turn it into a pdf, and there's a plugin that the GP chat, GT four beta has affiliation with a little web app called Ask your pdf. You can upload the PDF and have ChatGPT, go read it.
Geordie (26:08):
Wow,
James (26:09):
That sounds cool if you're thinking about a PDF of this transcript, but also consider that it could be a 1000 page PDF and it'll read that pretty fast too. So the things that you could do now that it can parse a PDF are pretty wild.
Geordie (26:23):
And just to clarify, so the ChatGPT, you get 3.5 for free when you create an open I account for 20 bucks a month, you get access to chat GT four. Yes. And all the plugin capability and the web access and all that kind of stuff, right?
James (26:37):
Yeah. So as of right now, we're recording in June of 23. I believe four is either still in beta or sort of public, but only accessible to premium subscribers. And I think that's 20 bucks a month like you said. And then the ability to search the web and the ability to use these third party plugins are beta features, but you have access to them if you're a paying subscriber,
Geordie (27:00):
Best 20 bucks you'll ever spend. Oh
James (27:02):
Yeah, I got my money's worth so fast. Okay, let's do another one. I think that ChatGPT has been really helpful to me in acting as what I called a helpful adversary. Meaning poke holes in my idea or read my article and tell me where it's missing something or what would you change about this. So here's an example of one that I did yesterday. I'm just going to read you the prompt and then we can talk about what happened. I said, your job is to play the role of an enterprise technology marketer in search of a partner to help with lead generation. I happen to know one of those. That's why I was asking, read this page and tell me what objections you might raise to working with this lead generation provider. And then I gave it the link to our lead generation sort of pillar page on the ActualTech media website. So it went out there, read this product page and came back and gave me a list of eight questions or objections that it might raise and a paragraph about why every single one of them is a legitimate objection that we've had real clients or prospects raise. And we have good answers to a lot of them, but just the fact that G P T was able to come up with those objections still boggles my mind. These
Geordie (28:20):
Are good. Wow. So these objections that you might have or concerns you might have as a marketer, you can, excuse me, these concerns you might, you can turn into an FAQ on your page, on your landing page. Yeah. So
James (28:37):
This is step one. What do you do with this? Right? Well, you go back and go over the product page and see where you can handle any of these objections in line. Could you copy, could tweak two words or add a sentence in the copy that just handles the objection right then and there. And like you said, you could do an F A Q at the bottom of the page or in a separate area that answers all of them. You could also take this and deploy those things in the content that you're creating and campaigns that you're building where you're proactively addressing those. Not head on, but you know what I'm saying, your messaging handles those before they even come up.
Geordie (29:10):
Wow.
James (29:11):
So it can inform your plan too. Well
Geordie (29:13):
This is humbling this, I wrote a lot of this copy. It's like, yeah, these are the concerns I might have and questions that might be going through my head as I read your marketing copy and it looks like I got a little bit of work to do, so I'll be back in
James (29:24):
A few minutes. You're fighting the robots now, but just think about you could have over the course of the next year had a real prospect bring up one of these objections in a conversation and realize, oh, I should go fix the page. So it says something about that and it would take you months and months and months of wasted opportunity, or you have it run over this, you got these eight, you could go handle 'em today and your product page is way better today thanks to this objection. So that's one example of where you could apply the principle, but the principle is the key here, not that one use of it. Where can you ask ChatGPT to poke holes in your plan or in your content or what you've created and then improve it before it even ever goes out to the world and it'll be that much better for it.
Geordie (30:09):
Amazing. What else you got?
James (30:12):
Okay, this one's pretty straightforward, but we talked about earlier ideation chat. G P T is a great brainstorming partner, so you can tell it, here's what I'm going to attempt to come up with, here's some background information. Help me come up with some ideas. And as I said that I noticed the way I said it would not have been a good prompt. One of the things you said at the top was be very specific. You don't want to leave any room for ambiguity or being vague. I said, give me some ideas. A way better prompt would be, give me 10 because especially when you're ideating, you probably want a lot. And if you don't say how many, it'll probably give you just a couple. So if I were asking that again, I would say, here's what I'm going to do. Here's some background. Give me 10 ideas.
(30:53):
Great go. I find that especially helpful for things that are not in my wheelhouse. So if it's something that I'm really good at, passionate about, I can ideate pretty quickly, but oftentimes I find myself working on something I'm not that good at or an expert at, and that's really hard. That's where I would really see this shine is give me 10 ideas about something that I don't know much about and then I can run with those. So as an example here, yesterday I asked it to write me a radio jingle for actual tech media and it gave me a three verse radio jingle that's like, it's not amazing, but it's pretty good. And it would've taken how retro it would've taken me all. Do you guys know what radio is?
