Here is a transcript generated by Otter.ai of The Global Marketer podcast episode about practical ways to use AI in content marketing:
Naomi Bleackley 0:00
I think there really needs to be strict regulations on AI. I think part of the issue is that AI is moving faster than the law is, so the law is not keeping up with all the different things that AI can actually manage to do. So there are no strict guidelines. Is AI plagiarism, if it’s taking, you know, inspiration, at least from a company perspective, I think you should hire someone like me, have someone in house that can kind of keep a general eye on this and make sure that the team is using it responsibly. We can’t control what other people do with AI, but we can control what we do with it. AI isn’t perfect at all. It’s not going to replace humans, mostly because it’s not a human. It doesn’t think like a human. It doesn’t work like a human. It doesn’t understand tone, it doesn’t understand nuance. It doesn’t actually have comprehension. It’s all based on statistics and probability. I very recently did a translation, and it had a play on words. There was no way to translate it literally. And I was like, Well, you know, who could maybe help me? My friend, chatgpt. I don’t consider it cheating, because the output is only as good as the prompt that you give up. We tested so many tools. A lot of work went into it, actually, but the reason we kind of went for chatgpt in the end is because
Shaheen Samavati 1:06
Welcome to the Global marketer, the podcast that helps you achieve success in international markets by learning from marketers on the ground and the tactics that have and haven’t worked for them.
Kegan Gates 1:15
This podcast is brought to you by Vera content, an agency specialized in creating web and social media content for brands across multiple international markets. I’m Keegan gates, marketing specialist at Vera content, and
Shaheen Samavati 1:26
I’m Shaheen Samavati, CEO of Vera content, on each episode of The Global marketer, Keegan and I have a conversation with a guest around a global content marketing topic.
Kegan Gates 1:35
Today, we have a fascinating discussion with our colleague Naomi Bleakley, editorial AI and Process Manager at Vera content, we dive into how AI is transforming content development, especially in multilingual projects,
Shaheen Samavati 1:48
keep listening to learn about the practical applications of AI in content creation, the balance between human expertise and automation and what marketers should consider when integrating AI into their processes. So that said, we’re gonna stop reading and jump right into the interview.
Kegan Gates 2:05
Welcome to Madrid.
Naomi Bleackley 2:06
Thank you. It’s great to be here.
Kegan Gates 2:08
So your role, editorial AI and Process Manager, is a new at VeraContent. It was recently created. So can you explain a little bit? What do you do?
Naomi Bleackley 2:15
Good question. So yeah, we basically created this role super recently. I think I started in October, only officially, but it kind of came off the back of AI task force that we had in the company, which I’m sure Shaheen will get into more detail about. But I realized I was very interested in AI, in how it worked, and also in how it could work with content creation, because I’m also a linguist, so yeah, basically that kind of happened, and then we created the role, and now what I’m in charge of is kind of implementing AI into our processes, finding ways to implement AI, researching different tools across all teams, Both like in content creation. And also, you know, more like administrative kind of processes, and yeah. Also on the editorial side, I kind of have a general overview of the quality of the work that we’re producing. I do like quality checks, make sure that everything is, you know, up to Vera content standards, and yeah. And I also train linguists and make sure that everyone is, you know, up to scratch and on board
Kegan Gates 3:25
with exactly. And Shaheen, I have a question for you then on that, no, you were part of the creation of her role. Why, then, was this an important step for us to take?
Shaheen Samavati 3:36
Yeah, well, so Naomi talked a little bit about about the task force that we started, which is at the end of 2022 and I think that was a time when every company in the world, I think, was scrambling to figure out how to implement AI in their in their processes, and I think a lot of a lot of us still are so but obviously working in content creation, I think AI for content creation wasn’t something that like something that we’ve actually been Looking to into in the past. Because even before chatgpt came onto the scene in 2022 we’d been looking at some of these tools, like copy AI and writer.com and things like that, but we never really took them too seriously, because the quality of the output wasn’t that great. But I think with chatgpt, it was like, wow, that’s it was like a leap in terms of, like, what kind of output you could get from it, and I think that’s what got everyone’s attention. And I was like, Okay, we have to do we need to get on this train. We need to figure out, like, this is gonna change things. Like, in our industry, we need to figure out how not to get left behind, but at the same time, like, stay true to our values, because and what we do as a company, creating quality content and working hand in hand with our clients to really represent their brand faithfully that’s so important to us. So it’s like we needed we decided to basically put this team of people together to look into how we can take advantage of tools like chatgpt and do it responsibly and in a way that reflects like the kind of work we want to do. And basically, Naomi was the star of this past. Of course, we were a cross departmental team where we basically spent six months meeting every two weeks, testing different tools on different aspects of the business people from each department. Naomi was representing the projects team, which is the team at Vera that works on client projects. And yeah, she really went above and beyond in researching the tools and really learning how they work, learning the technology. And that’s why we decided to create this role, role, especially for her, because we saw it wasn’t going to end there. At the end of the six months of that project, we needed to have someone responsible for actually implementing with the project. So now she’s working with all the project managers and identifying like, Okay, where can we on each of your projects and your processes? Where can we use AI to help you through your job better? Yeah.
