Full Transcript for #4: AI for Everyday Work: Useful, Risky or Overhyped?

those different things that that we're you know we're we're going about about talking about. We we want to come up with something that's that's genuinely useful in in our day to day. And in in my case, we were working on a really interesting project over the last couple of months that we just wrapped up this week. we had built an AI chatbot for one of the clients that we were working with. And it's an AI chatbot that actually Sinéad Walsh: That's really cool. Denis Jastrzebski: Yeah, and it's it's it's one of these ones that actually lives on your website. So we're not talking about ChatGPT, but it's it's that little bubble in the bottom right of of websites that you visit. And it's kind of something that maybe Sometimes you use if you're on a website. Most people probably you know aren't used to maybe interacting with chatbots on a website, but this one is actually really useful because we've built it in a way where it can answer all the questions somebody might have about a service or the product that this business is offering. it can also kind of bring in and collect information about them. So if they want to have like an inquiry form, instead of you having to fill that out manually as a potential customer, you can actually just have a conversation with the chatbot. First of all, to kind figure out if the business offers what you you know, what you're looking for. And then you can actually talk to the chat button and it'll ask you all the questions, which will they get sent to the business. So it's kind of an interesting way of of capturing lead information that's a little bit different from our traditional means. And I suppose with that, we're kind of talking about making more time available to the actual business owner or whoever's operating the business. So now they're not getting asked that same question, or maybe they don't have to ask those questions to the potential customer. So what I'm gonna talk about overall today is is our kind of business processes. So we're all gonna look at a slightly different angle in in the episode. but business processes is gonna be a big one because that's where we start to save time and we can reinvest that time into more kind of valuable things like maybe it is our our marketing or kind of messaging to make sure we optimize that, or maybe we haven't done paid ads before. And now we want to look into that. So there's lots of ways we can reinvest that time. Okay. So yeah, it'd be great. And Sinéad Walsh: Aisen, posen. Denis Jastrzebski: yeah, we we want to look at I suppose some more details into everything that that we've kind of mentioned so far. So I I was wondering, Elish, is there any kind of particular process in in in what you mentioned earlier that that you'd like to to focus on? What what do you think might be a kind of a nice entry level point into that whole process that that you want to talk about? Elish Bul-Godley: Yeah, so and back to the lesson I just did yesterday. I guess the the process was that and I also did a a webinar once called Beyond AI Slop, right? So so so the I think for me, obviously having a content background, I'm really tuned in to AI Slop and from a so not just from a professional perspective, but very sensitive to audiences AI fatigue and stuff like that. So So a process that I do drill in a lot when it comes to creative work is that making sure that people inject and train the model first with their brand. That making sure, and of course we can talk about later how you do that, but that making sure that before you create something for the public domain that's audience facing, that you do build in into that creation process, their content marketing process. the training with the with your brand style guide, with your brand voice, with your brand and then of course coming out at the end and just making sure that your logo is in the right place. Cause even when I did infographic like we talked about, I had to actually go back and tweak the logo because it wasn't positioned properly. So there's always that little bit at the end, even not just with the as Sinead said, with the data, but also with you know the with the visuals, right? And which brought me and just while I'm talking about process, like I was thinking about that chatbot that you just said, that's a content creation space that you're in, right? I mean, there are people's jobs out there that are copywriting for chatbots, you know? So if you were a copywriter and I've seen a lot of stuff out there, but copywriters in the content space munching their jobs are still there, like you know, you still have to copyright for the chat bot. If your brand manager or if you're the business owner in that process with like just creating that chatbot alone. which is audience facing, you still need to look at the language the chatbot's using, to moderate how it's responding and and you know, look at the quote when it comes out. So I think there there is like even Claude the other day, Anthropic actually, Anthropic had a job out there for 400K for a copywriter to work for Anthropic. So so to train their chatbot to yeah, that's like it was it made the news because it was like the highest paid copyrighted job on the planet at the moment, I think. Yeah. So yeah, so like just saying that there is processes, but there is also your creativity and your voice that has to come in. Even Anthropic is playing is paying somebody for voice. You know what I'm saying? So Denis Jastrzebski: Yeah. Elish Bul-Godley: yeah, so I think that's my take on the the the creative process and AI. Denis Jastrzebski: Yeah, no, definitely. And and that's it gives us a kind of a a really good I suppose we'll call it a hint, but it's a pretty obvious hint. If if massive all if all the massive AI companies are are investing in into kind of traditional roles, sure I'm we Elish Bul-Godley: Yeah. Denis Jastrzebski: we're like confident they're using AI internally in those roles to make them more efficient and