Geordie (31:32):
Yeah.
James (31:33):
It's like, okay, anyway,
Geordie (31:34):
It's what we're doing now.
James (31:36):
It would've taken me all day to do this. And is there a point? No, not really, but my point was it was easy for it to do something that was hard for me.
Geordie (31:44):
Yeah.
James (31:46):
Okay. I got a couple more and then maybe we should move on. I talked about sucking in lots of data and making meaning, meaning out of it being easy for AI and hard for me, something that I've been playing with is we run a lot of surveys and then a human does the analysis on the surveys. I've been playing with feeding the survey results to ChatGPT and having it do the analysis. And what's interesting about that is it just can have a much broader view and see it all at once in a way that I can't. And I could see either ChatGPT or some kind of custom built tool that's just for this. I've been using, I've been playing with one that's called Deep Talk. It's specifically for analyzing big data sets like this. And so
Geordie (32:31):
When you say you fed it, the survey data, what did you feed it? Like a spreadsheet of the raw results?
James (32:37):
Yeah, in this case it was a C S V where there was the question and then each line had the open text field answer and it didn't honestly have to be super clean because I guess it's smart enough to figure it out. But yeah, I uploaded a CSV of an export from the survey platform and it read that you fact, I'm talking about deep talk chat. GPD can't read a CSV right now.
Geordie (33:07):
We do a ton of webinars and you got to write an abstract, you got to write the high level value, all that kind of stuff. Could you feed it a PowerPoint and have it that my SME is going to present and have it create at least a framework that I could write an abstract based off of.
James (33:25):
You can give it anything. You can pdf. So what you could do is just PDF that PowerPoint, have it read over that and then write an abstract. Oh, a hundred percent that would work.
Geordie (33:34):
There's a time saving.
James (33:35):
Yep. Yep. Copywriting of all kinds. So you just mentioned landing pages. That's one I was playing with recently is like go read over these 10 actual tech media event landing pages, understand our tone, our style, and then here's an upcoming event. Write me the landing page copy for that one. That's great. Another one is any kind of ad copy. It's better than me at writing ad copy a hundred percent. And I've studied how to write good ad copy.
Geordie (34:09):
You have an Ogilvy book on a shelf somewhere, right? A
James (34:11):
Couple.
Geordie (34:12):
A couple. And so too, I guess that comes to ad copy for a content asset that you're advertising for content syndication, something like that. So that could be, so you could feed at the white paper and then once it reads through the white paper pdf, you could have it right an abstract and bullets of what you'll learn for the landing page for that, as well as the pitch email as well as some social snippets that it can use to promote that content asset.
James (34:46):
And the order you do that in is important because remember, it's learning as you go. So you do it in the same way you would need to do it first. You give it the white paper and say, read this, summarize it, give me the key points When you're satisfied with the key points, then you say, now write me a one paragraph abstract for this piece. Now write me some Facebook ad copy for this piece. And because it's already summarized it, now it's sort of building that ad copy based on its summarization, if that makes sense.
Geordie (35:14):
Yeah, that makes sense. Good stuff. So the steps, the order you use is important. That's
James (35:17):
Good. Yep. What else? Competitive intelligence gathering. So this is fun. Yesterday I asked it, here's a couple of my competitors, go and do research on us and our competitors, find out where our strengths and our weaknesses competitively are. And then it gave me a pretty good list here of things we could say about ourselves, where we stand out and things where they might stand out and we should avoid that topic or bolster our story there. So something that would be really easy to do. It worked it a company level where I think this would be really powerful is at a product level. So if you say, I sell a firewall, here's five other firewalls in the market, tell me as you go read the product pages where you see us strong and where you see us weak. That'd be a really useful exercise.
Geordie (36:06):
Can I ask an idea just came to mind, could you, could you have it build an ROI calculator? So for instance, I know that it writes code, could it code an ROI calculator for you based off a spreadsheet where you had the raw data?
James (36:26):
So I don't know enough about writing the code myself to know whether it's good, right? My guess is the answer is yes as long as you had a developer to oversee it and make sure that it was actually working correctly. But hypothetically what you're describing, yes,
Geordie (36:44):
Because I mean there are companies that do this. You could and if you could do it. Okay, just curious.
James (36:51):
Yeah. And that principle applies to all of this is the scary amazing thing about this is probably everything we've talked about. There are companies that do this and what they need to figure out and what we need to figure out, and probably everybody listening needs to figure out is if it can do all this, where do I come in? And the answer is, you mentioned it, you need to be the one with the idea, the big thinking to tell it what you want it to go and do you do the thinking so that it can do the doing
Geordie (37:20):
Master the prompts rule the world.