Kegan Gates 5:47
So then going back to this task force that sounded super cool and really fun, actually. What tools did you test, and why did we decide, as a team, the task force team that is in as read content. Why did you guys decide to focus on the tools that you did, which specifically, spoiler alert, is chatgpt, yeah.
Naomi Bleackley 6:08
So basically, we tested so many tools. I think it was maybe over 20 tools or something. We tested a lot of tools. A lot of work went into actually, we tested tools like writer, like Jasper, AI copy AI tested lots of different tools, but the reason we kind of went for chat GPT in the end is because it really does just kind of cover everything. You know, all the other tools, they’re quite specific. So writer is, as the name suggests, very good at writing content. But then to, for instance, create a social media content calendar. Maybe it’s not that great. Okay, so copy AI is really good for short form content, but you know, if we wanted to then write a blog post, maybe it’s not the best fit. So in the end, after researching all of the different tools, we decided that chatgpt Actually probably had the broadest range of functions, which adapted to our team the best. And also it’s, you know, it’s important to say that chat GPT price wise as well, was more affordable for our team because tools like, you know, writer or Jasper, they come with a price tag as well. And for a small content agency like Vera content, you know, sometimes we also have to make these kinds of decisions where, you know what’s going to make most sense, right? And
Kegan Gates 7:21
we have some really good writers on the team, they must say. So like, writer may be good, but come on, we’re good. So
Shaheen Samavati 7:29
yeah, and I wanted, I wanted to chime in on that, that when we first started researching the tools, we were kind of looking for alternatives to chatgpt, because we we wanted to have collaboration tools that, like, different members of the team could work on, work with. But then chatgpt introduced the custom chatgpt feature, the custom GPT feature, I should say, and that made us look at it a lot more seriously, because it was a way for us to create specific kind of well, I think Naomi can explain this a lot better.
Kegan Gates 7:58
Well, yeah, and as well the challenges that come with that, right? There’s a lot of challenges with it. And, I mean, we’re what, what is AI? Is it gonna replace me?
Naomi Bleackley 8:10
No, oh, great. Now, no, AI, the at least at Vera content, the way we want to see AI is as a tool, as a friend, as a helping hand, something that can actually not only make our jobs easier, but also make our work better. Because the reason AI is so controversial is because a lot of people think that, oh, AI is going to take my job. AI is just going to do what I do, and then, why am I even here? That’s not what we’re looking at it as. That’s not the kind of approach we want to take to AI, yeah, we really want it to kind of assist our linguists and also help us produce consistent content as well. So the different clients that we work with, they all have a very specific tone, very specific brand guidelines using chat GPT, and especially custom gpts, where basically we it’s just to tell you kind of what a custom GPT is. Basically it’s a it’s like chat GPT. So if you know how it kind of is, more or less it’s basically a chat bot which generates content for you. The custom GPT means that it’s a chat bot that takes into account some kind of pre trained instructions. So you give it the instructions. So basically, like always create content between 709 100 words, for instance. So the output that you’ll get will always be according to his instructions, so you don’t have to re feed it all the instructions from the start. So it really helps with consistency. Really helps us create content that you know is up to a certain quality to say, and then it helps the writers afterwards edit.
Kegan Gates 9:59
So earlier, you told me, AI is not going to take my job, right? And but then you just told me how amazing it is. So what are the limitations, and how are we defeating those?
Naomi Bleackley 10:08
So AI isn’t perfect at all. It is. It’s a great tool to have, but it’s it’s not going to replace humans. Mostly because it’s not a human. It doesn’t think like a human, it doesn’t work like a human. The answers that it produces, they may look like human answers. They look very fluent, but the reason they look so fluent is because it’s trained on like, you can’t even imagine how much content like. Well, it’s like petabytes of content, like so much content is trained on that basically what it’s done is it kind of has an internal it’s called an algorithm that basically means it can predict what word will come next, wow. So it will basically, you know, after the word with it’s very likely that there’ll be a name after it, or an object or an instrument, something like that so it can predict things, which means that it like the answers it gives you can be like, yeah, like, that sounds legit, but it’s probably not, or it might not be, because it’s not really a human that’s understanding. Doesn’t understand tone, it doesn’t understand nuance. It doesn’t, you know, it doesn’t actually have comprehension. It’s all based on statistics and probability, okay?