all that stuff, but you know, even these AI companies, what they're focusing on is is the people behind you know, the product. So the product is one of these chat bots like Claude or Gemini or ChatGPT. But at the core, we need human creativity all aspects of a business, even one that supposedly is completely AI driven. So that's a really good point. I really, really like what you mentioned there. And we'll we'll definitely come back to it. So Sinead, how about you? What what what's one of those things that you've mentioned that's kind of an easy entry point into maybe doing a bit of analysis with with some data using Gemini or Chat GPT. How do you go about doing that usually? Sinéad Walsh: Yeah, it's a it's a great question. So I think the first thing I would do, even if you are not a data whiz or a data expert, right? First of all, when you download your data, let's just say we're talking about a meta campaign, okay? or a series of campaigns, and you want to get a monthly report, you download that from a CSV, and then the first thing I always do is spot check, right? So d and do a data cleanse. So what I mean by that is And generally, to be fair, something like Meta or Google or whatever will give you good quality data. You're not going to be missing a lot of rows, columns, headers, and that kind of thing. But it's always a good idea to just check. Also, formatting in a CSV can go a bit wonky. So sometimes things like dates will come up weird and and that kind of thing. So it's always a good idea to do a quick scan. and then I would also encourage anyone to understand exactly what each header means, right? So so like your headers and your columns and and things like that. Okay. So and and also that your date range is accurate for the thing that you want to report on. Now, for me personally, sometimes what I'll do is to like get it just an idea of whether the data is okay is I will make a quick pivot table or I will like do a chart or whatever. But I know not everyone is going to be doing that in Excel and that's 100% fine. But I will do that because if a quick chart or a pivot table ends up, excuse my French, arseways, I know there's something, you know, there's a fly in the ointment, shall we say? So but yeah anyway, a quick kind of visual scan at least would be the first imperative. Then Now, like I said, I use Gemini a lot, right? And I have used Gemini for, I was just doing a quick analysis on it earlier. I think it's like probably just over two years now. So when it's gone from like Googlebard to Gemini and what it's become ever since. So I will upload it. Now your prompt here is key, right? So we know this, we talk about it a lot in our in our in our classes and stuff. So it's a good idea to you know, give it give it a role, right? So I want you to act as a digital marketing data analysis expert, whatever, like invent the job title. I don't care. Give it a focus role. The next is to like give it a an action verb. Okay. So what do you want

it to do? I want you to analyze this data and give me, you know, a breakdown of opportunities and things that went well or whatever it might be in a bullet point list or in a chart form or whatever it might be. So be very, very clear on what you want that to do. and then because we're talking about data, we don't necessarily need to worry about things like tone the way we would for social media captions as an example, which I'll get to later. But yeah, be very clear about the output that you expect from the tool. And sometimes what I do as well is I will put in the defined or I will put in, I will define for the tool what audience I am going to be presenting this data to. So if it was you, Dennis, for example, if I was presenting this data to you, I would say. person or the client is well versed in data, so don't be afraid to use X and Y terminology and and that kind of thing. But if I was doing it to a client that maybe wasn't so OFA with data, I will ask it to give it give me those points in kind of more simplified language that's going to resonate with the client or the audience or whatever it might be. So yeah so what what I will do as well is just to say the I I don't trust AI to do everything and just nail it all, right? So similarly, if you asked AI, I want to buy a pair of Nike Air Macs, what are the pros and cons? because I've got, I don't know, flat feet, right? What are the pros and cons? I will go and verify the claims that it makes. I will similarly do that in data. So if it says this campaign did not work or this campaign was excellent, I will go and spot check that in the data again. So I spot check it first and then when it makes claims, I go back and check it again. and so what I try and do is I use the likes of Gemini as like a metal detector, right? So it's gonna cover a lot of lot of ground for me and process a lot of data, but I want it to be able to point out the kind of pit the the strengths and and weaknesses of the campaign and where it kind of goes beep, beep, beep. I myself will delve into that a little bit more just to verify that it's not making mistakes and it is doing it correctly. So yeah, so As we always say, AI should be your co pilot, not your captain. And when it comes to data, that's kind of how I approach it. Denis Jastrzebski: Yeah, perfect. and it's it's a great approach and it it to be honest is it's almost the only approach. because the next thing we were going to talk about and move on to was was that kind of trust aspect. You know, when it comes to using AI in any of the the ways and methods we've talked about so far, you know, can we trust it to what extent can we trust it? So what I heard from you, Sinead, is even though you're getting it to do lots of this analysis, you obviously understand the structure of that data first of all, you understand what's there. You obviously have experience in dealing with that data without AI. So you can't really if if you're somebody that that doesn't know how to to

interpret that data in the first place, you're not really going to be able to verify if if AI has done a good job or not. Right. And Elish, same with you, you were saying about obviously that copywriting role that's come up. chatbots have to have good good content, they have to have a brand persona and all that stuff. Any content that we produce has to line up with our brand, like you mentioned the client's brand there. So from your perspective, Elish, when we're talking about the the kind of risk of of Elish Bul-Godley: Yeah. Denis Jastrzebski: using AI and those things, what's what's the the biggest thing that you think people miss? because it seems like even though you're doing a lot of creative work and there's quite quite a lot of moving parts in in that whole process, you're checking over things constantly within that process. So you're like making sure that the AI is doing what what it should and and you're verifying that. So what what what's kind of the approach behind that or what's the thinking behind doing that? Elish Bul-Godley: Yeah, and I think there are two sides of the coin here. We brought up and I brought up trust at the very beginning of our podcast series, and and you know, it's coming up now. So the other side of the trust coin is risk, right? So if you don't enable that trust, what you're left with is a lot of risk, in that people don't believe it's you, they don't believe it's real, they don't believe you gave a toss, like you didn't try hard enough to persuade them and you're they know you're using AI, the the mention of AI slop and AI fatigue, AI exhaustion, which is also the risk on an employee level as well. And so there are if you were thinking in like a governance terms and you did a risk assessment, right? Like they do in board meetings, like you you'd have to break your your content into parts, right? So so where are pieces of content that are higher risk in terms of We can't afford to sound like AI here because it's a death knell for our brand. And where are the places where, yeah, yeah, let's throw the AI copywriter in the mix? It's not going to make a big deal. And like if it's a training manual, if it's a product manual, I'd say your risk level goes down, right? From a content perspective. Because people just want that technical information. They don't need to be persuaded. They don't need any emotion in your employee manual, right? Right, Sinead. Remember, Sinead said, emotion Sinéad Walsh: Yeah. Elish Bul-Godley: sell. So if You're trying to sell, you're trying to persuade, you're trying to have a brand activations, you're trying to move people, then the risk starts to kick in because that's your four your future your audience facing stuff, whereby the risk is high if you don't sound human, you know? So you gotta choose your battles here with the AI. You can't like want to be a copywriter when you're writing the employment. That's the AI is much better at you at being clearer at these things. Like it's better