James (37:21):
Right? Okay, so we're out of time. Let's talk about prompts because prompts are the key and that applies very much to ChatGPT, but also to other things like if you're doing image generation with Mid Journey or Dolly or something applies there, you're going to get the quality of results that matches the quality of your prompt garbage in, garbage out. So let's talk about good prompts real quick. One thing I find to be really helpful is to tell ChatGPT to assume a persona. So then it thinks within a certain context. So I'll say imagine that you are an expert enterprise technology marketer. You work in a company that is this size at this stage. Now your task is to go and write me some ad copy, but when I put it in that box, then it thinks that way instead of writing some ad copy that would work for a B2C brand.
(38:21):
One way that I found helpful to get better at that is I found big list. We can maybe put this in the show notes. There's a GitHub that is cataloging. There's probably like 200 different personas and the prompt that causes it to assume this persona and there's everything. It's like airline, pirate, pilot and radiologist and brick mason, all everything you can think of. And I'm not going to use those necessarily, but learning how to get it to be those helps me to learn how to make it be the ones that I want. So right. That was really cool. There's also a tool that I found that will just do it for you. Perfect. The prompt for you. It's called Prompt. Perfect. Immediately enough. And we'll
Geordie (39:06):
Stick all this in the show notes.
James (39:08):
Yeah, we'll put this in the show notes, but prompt. Perfect. You'd go just right in the box what you're thinking, here's what I'd like it to do. And then you say, perfect my prompt and it goes and rewrites it in a way that is going to be more effective for ChatGPT. Nice. That's one of the plugins that the beta plugin integration right now has. So I'm personally doing it on the Prompt Perfect website, but you also can just have that happen right in ChatGPT. So you send your prompt, it reaches out to the third party plugin, perfects your prompt, sends it back, and then actually runs, executes the prompt. Wow, that's a good one. I talked about sometimes you want to provide way more in a prompt than ChatGPT can take at once. We're going to put a link in the show notes to a little tool called chat, G p T, prompt Splitter. And basically what it does is it writes you the instructions to tell chat G, to tell ChatGPT that you're going to send more than one thing and then it chunks it up automatically for you. So all you have to do is say, copy from the prompt splitter, paste it into your prompt, enter copy, paste it in, enter, and it runs through the whole sending it the entire prompt and all your source material. So I found that really helpful.
(40:22):
Querying PDFs we talked about earlier, again, there's a third party tool called Ask your pdf. You can use it directly on the web app or that's also got a plugin that sometimes works, sometimes doesn't because it's in beta. But anything that you can pdf, you can have ChatGPT go and read. So your example was a PowerPoint, pdf, the PowerPoint, give it to ask your P D F and then say please summarize this presentation for me and it will go and read all of that and then come back with a summary with a couple bullets. Awesome. And then one more, we didn't talk much today about image generation, but Mid Journey prompting or any of the others is also totally an art
Geordie (41:03):
Mid journey is Chachi p t for Im creating Images basically,
James (41:07):
Right? Yeah. More or less. You tell it what you want in English and it gives you that in Got it. A graphic. There's a tool that we'll put in the show notes that I've played with that creates better mid journey prompts for you so you can get more of what you want from there too. So
Geordie (41:22):
Nice. So you're not so freaking out about stock image rights and all the rest of
James (41:27):
It, right? Yeah. I mean that, that's one of the most perfect use cases in my mind is, Hey, I need a blog header. So you give it some information about the theme of the blog aspect ratio, go and it'll create you a blog header
Geordie (41:39):
And maybe we get away from having the same data center pictures from Shutterstock that everyone else is using. Yeah,
James (41:44):
Exactly. There's
Geordie (41:45):
Only 10, we get the point. Oh James, this has been amazing. So this is only the beginning. We've got so much more here. I mean, this is half the document that we talked about, so let's do part two soon. But thank you so much for the time you invested to play around with this to help us learn how to use it. I think this is how to 10 x yourself as a marketer with a smaller team with less resources, less budget. This is the answer. And
James (42:12):
Yeah, I think unfortunately we're looking at a market right now where we're seeing marketing teams in the technology companies run pretty lean, and that's hard. But at the same time, I think it's an opportunity for you because if you can learn how to really optimize and maximize what we talked about today and get a whole lot done, that's going to give you a serious advantage in the marketplace in a time where you need one.
Geordie (42:42):
Awesome. Well thanks so much James. That's it everybody, thanks for listening to Tech Marketer Live and we will see you next time.