Kegan Gates 11:20
Okay. And the statistics of probability clearly don’t have feelings. So if I’m smiling in a happy way or in a sarcastic way, it wouldn’t be able to pick up on that exactly, not yet,
Shaheen Samavati 11:29
especially not touchy. BT, it’s not looking at you yet. Well,
Kegan Gates 11:33
thank God I’m in my home office. Nobody wants to look at that. No, okay. But Shaheen, so then at an agency like ours, at Vera content, what are some of the use cases that we can use and that we can implement? Yeah,
Shaheen Samavati 11:47
well, I think we were talking about a little bit obviously being a content creation agency, like using chatgpt at different aspects of the content creation process is something that we’ve been looking into, but we’re looking at how to do it, basically as Naomi was saying, how we can use it as a tool, not as obviously a replacement for human writers, how to make it an assistant for them. But I think Naomi can probably speak better to like specifically at which places in the content creation process that we’ve actually used. Yeah.
Naomi Bleackley 12:17
So I actually want to talk about something really interesting that we did, and it was actually part of the AI Task Force, one of one of the initiatives that we had, we wanted to create, well, kind of customize a tool that would have a client, a specific client style guide integrated into it, so that when we were writing it would pick up any specific Like style guide errors, or anything that you know wasn’t quite right with the style guide, we actually tested, we tested a lot of tools for that as well. Actually, in the end, we went with Grammarly. Grammarly only works in English, which was one of the main limitations that we had. But for this specific client, it was very useful, and we’ve actually had really good feedback on it, because basically, you give it a set of rules. So, for instance, you know, you can’t say this word, like, this word is banned for this client. For instance, sure. So every time you write that word, it will actually highlight it in red. So like, No, can’t use this word for this client. So this is, like, one of the capabilities it had, which was really useful, because we do have some clients that have, you know, very specific style rules and things that can’t be said, or things that should be worded in a certain way, or, like the names of institutions, things like that, or, you know that you can’t use acronyms instead of full names and things like that. So it’s been very, very useful for that. But then Grammarly also has more, like aI capabilities integrated into it. So it has, for instance, like, a rephrase option. So when you’re writing, you know, sometimes you write a sentence and you’re like, something, I know, I know this could be better. I don’t know. I don’t know why, I don’t know why I don’t like it, but I know it could be better. You can, actually, you don’t have to tell it to just, oh, rewrite it for me, and that’s it. No, you can actually tell it, give me a few options for this. And you can even, like, cherry pick, like, Oh, I like the beginning of this sentence, the middle of this sentence and the end of this sentence. And then you create your own sentence that actually, you know, it gives you more ideas. Sometimes we get into, like, writer’s blocks, and, of course, things like that. And it can really, really help you, you know, kind of pick up these things. And then another cool thing that Grammarly does is it’s not just like a spell check, it’s also a grammar checker, and it gives you suggestions, not just on pure errors, but also on things that you know. Like maybe this could be worded differently for clarity, or maybe if you word this a different way, it could be more effective, it can sound more confident, more confident, more positive, things like that. It gives you all of these suggestions. They’re all actually color coded as well, which is useful. Yeah, absolutely. And, and you don’t have to use the suggestions either. You
Kegan Gates 14:51
can’t, because I, you know, a little secret, not only am I a marketing specialist, also on the projects team, so I’ve been using this and I was part. Of the like, testing of it. As you know, the team started to decide whether or not we’re gonna use it, and sometimes they would pull up some random things that had nothing to do with what I was saying or what we wanted to say. And so it made me feel secure that there were Arsenal limitations with AI. But it’s a very helpful tool, because sometimes, like you, when you’re editing something, it just you may miss it too, and it, yeah, picks up on, like, little nuances that might Exactly. But
Shaheen Samavati 15:24
are you saying, Keegan, that the kind of the Rephrase feature you didn’t find to be that useful? So
Kegan Gates 15:28
the Rephrase feature, I like a lot because it well, specifically with one of the clients that we use, they don’t like passive voice. I mean, not many people like passive voice, so it’ll flip them, it’ll flip the sentences, if there’s passive in there, or a lot of times, it finds important things. So sometimes I’ll use it, and then sometimes I won’t. Because, you know, if you use the there was a specific one that I just can’t really remember it, but I was like, Oh, this was quite interesting. Or or this is the key concept. And then they change key to crucial. And for me, key concept and crucial concept are very different. So why are you trying to change that? So, you know, you have to be really careful, and you really have you really have to use it. But I think that’s one really cool thing about our company, is that we our processes make it so that there’s a first person who writes the actual content, and then the next person makes sure that it’s up to quality. And so then now we have this like third wall that, you know, and I just really think it speaks to the quality of our of our content, yeah, which is cool. And then also the linguist gets to decide, or the second editor we call them gets to decide, okay, I’m gonna use this rephrase option. Or, nope, my the original writer wrote it better. So
Shaheen Samavati 16:32