at code, it's better at all of that stuff, right? it's better at structure. so for me as a content creator, right? I use the AI in a low risk situation. So when you took visuals, and I talked about visuals in my presentation decks, the low risk scenario was forming a table. AI can do that for me, you know. And the low risk situation was creating an infographic, their icons. I can program those in. But the high-risk situation is like if I wanted to move somebody. with a big image, you know, I need to pick the right stock image or the right piece of photography. I don't trust AI to do that. I I will pick that slide. You know, just if you're thinking like a slide deck already, right? So like I I you just gotta if you want to do proofreading, that's great. When it's copywriting, let's talk about structure again. It's not visual, but it's still important. If you want to do advertising taglines, you know, like you wanna be a bit more, you know human for that one, right? So like stress testing, AI is great at that. Structure, AI is great at that. But you know, when it comes to storytelling, that's a whole other ball game. You're talking about human beings and lived experience. You're the only one with that experience. Storytelling eats content for breakfast as we all know. So don't rely on the AI when you're moving into that space with your work. So your work can be divided up. You know there are places that have less risk. Go on, have fun with your AI there. But like, yeah, be a little bit more process led, like you say, Dennis, when you're doing the creative front facing stuff. Yeah. Denis Jastrzebski: Yeah, perfect. No, and and that's it. I I love that term stress test because we could really apply it to absolutely anything that we do with AI. So it it should be able to stand up to whatever test we we kinda have in mind. And we ideally want to have that test in in mind before we do the use AI for the task, don't we? So yeah, I think with content that's really important. Sinead, you've already mentioned about data, how important that is. and if we just talk about I suppose the the other risks, if we're talking about a chatbot in this case. now we have laws that are coming in about AI. So the EU AI Act is coming in. it's not something we're gonna dive into today, but we have an an episode, the next episode coming up is gonna be all about AI governance, so that'll be quite interesting. But I just want to touch on one part of it, which is which is kind of to do with the stress testing side of things. So we're we're supposed to kind of assess risks for the different tools we might use within our business. So if you're regardless of the business that you are. You know, if we use AI for things like recruitment, there's a bunch of stuff that we have to to be aware of there because we can introduce some bias into the recruitment process. And AI is kind of a little bit ambiguous in in what decisions it makes sometimes. So we have to have a really kind of robust set of guidelines and guardrails for it. So we have to once again think of of the bigger picture, think ahead a little bit. and with AI bots, for example, just in August. the the new phase of the EU AI Act came in. And now if you have a chatbot living on your website and it interacts with customers, people have to know that it's a chatbot. So some businesses or organizations wouldn't have really maybe thought about that, and it's just kind of there. If if you're not really familiar with AI or how it works, you might think you're talking to a person or a customer service rep. Some organizations were kind of in that gray area, obviously trying to hide the fact. They're using AI a little bit, so they'd have, let's say, a profile picture of a customer support agent that doesn't actually exist while the person is talking to an AI. So with the new phase of the AI Act, that's not allowed. So now everyone has to know you're talking to an AI chatbot. So a lot of people can be caught out with that. So it's kind of similar to when GDPR came in. And there's there's a massive amount of information, legislation, it's hard to kind of wrap your head around everything. Not everything is relevant to every business. but it's one of those things that a lot of people got cut out by because it's it's kind of difficult to understand. and the EU AI Act is a little bit similar. Now thankfully they're giving us the time to kind of phase in different requirements, and none of them are particularly difficult to actually implement. it's kind of a a little bit more common sense than anything else. you shouldn't hide the fact that you're using AI when you're talking to somebody. It kind of seemed like common sense to a lot of people. but that's something Sinéad Walsh: Yeah. Denis Jastrzebski: we will touch on in the next episode, so it'll be really interesting. But there is risks that inherently come with using AI. So we have to kind of figure out if there's a trade-off of the time it takes to maybe assess the risks to actually use AI in any sort of process. So a big thing we we've kind of seen over time is that people and companies invest in AI and sometimes in quite a lot of money and quite a lot of time, but without kind of properly planning out. if the AI that they're implementing is actually going have like a tangible result for them. So you could spend all day creating a agent, an AI agent that that it you know looks after a whole content creation process. But at the very end of that process you have to spend four hours editing all the content that it created. So you know you kind of think did was that actually useful in the first place? So I was wondering, Sinead, have you seen that in in any of the the kind of projects that you've worked on, whether it's social media or paid ads? are are clients or different people that you see online, are they using AI in a way that's potentially actually making their life more complicated and not really saving time? Is is that