I think, like, million dollar question. So we know that, like, we have all these tools that we can take advantage of, and everyone wants to use them to be more efficient. Yeah, is it actually up to now? I guess making us more efficient, does it take less time to create an article if we if we use something like Grammarly as part of the process, or a custom GPT or chatgpt,
Naomi Bleackley 16:50
I’d say Grammarly maybe, rather than making it more efficient, is actually more of like an extra quality layer, especially at least in the way that we’re using it. With the star guy with the grammar checker, it might not make things faster, because actually, if you’re rephrasing things and you’re like checking through different options, you might actually take a little bit longer to write an article, but I think the end result will probably be better. But in terms of custom gpts and things like that, I 100% think that it can make the process of writing articles or just producing content a lot quicker things well, GPT in general, I think it can really help with creating ideas or creating, like an outline for an article. It’s never going to be the finished product, and that’s not our aim. We don’t want it to be the finished product, because we want that human element in there. But if we can kind of say to it, okay, I want to write an article about, I don’t know, education, give me some, like, talking points, for instance. So I can give you those key talking points, and then you, as the human writer, you then flesh it out. You can even something I’ve actually used it for, is I’ve written a whole article myself, and then I’ve been like, Okay, I’m not, I don’t know, like, Could this, could this be improved? So you could, like, give the chat, GBT your whole article, and then it can maybe, like, pinpoint things that, oh, like, this sentence could have maybe been worded differently, or this isn’t actually very clear. This is ambiguous, things like that. So I think in that sense, it can actually make things faster, not always going to make things faster, but if you use it right and you have the right processes in place, I think that it can not only improve efficiency, but also quality. And
Kegan Gates 18:35
what about for translation? Let’s go there. Let’s go to translation and how the AI uses that and works with that as someone who also works in translation. Am I in trouble?
Shaheen Samavati 18:49
So is the question like, how? What’s the use case for translation in AI? Or is the question about, is AI taking translation jobs?
Kegan Gates 18:56
I think we should tackle both. What do you think Job. Shaheen,
Shaheen Samavati 19:06
well, so again, that’s kind of going, okay, that’s going into a, kind of a different topic, but Okay, let’s go there. No, I think, I think it’s changing jobs, absolutely. And I think that was something we’re kind of speaking to earlier. It’s like, I think everyone is racing to adopt AI at the same time, everyone’s a little bit afraid of what it’s going to bring. And obviously we work with a lot a lot of linguists, both writers and translators, and a lot of them have been resistant to adopting these tools. But I think increasingly, as time goes on and we see how mainstream it’s becoming everyone on our team is realizing, okay, we need to embrace or get left behind, essentially 100% and I do think we saw the same thing with machine translation a decade or so ago at this point, and that really changed. It completely transformed the translation industry. And the nature of a translators job today has completely changed. I think for translators, they. Were already disrupted, essentially. So I think maybe this next stage of generative AI isn’t impacting them as much as it is writers like writers, of course, they thought, oh, like, that’s not, yeah, creative people, we like machines can’t do our job, and we’re starting to see machines are becoming able to do some pieces of creative work. We’re seeing, obviously, writing being disruptive. We’re seeing graphic design and video editing and all the stuff that that AI can do now. So but pass it to Naomi if you have any thoughts on it. Yeah,
Naomi Bleackley 20:30
I would say for me, for instance, like as a translator, I would much rather work with AI than with just playing machine translation, because, by machine translation can be really bad, like it can be absolutely, very, very bad, whereas, if it’s, I mean, it’s still machine translation, even if we’re using AI, but if it has the AI element to it at least, at least the outputs gonna sound okay. It might not be accurately, you know, it might not be good, or it might not be 100% what the meaning is, but it means that the the output is at least going to sound a lot more fluent than using machine translation. And use machine translation, just pure machine translation. A lot of times the output is it’s just gobbledygook, like it doesn’t it really makes no sense, whereas with AI, at least it’s gonna make sense. It might not be correct, but it’s gonna make
Kegan Gates 21:20
sense. Did you give us, sorry, a different like, what? What is the difference here between purely machine translation and using AI as translation? Because there are some people out there who probably have no idea what that means. That’s a that’s a whole or something, yeah. So like, can you maybe just, like, drop a name, what’s a machine translation? Like Google Translate. Google
Naomi Bleackley 21:39
Translate uses machine translation, okay? And
Kegan Gates 21:41
then an AI translator would be something like chatgpt. Chatgpt can
Naomi Bleackley 21:44
do translations. I think a lot of like companies that were classic machine translation are now integrating AI into their tools. So deepl, for instance, now has like deepl AI. So they’re kind of combining both things. So combining the, you know, machine translation network, or however you want to use it, yeah, and AI capabilities as well, to then make the output more fluent, whether as a translator, whether that’s better or not, I don’t know, actually, because, at least for me, I prefer it, but because I’m the kind of person I like things getting done better quicker. That’s just the way I am. But it is true, though, that in terms of creativity, it probably does, for a translator, probably does feel like, you know, your creative agency is kind of limited from you, yeah.
Shaheen Samavati 22:32
But as we said, going back to the idea generation side of AI, I mean, it’s it gives you a suggestion you don’t necessarily have exactly, and so if it gives you a better suggestion, like, great. The thing is, as you were saying, it’s not always accurate. So, yeah, maybe straight machine translation, in some cases, might be a more straight translation that’s more likely to be at least somewhat accurate than AI, which might give you something crazy, although you did an interesting test. I wanted to mention, I like testing these, testing AI output for, like, slang translation. Yes,
Naomi Bleackley 22:59
interesting. Yeah.
Shaheen Samavati 23:01
And what did you find there? So,
Naomi Bleackley 23:03
yeah, I did a, I wrote my own, kind of just quick text, and it was very, like, very slang, very Gen Z words, things like that, trying to, you know, trying to make it as complicated as possible. And I tested Google Translate deep out and chat GPT, I think I just did those three. And I kind of did, like an analysis of basically, which output was the best, and chatgpt output was definitely the best. Like, by a long shot, picked up on the pensie exactly it picked up on the sign because it’s been trained on that kind of slang, whereas machine translation has not, probably hasn’t. Yeah,
Kegan Gates 23:36
have you seen the talk show host? I don’t know this could be I could be showing my American here, but I think it might be Jimmy Kimmel, where he puts things through machine translation and then, like, three or four times, and then reads it back and it’s just quite No, nobody, no
Naomi Bleackley 23:50
or like, the
Kegan Gates 23:51
song lyrics. There’s also that where people put song lyrics in it into machine translation a few times, and then bring it back into English and try to sing along with the song. It’s so funny, the things that come out of that, definitely. But
Shaheen Samavati 24:01
when was that right? Because now maybe
Kegan Gates 24:05
I don’t know, we should test it.
Naomi Bleackley 24:07
They make songs with AI now as well. So you never know. Even some actually really interested that you said, Shaheen, that was that if it gives you a better suggestion, then you can take it. I actually think that’s where a lot of the problem lies, that people, at least as a translator, you might not want to necessarily admit that it’s a better suggestion than what you thought than what you thought of. You know, like, I think it’s sometimes really hard to be, I mean, it’s writing, so you can never be fully objective. It’s always a subjectivity kind of element in there. But, you know, sometimes there might be something that it does word better, but kind of, I don’t want to say people are too proud, but kind of as a sense of, like, Oh no, but I definitely can come up with something better than that. You know, I’m human. I’ve trained for this. I’m, you know, I’ve studied this. So I think that could be part of the resistance as well, that it’s kind of like sometimes the thing. Things that AI produces are actually decent, yeah, but do you actually want to admit that they’re decent? And how much do you actually want to involve yourself in, you know, the output that it gives, and how much do you actually want to edit? It’s kind of a and
Shaheen Samavati 25:14
sometimes as well, I think it can make people a bit lazy, because it’s like, it gives you a good enough suggestion. You’re like, Okay, I’ll just go with that. But actually, maybe you could have thought of something better, yeah, yeah, there’s
Naomi Bleackley 25:24
both sides. Do it really, yeah, well, and
Kegan Gates 25:26
that’s, I guess maybe that also is to say, the difference between translation and transcreation, right? And so if you’re just doing strictly, like, okay, they want this as tightly possible, as tight as possible, to the original, then that’s one thing. But if we’re going with transcreation, we want to make it sound as, you know, unique as possible, and as if some really awesome native writer wrote it, yeah, and like, that’s the line that Yeah.
Naomi Bleackley 25:53
Actually, I do think chat GPT can be very useful for transcreation, funnily enough, because transcreation, especially when used in marketing. So a lot of times there’s slogans, there’s play on words, things like that. Um, I very recently, I’m not gonna name any names or anything, but I very recently did a translation where there was basically, like a title of something, and it had a play on words, and I was trying to come up with, you know, alternatives. And I was trying to think, like, oh, how can I, like, have this same play on words, but then, but in another language, because it didn’t work in I was doing Spanish into English. It didn’t work in English, like, there was no way to translate it literally. And I was like, Well, you know, who could maybe help me? My friend chat, GPT. And I don’t, I don’t consider it cheating, because I, firstly, I had to put in the prompt. I had to say, you know, like, Okay, this is the source. I want it to be this kind of play on words. It needs to be related to the same topic. But it doesn’t have to be exactly the same. You know, you do need to give it very good prompts, because the outputs only as good as the prompt that you give it. Oh, love
Kegan Gates 26:59
that. Yeah. Going back,
Shaheen Samavati 27:01
yeah. Going back to that topic. I was wondering, because we started to talk about custom gpts, but I was wondering if you could talk a little bit more about what goes into setting up a custom GPT, and when you’re when you have multiple people working with it, I mean, like, what do you need to keep in mind when you’re creating kind of a custom GPT for a team to use?