something you've seen? Sinéad Walsh: Yeah, I think like I've definitely seen maybe not so much my clients, thank God, but companies and like large companies as well, particularly on social media, excuse me, large companies on social media trotting out content or captions that Elish Bul-Godley: Mm. Sinéad Walsh: have not been spot checked or not been built in such a way that they are true to the the brand voice to kind of touch on what Elish was talking about as well. And as a result, what is happening is people in the comments, as people do on social media, light up. They have, you know, they it opens up a whole different kind of branch of conversation. And then it becomes reputation protection and management for the brand or for the business or whatever it might be. now, luckily in this day and age, you know, you only need to wait kind of 60 seconds before another brand disaster kind of comes along or whatever. But I've definitely seen things like that go awry so much. That I'm like, well, you know, if you had just got John to sit down and write the caption himself, I think you you would have saved yourself a lot of heartache here, you know. So yes, absolutely I agree with you. And like the other thing I will say as well is we are so we come across SME owners every day that are so hellbent on using AI to plug in the gaps that they have in terms of their knowledge or their skills, which is something I was mentioning on I think it was the last episode. Where actually you need to do the evaluation and and kind of understand like you upskilling versus you lumping it all on black in terms of, you know, getting AI to try and do this for you or try and do that for you and then manage the outcome of that. Like what is the time saving there, you know? So I think it's very, very easy to believe that AI is that kind of magic wand that you're gonna wave and and solve all the pro you you're gonna have a 20 hour work week in the blink of an eye. That is absolutely not the case. I've been on the receiving end of reputation management over the years of my career in in comms and marketing and stuff like that. Trust me now, when you are firefighting, it's gonna take up a lot of your time. So yeah, I would say it it happens. It's you know, brands misstep, they don't pay enough attention, they don't, you know, put in the work at source to ensure that what comes out of the AI tool or whatever is true and accurate, et cetera. So yeah, proceed with caution. Denis Jastrzebski: absolutely. You know, really, really good points there and and I saw Elish, you were just nodding the whole Elish Bul-Godley: Yeah. Denis Jastrzebski: time. You you completely agree with Sinead. I'm Sinéad Walsh: Yeah. Denis Jastrzebski: curious what what's happening in in your head around that? What what do you think? Do you find that, do you see that around you? Elish Bul-Godley: Well I d you know, I I thank you, Sinead. You just helped me create another tagline because my new tagline I'm just

invented right now, guys, is Sinéad Walsh: Ha ha. Elish Bul-Godley: it it takes time to save time when you're using AI. Denis Jastrzebski: There you We have we have to trademark that. We have to trademark that right now. Sinéad Walsh: That's great. Elish Bul-Godley: Trademark it now. Go talk to the to to the IP people. Okay, so yeah, because I just had this experience. Yes, do my lesson that so you you know, so and we like it was hilarious because like the the title of the lesson was your 20 minutes social media 20 minutes a day is all you need, right? Just like such a misleading hook. And like Sinéad Walsh: Yeah. Denis Jastrzebski: Yeah. Elish Bul-Godley: we did it. I'm sorry. And then the first thing I did in the lesson was like, I'm sorry, I totally lied with the title. So basically what I What I showed them, right? What I showed them basically in the lesson is yes, you can, you can do your social media 20 minutes a day if you had spent two hours at the start of the week planning your content. So that was the caveat, right? But back to AI, right? So yes, you can do stuff really quickly if you had taken the trouble to train the AI properly in the beginning. Sinéad Walsh: Yeah. Elish Bul-Godley: Otherwise, you're just gonna end up with slop that looks like everybody else's slop, and then it's just there's a and the other thing is graphic design. So the reason I was nodding, because I'm I'm part of this, like my my algo in LinkedIn has thrown up this thing, which I totally am involved with right now, and this all of these threads are coming out from graphic designers. The biggest bugbear I have right now is posters made with AI, right? I just I see you guys nodding. I see you guys nodding. I swear to God, I swear. Denis Jastrzebski: That's my biggest crap as well. It's it hurts. It hurts. Sinéad Walsh: Yeah, same. Elish Bul-Godley: I catch any one of you doing this and c I will personally come over and slap you myself, right? So posters made with AI, right? They're great when they're in for graphics, right? And they're like they're on a wall. But but if they're posters for marketing, do not let the AI do that because it's going to not follow the rules of graphic design, rules like visual hierarchy, negative space. AI has no sense of negative space. AI could never have invented Apple's advertising campaigns because that took a human being's sense of negative space to create that Apple branding, right? So that that that's what I'm talking about, negative space. So like that sense of like the title font versus the by the and then the placement. There's like so many things on the poster, tiny little infographics. Like I'm just having a rant right now, guys. Just rant warning. Okay. So so like that's my biggest part. Yeah. Sinéad Walsh: So unlike you. Elish Bul-Godley: Yes, so unlikely. So that so it's all coming out now in episode four. so yeah, so basically, right, if you wanted to get away from that AI, yes, you could get the AI to do a really nice

poster, but you would have to train it for I'd say a solid hour. like a solid hour tweaking that design with AI. I tell you, no lie, folks. If I you let me loose on Canva with five minutes, I would have given you a better template. Right. So the human stuff that's available. There are AI led tools like Canva, for example, whereby there's beautifully made templates built by graphic designers that you can tweak and pull and takes you five minutes. But like if you try to do this from scratch with AI, you're actually wasting time. You're actually using way more time than if you just use the human approach. Right. So so yes, AI has its place, but like not everywhere. Right. So that that would be my thing of like it takes time to save time. Do you have that time to save time with AI? Or do you just wanna learn how to get it done right the first time? You know? That's my take. Sorry. All right, well that's over. Yes. Sinéad Walsh: Yeah. D Denis Jastrzebski: Yeah, no, no, you're absolutely yeah Yeah but yeah Sinéad Walsh: Ha ha ha ha. Denis Jastrzebski: it's it's great points. everything that you've made that you said there makes complete sense. and I Elish Bul-Godley: Yeah. Denis Jastrzebski: think we're obviously we're we're we both agree. so it's it's one of those things that you know it seems like one of the most basic things we can use AI for is to create images, let's say, for social media, but ironically it's one of the worst ways you can use AI for for marketing in in that particular area. So there's a lot of like basic things that we we can use AI for. And it's usually the boring stuff, which is Elish Bul-Godley: Yes. Denis Jastrzebski: the stuff that yeah, we don't really want to do, but we're kind of more excited about using AI in like graphics and creative things because it's like it's visual, it's more engaging. But usually AI tends to to do a better job in in things that are boring, right? So I I was just thinking in in terms of business processes and and automating business processes. A lot of companies invest a ton of money into like a massive AI driven platform, maybe internally. maybe it's like a CRM, they're changing everything just to have AI in there. But it's kind of overly complicated. most of the time, the biggest time savings that you get is from automating the really basic tasks. and usually in in the kind of three areas we see is customer support, being the first one, things like an AI chatbot on your website. then we've got marketing, because there's lots of things like data that Sinead's been talking about we can automate, we can kind of help help ourselves to understand things and speed up our workflows. and then the other one just being admin heavy tasks. So AI works really well when it has to do something that it that we repeat. It's kind of like a a pretty set process. It's it's got its confines and it's just kind of doing that or we have to do that every every week or every day. So AI does a really good job in in doing that. And it can't really be creative, which is the point that you're making Elish as well. If you look at those posters that that we're talking about and We might include one or create one in the show notes just so you see what we're talking about. They look kind of first glance, they look interesting. That's like a a really detailed graphic design. But when you look further from a marketing point of view, we're losing so much of the message that we're trying to put across. There's usually just a lot more information on that poster and nothing really sets itself apart. We don't really know what the general theme of the poster is because there's so much information. So AI doesn't really understand that part. of of creativity. And and generally as as a principle, all AI, it's it's designed with the this idea that it'll average out anything that you ask it to do. So it works on averages and it just in the background it's like mathematically that that's the the way it's designed and you can't really change that. But in terms of it its output, that's the best way I can describe it. Is is it it makes an average of of anything. So if you ask it to generate a social media report on a particular client or some data, it's going to use the average of all its knowledge to come up with kind of a standard format. And same with posters. That's why all the posters end up looking pretty much the same, even though they have different images, different colors, different fonts, but they have this AI feel to them, which if you if you look at five of them side by side you'll realize what the AI feel is and it's that there's it's cluttered and it's not really driven by kind of specific things that stand out to you. So we kind of see some things that AI is good at, some things it's not. So we've kind of identified creativity. It can't really be creative. It can be really good at structured things, like you said, Elish, and even writing content or even writing content outlines if you give it the proper instructions and if you give it a brand voice. And Sinead, obviously, data and spreadsheets, they're all really structured, so it makes sense. It does a really good job in analyzing all of that. So We've kind of looked at a few more examples. I'm I'm wondering, is there any other examples, Sinead, that you have of let's say a different sort of a process that you've used AI for? Whether it is kind of to do with paid ads or or data, are there any other processes that you think, that this actually saves or could potentially save an SME owner an hour or two hours a week? Sinéad Walsh: Yeah, definitely. One thing I I really and again, this is because I'm using Google Gemini, but you mentioned paid ads there. So keyword research, I think, is one of those things that business owners are mad for in terms of so I here's where I think this comes from. Okay. So you're a small business owner, it's kind of just you, right? You're the one setting up your business, running your

business, getting your website going, all that kind of stuff. You know that SEO, you're not totally sure what it is, but you know it's free. So that's helpful, right? and it can help you appear on Google, so that's great. And then people get kind of obsessed with this keyword research idea and all that kind of stuff. But like they're not always totally sure how to go about it, right? And then the other thing is if people do want to run Google Ads, like a little further down the line, they want to invest in in the kind of digital marketing aspect of their business and they decide that Google Ads is best for them, again. It's all totally based on keywords, but they're they're kind of not really sure. So, like hardware store Navan is your starting point, but how do you pull it apart? And also, as we know, how do you make it? there's keywords and there's search terms, and they're two very different ball games, right? So there's keywords that describe your business and and that kind of thing, and that you want to rank for, but there's search terms that the human is going to come along and be like, you know, claw hammer navan near me, whatever, you know, things like that. So I think AI, and again, I'm not I'm not on a I'm not on a commission here for Gemini, but I do find it really, really useful. However, just again with that process, you need to prompt, you need to give examples of where your head is at in terms of keywords and search terms. You need to do the research if possible. So if you have something like Google Search Console already in your i installed on your site. Can you look at those reports? Can you see over the last thirty, sixty, ninety days what terms have brought people to your website and that kind of thing? And all this stuff takes time and history and you need your website going for, you know, a good amount of time to to take those learnings. But yeah, I do really think I I think business owners get obsessed with keyword research, but really aren't sure how to go about it. And I definitely think the LLMs are are very, very useful at that. Again, particularly Google Gemini, 'cause it's there. their system. But yeah, so that's that's my thoughts on another process. Denis