Naomi Bleackley 27:19
Yeah? So it’s actually pretty easy to create a custom GPT, at least on open AI side. The way the interface kind of works is you basically have a conversation with it. There are two, there are two sides to it. So one side is you can basically almost, it’s not programming, but it’s more, you know, like you, you put in bullet points, and you say, right, this is this. This is this. But then it has another interface where, basically it’s, it asks you a question. It’s like, How can I help you today? What would you like this GPT to be? And then you say, Oh, I’d like it to be this. And then it says, okay, great, let’s name it this. And it gives you like suggestions, even gives you like an image, like, what do you like this logo for your GPT? And it like creates it for you everything. So it’s actually quite easy to do. And I feel like it’s something that a lot of agencies could implement themselves. I don’t think you don’t require, like a programmer or anything like that, to to create a custom GPT. So it’s it’s very easy to use. All you have to do is just implement the instructions and test and then if the test that it gives you isn’t what you want. Try again. Word it differently. Keep going. You have to. There’s a lot of trial and error in it, a lot of experimentation, and a lot of time goes into it, but it’s not necessarily difficult, so you do have to keep trying, and it won’t. It will definitely not get it right on the first try, like 100% won’t. But then you have to tell it what it did wrong. Give it examples. Give it feedback. Say this part you could have actually worded it like this. Please implement this going forward, things like that. So
Kegan Gates 28:50
and it takes feedback. Well, it does. It’s very Yeah, contrary to some humans, yeah, it
Naomi Bleackley 28:54
doesn’t get defensive at all. It says, Oh, I’m sorry. It apologizes and everything. And then it fixes it and then it fixes it
Kegan Gates 29:02
and comes up with a better solution, exactly. So it really listens, yes. And
Speaker 1 29:05
sometimes the better solution isn’t still, it still isn’t what you want. So you just have to keep trying and trying and trying. And even once you’ve kind of finished the creation of, like, what you want, you’ll still have to, you know, go in. And even so, imagine your custom GBT is created you’re making, you know, you’re trying to get it to help you come up with and, like, an outline of an article, um, and it might still even, even if you’ve given it like the instructions, it might still not follow them. So it’s so then you’d have to, like, go back and prompt it again and say, remember the instructions? You know you didn’t do this. And then it will say, Oh, yes, I’m sorry, because it’s so basically, if you have like, a full list of instructions, it will start creating content, like, straight away, and it just starts reading the instructions so it might not get to the last instruction before the contents already been created. Created. Wow. So you really have to or another way to program it is to say, Okay, make yourself a checklist and make sure that every point on the checklist is completed before creating the content, things like this. So you have to really be very specific with your prompts and be very clear no ambiguity at all. Because if there’s ambiguity, it probably won’t do what you want it
Kegan Gates 30:23
to do, because it’s robot, because it’s like, that doesn’t matter to that. It’s a robot, after all. And so, I mean, that takes a lot of time, but once that you’ve done this and time after time and trial and error and trial and error and trial and error, you have it right. And then that’s it, and then you can continue using it, and it’s going to do
Naomi Bleackley 30:37
exactly what you want it to do. Well, I wouldn’t say exactly what you wanted to do. That’s been generous. The thing with the answers that that chat, GPT, or these kind of things give you is that they’re randomized, so you almost never will get the same like two same answers to one same prompt. So because it works on statistics and probability, so it’s always using these it’s always using this database of statistics. So you might say to it, I don’t know, write me a song. It will write you a song. You say it again, it won’t write you the same song twice, which is good, like in terms of creativity, because it gives you various options. But in terms of consistency, it can be an issue. So you really do have to keep an eye on what it’s giving you, like, the output, and make sure that it’s always following your instructions you have to, it’s like a naughty child that you have to, like, you know, repeat to like, No, I told you don’t do that right then and then, yeah. That’s why our jobs are safe, exactly. That’s why you always need a human interacting with,
Shaheen Samavati 31:39
yeah, yeah. One of the use cases that we didn’t talk too much about is content repurposing. And I think, like that’s you were kind of talking about how the the kind of result that you get from chatgpt is only as good as what you put in. But with repurposing, obviously you’re providing the content and you’re asking it for a different version of that. Yeah, what kind of like possibilities have we seen for that, like in content creation, in what we do?
Naomi Bleackley 32:02
Um, so for repurposing, it’s, it’s very useful, because you really can, you know, take a completely different format and putting it, put it into, like another one. So, for example, imagine you have, like, a script for for a podcast, you could give it this script and say, Okay, make me a LinkedIn post based on this script, and then it will give you, like a LinkedIn post that you can then attach to the podcast, for instance. Or you could another thing. It’s not exactly repurposing content to publish, but something that’s really useful is sometimes we receive reference documents that are, like, 300 pages long, and you can give it this and say, okay, read this and give me the key points, because maybe you don’t need to read the whole thing. And it’s really useful for that, because it basically summarizes, you know, a whole document which that would have taken you hours, if not days, to read, and it gives you the key points that you need to know. So you can’t be 100% like, if it’s something very important and something that you know that you really need to get right, don’t do it. But if it is just to get a general gist of something, just to get the idea, then it can save you so much time. So
Kegan Gates 33:16
if you ask it to do a LinkedIn post and an Instagram post and a tweet, will it know all three and be able to take the tone of all three differently? Will you get three different you will, like, really on point kind of things, or
Naomi Bleackley 33:29
whether it’s on point or not? That’s that’s up for debate, but it will, it will give you three different outputs, okay, and but it will only do it on its understanding of what a LinkedIn post should be, or what a Tiktok post should be, or something that be your brand, which might not be your brand Exactly. So the best way to do it is to give it very clear prompts. So to say, I want you to make me a LinkedIn post and it needs to sound professional but approachable, and a Tiktok post which needs to include Gen Z slang, and a Instagram post which needs to be esthetically pleasing, or something like that. And then that way, then you guarantee, not guarantee, but you you know there’s more probability that you’ll get the result that you want, rather than just giving it general instructions, like, make me this post. So really,
Kegan Gates 34:11
it’s not stripping our creativity in a lot of ways, because you have to be creative in how you’re writing the prompts Exactly. And that’s kind of a new version of creativity, a new version of fun. If we’re gonna look at Silver Linings here, of like, AI being that, that’s just me. I think, yeah, absolutely, yeah.
Naomi Bleackley 34:29
I think so too. I think actually, it can sometimes even make us be more creative. Because, yeah, especially as writers, not everything we do is creative. A lot of it is also, like, very like, formulaic and things that you know, especially like when we get a brief that’s like, Okay, this brief always needs to include, or something like a meta description, or, you know, things like that. But let’s be honest, writing a meta description isn’t the height of creativity, so getting AI to do these more, I don’t want to say boring. But these more, yeah, these more like mundane, repetitive tasks actually, then leaves right as the time to, you know, invest more time in the more creative aspects and in the more you know, the more fun parts of the job as well, because AI is doing the legwork.
Kegan Gates 35:14
Okay, so now a very important question is, What, then, is your advice for marketers who want to get the most out of these generative AI out of these generative AI tools, especially when we’re working with teams, to both of you, yes, because I know you guys both have really
Naomi Bleackley 35:38
good I would say something really important is training. You need to have all of your team trained, and you really need to document your processes, which is something that’s kind of part of my job role now, Vera is to document processes and make sure that, you know, we have everything very clearly written down and accessible for all our team members. It’s very important to set clear guidelines, especially when using AI, because there are, like, ethical considerations, and there are sometimes, you know, issues that you could run into when using AI. So it needs to be very clear how we want to use AI, like we said, as a tool and not as a replacement, and the team needs to be using it responsibly, and we can only ensure that through training. I think,
Shaheen Samavati 36:19
yeah, well, I completely agree on training. I think, I think, like, a challenge that pretty much any marketing team is likely to face is just like, how to, like, there’s no problem with getting people to use AI, like, everyone’s using it behind the scenes, but it’s like, how do you make sure that everyone’s using it the way you want them to? And that’s really like, Naomi’s entire role is like setting those standards, setting those policies, making sure everyone has kind of been told the expectations like this is how we want to use AI. We want to use it like this. We want to use it for idea generation. And this is like the best practices. This is the best like, how you can get the most out of these tools and to get the results that we all want as a team.
Naomi Bleackley 36:59
Yeah. Definitely. Also custom gbts kind of come into this too, because if you just tell them, Okay, use chat GBT, then it’s it’s so broad, whereas, if your company is in control of the instructions that the GPT has, yeah, and that it and the kind of content it’s allowed to produce, then you get to control it a lot more, and you get To make sure that you know the kind of content that you’re producing is in line with what you want and with the your brand guidelines, and
Kegan Gates 37:27
you know well. And it’s constantly updating, right? I mean, it’s moving really quickly, it’s learning so fast, and it’s constantly updating. So then we also have to stay constantly updated as well, right? We have to keep pace with its move in, yeah. And so that comes down to training and best practices as well. Yeah. So, speaking of constant updates, where’s the future? Where are we going? What’s the future of generative AI, what’s the future of how it’s going to impact our market? What’s the future of how it’s going to change our industry? Where are we going?