Jastrzebski: That's that's a great point. And it kind of reinforces again if we're thinking of averages, keywords, and and what people search for. SEO is kind of about finding out or trying to realize what keywords people are using to find us. And regardless of kind of what country we go to or what area we're we're targeting, people tend to use certain keywords or or words or terminology to look for things. So once again it makes sense if AI is obviously really good at at assimilating data and it it has all the keywords, it knows how people naturally search for things as well. It's part part of its training data. Makes sense it'll it's it's a great brainstorming tool in a way, isn't it, for for keywords? Because there's there's terms and synonyms that we might not use on a daily basis that we just happen to to be ranking for and then we realise, actually that's something that we've never targeted before, but people actually use that. Or sometimes we're too close to the business and we think, yeah, this is exactly how people are searching for it. And I know all the technical terminology in my industry. But maybe people have no idea what that terminology is. And they're using different Sinéad Walsh: Absolutely. Denis Jastrzebski: words, right? So it makes sense. AI is really good at that. So that's a great point, with keyword research. how about you, Elish? Is there is there anything else that you've seen that that's useful? Elish Bul-Godley: terms of using AI in my Denis Jastrzebski: Yeah. Yeah, any other process or even another part of that whole process you mentioned at the beginning. Elish Bul-Godley: Yeah, I mean there it's actually many points in the process. So so I would definitely and they're different models, right? So I'm really interested in what Sinead said, by the way. you Sinead, I'm just gonna ask you a quick question. so the SEO side the SEO the search term research that you're doing, you Sinéad Walsh: Yeah. Elish Bul-Godley: you feel Gemini is well poised because it it's part of the algorithm already. Is that why you're Sinéad Walsh: yeah, no, that's just a theory I have. I don't have the only kind of thing I have to back that up is that I have done keyword research off the back of Gemini, Chat GPT, and Perplexity as well. And I have I have used the pro versions for all of those, that kind of thing. And I've always preferred Gemini's output. And the other thing that it can give me, but I I have only spot checked this a few times against their keyword planner. is like the forecast or the rough searches per month and and or search volume is the word I'm looking for there. and that kind of thing. So yes, where I have done and I will say limited research, but where I have done my research and my comparisons, Gemini has come out on top of those three models. Now I haven't used Claude and I know both of you guys are like pretty big fans of Claude. So I just I haven't waded into that yet because I feel a little bit saturated with AI tools at the moment. But I think I will wade into Claude 'cause I know like I respect you guys so much in terms of your knowledge of AI and I know you're both big fans of it. So I will get there eventually. Elish Bul-Godley: I'm yeah yeah, I'm still learning. But yeah, so back to my process. So what we can kick off where she Sinead left off, which is so say I've got to write a blog piece about a topic, like I don't know, there was a seminar and cybersecurity is coming up in Dublin, right? So so say I've got that job, right? so where would I begin, right? So back to Sinead, that's kind of the question I was asking her with the search terms. What are people asking about these things as a as a content writer? and you're gonna create a decent blog that gets traffic. So I

would Consider Uber suggests or Answer the Public. And I now I would go and ask it for that sort of research, that data that it has already crunched and repackaged into these third-party tools. So that you'd have something like, in Answer the Public, you'd see all the questions people ask around cybersecurity, for example. And then you might they divvy that up between social media queries. Google search queries, and now they've got AI chatbot queries as well. So questions that people ask AI. So that's a nice little AI tool that can kick off your creative process because it's pointing you to the most commonly asked questions. Remember, we said content writing is not for yourself, it's for other people, is to answer their question. So answer the what's in it from a question. What's the best way? You know, look at what the questions are, right? So that will kick off my creative process, and then I start thinking about angles. A lot of angles will come in through my head. I'd have to do a lot of personal reading and listening. Sometimes the creative process is just listening and reading and then stopping and going for a walk and all of a sudden there's flow. but then to help me with the flow, I might go to Claude. so Claude it looks at things and it's very structured in its approach when it comes at and what I like about Claude is like it it will tell you how it came at that problem. So you know, you know, why it's telling you what it's telling you. So I would say, Claude, I've got an idea for like this start to the essay, and I want to bring in these topics. And I like that sort of metaphor that's come into that creative metaphor I have in my head. How do I weave that creative metaphor? Have you given can you give me some approaches, some angles to bring that creative metaphor that intuitively I feel is going to work for that article? But let me see, how can I pick up certain strands and so that's that's the space I'm giving Claude with my conversation where I said me and Claude had a conversation earlier. That's that conversation we would have, right? So and then of course I'll start breaking down how I want to structure the the thing and then I might get that blocked into like, you know, ideas blocked into paragraphs, bullet points, things like that. Then I'll go off and I go free form. I go freeform with it and then I might get some one of the AI models to stress test it. Maybe is it answering any questions? Are there gaps in what I've just written? can you proofread my grammar? so you see all those low risk places. Yeah. But that lovely metaphor, that was me guys. You know, that was me going for a walk, Sinéad Walsh: Ha ha ha. Elish Bul-Godley: you know, watching the swans in the canal. So so yeah, so like so that's how I think I'm weaving it in at different phases using different tools. And of course, the brand, the brand voice, and and the styling. using a style block, then I go, can you, you know, this is the style of the company. They've given me some a lot of companies or clients we have which are larger in size, might already have a ready made brand style voice that they've listed. And what do we mean by that? We mean it's a it's like a list of things you you can say, you can't you can say. Words that we like to use, words we don't like to use, you know, prepositions we prefer or pronouns we prefer, things like that. So there's this like a structured parameters around how that that essay should sound, like them. and then of course I would make sure that's working towards that as well. But I've got the main idea, I've got the main polemic, the main argument. Polemic is usually something I come up with. Polemic is a term for making an argument. And that usually comes from inside me, but sometimes you can get help with that. But you know, the the the persuasion in the language, that's something you bring to it, and that's something the brand brings to you. So if you're a small business owner, you may not have a fully fledged brand style guide or voice lined up in a docking, but that voice is you, that's literally you when you open your mouth. So, like, so make sure that when you're rewriting that that essay out or that article, that piece of blog, that you're you're writing it out as yourself. I have a really good tip for copywriters and for people. learning how to copyright for their businesses. if you want to say something that doesn't sound like AI, take the thing that you're given and pretend that you're explaining it to your friend in the letter or in a text or whatever. Just like this is a little practice, right? So a friend of mine said, I'm writing emails. She works for for Big Bank and it's like she's so sick of people sending her emails that sound like they're written by AI because they have too many words in them and like they're repeating, you know that That AI thing where it's just like it's like this long when really the message is really here, right? Sinéad Walsh: Yeah. Elish Bul-Godley: So, so like I said to her, just I don't want she said, I don't want to sound like that when I'm writing. I said, Yeah, well, just pretend you're explaining it to me, like you're writing me a note, and you're trying, and you will never waste your time and pretend the bus is coming, right? So like you're standing at the bus stop Sinéad Walsh: Yeah. Elish Bul-Godley: and you're trying to explain it Sinéad Walsh: Mm. Elish Bul-Godley: to me, and the bus is like a hundred meters away. You're not gonna waste your time, you're gonna get your point across and Everybody who's on the other side of that email will thank you for it, right? Because you didn't waste my time. So like that's the problem with writing. I'm just we talked a lot about design and graphics. I like from the contents at the writing side. That is a good way to think about your process. You know, use the AI for the research, use it for the structure, and then bring the bring your your argument, your voice into it, you know.

Denis Jastrzebski: Mm-hmm. Yeah, excellent, excellent. No, th those are really, really good points. And I like the the copywriting is an important part Elish Bul-Godley: Yeah. Denis Jastrzebski: of of marketing because yeah, we did talk a a lot about visuals and you know kind of maintaining personality, but it's true, it at the end of the day, it's all about your writing and and how your personality or brand personality comes across. So in the previous episode, if anyone is looking at the show notes, we do have a prompt for for kind of helping AI learn your particular writing style personality. So you can use that in in combination with Elish's approach. of writing a a letter. I think that would be really good. It's it's a n a nice exercise maybe to get a bit of practice writing without AI or using it to you know to a certain extent. So that's really, really good. And and I think just to summarize what we've talked about so far and and throughout the episode really mainly about the the different processes we can use AI in, we want to kind of be pretty meaningful about the processes we choose because not all of them are going to actually have a a positive result. So whether that means the the content itself, like maybe the AI posters, aren't gonna actually represent our brand well and is probably gonna put people off. That's obviously not a good way of using AI. But we found ways within the creative process and also within advertising and and analyzing data that that works really well. Using chatbots is is something that it's it's pretty useful from a from a customer service point of view, cuts down that time you have to spend. But all of those decisions were made, I suppose, before you actually started putting AI into any sort of those processes. You're trying to figure out, you know, what do I spend a lot of time on that AI can help us with. And the integral part is that we're always part of that process. So we're checking over things. We should have enough knowledge to verify that the quality of of what AI has produced is good. So those are some some kind of high level takeaways I suppose that we have from today's episode. And and what we do want to also give you is is as usual are kind of tool or recommendations. for using AI and in this case how can we kind of figure out what process we should prioritize and using AI for? What's the best way of doing that maybe from a content perspective? Sinead is gonna talk talk about data or maybe social media stuff as well. So we're gonna talk about those key takeaways. So maybe Sinead, what was your tool that you were going to recommend? Sinéad Walsh: Unsurprisingly, to anyone who's been listening to this episode so far, I would Denis Jastrzebski: Yeah. Sinéad Walsh: strongly recommend Gemini. And I would strongly recommend the paid version of Gemini. What they're getting better at is developing the different models, obviously, for different uses. And they've got some really, really great I suppose variations there, okay. And and kind of since LLMs have come to the mainstream, it's been like ChatGPT and Gemini, ChatGPT and Gemini, and then You know, one is always a little bit more ahead of the other. But I think what works for me is Gemini is integrated into my everyday work system. So it even when it is writing emails or it's it makes the auto suggest, it does pick up on Like for example, Dennis, when I'm emailing you from my Gmail account, it'll say like thanks D as opposed to thanks Dennis, because I call you D a lot, right? Or, you know, like my partner or whatever, it will like, you know, it'll it'll pick up that I'm emailing my partner back and and use their nickname or whatever, you know. So like things like that. So but from a data perspective, I love Gemini for the clarity that it has. It it I think it presents the output is always a lot more digestible for me. And obviously, of course, I play a role in that by telling it how I want it, but it's very action oriented. It's a good business tool, right? Like I think ChatGPT is kind of better at like writing assistant kind of stuff, more so and natural language processing, where I think Gemini is just better for business led output, in my opinion. So yeah, Gemini. and we were talking about social media captions and that kind of thing. And I was saying like I always