Shaheen Samavati 37:56
I mean, personally, I’m optimistic. I mean, I think we’ve touched on it a lot already in this conversation, but like, I definitely see AI as opening up opportunities for more creativity. Like, I think it’s a tool that frees us, frees writers up from doing some of the tedious tasks, lets us focus on, like, the actual fun creative parts. And for me, personally, I like figuring out new things, and I’m excited about, like, the opportunities it can bring a it can bring. Are there some scary aspects to AI? Yes, yes. And we really don’t know how it’s gonna like, we we’ve seen how quickly things have developed up to now, and we have no idea like, what the future is gonna bring. I mean, it’s like, it’s advancing so quickly. We say, oh, there’s some things that AI can’t do, but maybe in a couple years it will. So, yeah, it’s a, it’s a question mark,
Kegan Gates 38:42
yeah, definitely, yeah.
Naomi Bleackley 38:44
I think as well that it’s like, you know, kind of what we were saying that, yeah, maybe, maybe it will end up, you know, understanding feelings and understanding, you know, being more close to what a human thinks like. But whether, well, firstly, should it should? Will people allow that to happen? I hope not, to be honest, I hope that there will be a moment where, you know, the people who are involved in, you know, advancing this technology actually say, okay, like this is where we stop. Whether that will happen or not, I’m not sure, but I do think that, you know, in order to kind of preserve, you know, human creativity, not just in written form content, but also, we’ve been seeing AI being used in art in, like I said, Before creating songs, there’s a lot of artists have actually sued people using AI to, like, mimic a song that could be for them. Yeah, yeah. So, you know, there is a point where we do have to say, okay, like, if it gets
Kegan Gates 39:42
into the wrong hands, I think that’s where my I’m with you. I’m positive. I’m like, Yeah, let’s go. This is gonna make my job easier, and it’s ultimately gonna make me better at my job, because I’m gonna have to, like, look deeper into certain things and then other things, and it might unlock a side of creativity that, you know, I didn’t know I had, or whatever. And then also. So I went with you in the fear, in the fear side, where I’m like, if it gets into the wrong hands, what’s gonna happen? So
Naomi Bleackley 40:07
I think there really needs to be strict regulations on AI, yes. And I think part of the issue is that AI is moving faster than the law is on these kind of things. So right, so the law is not keeping up with all the different things that AI can actually manage to do, yeah. So there are no strict guidelines. There’s not really, like, I think they are starting to, you know, trying to come up with, like, is AI plagiarism, if it’s taking, you know, inspiration from from other texts, things like that. So at least from a company perspective, I think you should hire someone like me, no, but have have someone in house, even if it’s not their entire role, but have someone in house that can kind of keep a general eye on this and make sure that the team is using it responsibly. You know, yeah, we can’t control what other people do with AI, but we can control what we do with it. So
Kegan Gates 40:59
good. That was just such a zinger. And, like, I think that’s a testament to how we as a company are moving forward and how we’re shifting with the times and and I mean creating this role for Naomi, because we saw something in her that we I wasn’t part of it. You Shaheen, I didn’t create any roles. I’m just part of the company. But Shaheen, you saw something in Naomi, like, okay, she’s really invested in this and really interested, and I think this can be used for our benefit, and I think that that’s such a good point. And,
Shaheen Samavati 41:28
yeah, I mean, we’re really lucky to have someone on the team like Naomi who could fall into this role, and I’m really excited to see how it develops. Like we said, it’s only been a month, and she’s already done so much so,
Kegan Gates 41:39
and I’ve just learned so much sitting here talking to you about this, honestly, about the way that it works. And I think it’s super cool, and I use it a lot, you know, yeah, because I do right for the company. So I think it’s just, I don’t know, I’m pretty proud of the stuff that we’re producing, too, and how we’re using the tools. I think it’s really cool, and I think you’re doing a great job.
Naomi Bleackley 41:56
Thank you. Okay,
Kegan Gates 41:57
we can see where we go. Yeah,
Naomi Bleackley 41:59
I think there’s Yeah, watch this space, because I really do think that we’re gonna, we’re gonna do some cool things. I think, well,
Shaheen Samavati 42:06
that said, I think we can wrap up this episode. Thank you, Naomi, so much for sharing your insights with us today.
Naomi Bleackley 42:11
Thank you for having me and in Madrid and in the podcast. So like
Kegan Gates 42:15
Naomi said, watch our space, because we have a lot of cool things coming up, and if you’d like to connect with us, or suggest topics for future podcasts. Please reach out to the global marketer@veracontent.com
Shaheen Samavati 42:26
Yeah, okay. Thank you. This meeting is.
Transcribed by https://otter.ai