use the example Even though I sit here every day and I use AI for my own stuff on Instagram, especially with comedy stuff, I will always, always, always write my own captions. And the reason for that is I feel like my friends or my followers will know from a mile off if I have used AI, you know. my sentence structure, the way I speak, like you guys know, we've worked together long enough now. So, you know, like my little buzzwords or my catchphrases or whatever. So, but I think if you are a business owner, I think Jasper is a fantastic tool. It's a Purpose built writing assistant. It costs a few quid, of course it does. But if you want consistency in terms of brand voice and and that kind of thing, I just don't think you can go wrong with it. It is an all thing and all dancing solution for everything from your email marketing content to your blogs to your social media captions. So they are the two tools I would recommend. And then of course I was gonna say Canva, but I'm actually gonna let Elish say Elish Bul-Godley: Mm-hmm. Sinéad Walsh: sing the praises of Canva because I trust that's where where you were going, but Correct me if I'm wrong. Denis Jastrzebski: Yeah, we've bet a tenor on it, so hopefully hopefully we're right. Go on. Elish, please. Your your recommendation. Sinéad Walsh: Yeah yeah. Elish Bul-Godley: Yeah, and just full disclosure, I'm not being paid by Canva. In fact, I'm the one paying Canva every month Sinéad Walsh: Yeah. Elish Bul-Godley: or every year. So Denis Jastrzebski: Yeah. Elish Bul-Godley: so yeah, so why am I paying for Canva? That's the question. Yeah. So yeah, so like I think you could make a

poster, you can make social media graphic on DALL-E, on Gemini, but you'll come out with that weird thing without the visual hierarchy and all that. So if you really want to use AI to make your graphics, if you really do, definitely start and Canvas are probably better place because it's got the design intelligence. It's been trained by all that templates that have been created by humans and it weaves your stuff into those templates really seamlessly. But the templates themselves are made. So, like if you really want to use AI for graphics, fair enough. But it's got lots of little AI tools for graphics where you can tweak things. So where I like using AI for graphic design, and yes, I do use AI for graphic design, is for the tweaks, right? So if I wanna take out your background. If I want to shift that flower from a rose to a daisy, that's a little literal. If I don't like the graphics that were there, or if I don't like the way you use that font, I'm gonna manipulate that font. There's a tool for that in Canva. So I can pluck something out that I don't like and put myself on my taste. People pay me for taste in fairness. That's my job, right? My job is to use my taste and get paid for it. So So that's basically, yeah. So if I want to re rebrand something like that's where the Canva tools allow me to do that. So that's where I'm using the AI. It's to actually do more of what I want and bring more of my taste into it. the other thing it's not so much a tool tool, it's a process, and we hinted at that. the paradigm is the unique value proposition. We brought that up before. So the styling block allows your unique. Your your to come through. So we'll probably put that prompt in there in our toolbox in in the website. But don't forget, in a unique value proposition, there's how I sound, that's how my competitor looks, and what my customer wants, right? So you've only attacked one thing with the style block, making it sound like you, but stress test it, that's the technique I would say. Stress test what you got against your competitor, you know. Stress test your blog or whatever it is against does it sound like my competitor? Can it do better than can it talk about things my competitor isn't talking about? That's the stress test. And then of course against my customer. This is persona number one. you know, Sinead talked about personas and targeting. So stress test that against persona number one, if this is written for her, does it supply her with the right lingo, with the right value points that she needs? And that's the three-way. stress test that you're doing the style block, competitor analysis, and then against your customer target persona. And that's a whole other day's job building that persona, which we'll probably talk about in another episode. There you go. Denis Jastrzebski: Definitely. Sinéad Walsh: I like to say that's a that's a different bottle of wine conversation. Elish Bul-Godley: Yes. Bottle of wine. Great for copywriting. Bottle of

wine I would actually recommend. Bottle of wine and copywriting come together. Denis Jastrzebski: Absolutely. Love that. Love that. No, that's that's fantastic, Elish. And you've you've made up made some really good points there. And I think it definitely is worth a whole other conversation because really when it comes to to AI, realistically most of us are going to be using it for marketing. Most business owners want to use it for marketing to some extent because that is it's quite a big role. You know, there's so many avenues of marketing and digital marketing that we have. It makes sense. It's tough to to know everything. So I think that'll be a great episode we could we can do in the future. So all fantastic points. And and what I wanted to touch on, once again, just reinforcing everything we've talked about, is one, as a principle, we don't want to use AI in places where we can't verify what that result is, if it's a good result, if it's a poor result, if it's accurate. So, first of all, whatever we're using AI in, we should understand first ourselves. We should be able to do that ourselves. once we start using it in in places where we have no expertise, no experience. and we think AI is just gonna do it for us, that's where you can start running into issues like graphic design, for example, or different business processes. So the business processes that I've talked about today, you know, kind of a little bit of about customer support. I mentioned just the admin role in itself, has a lot of these repetitive tasks that AI can help with. So what I'm gonna do is in the show notes, along with Elish's and and Sinead's examples and the prompts that we're gonna share, I'm gonna have a couple of templates linked. to a tool called zapier which I mentioned in in a previous episode around kind of business automation. And there are going to be a few basic admin templates that that automate a basic task. So it could be you know maybe even just assessing an email that comes into your inbox and categorizing it as a lead, as a promotional thing or or something else. Or or maybe it's actually just drafting an email. So along with that I'm gonna show include an example of of one of the projects we've done recently The same chatbot that we that I was talking about at the beginning that lives on a website from one of the clients that we recently worked with, that same chatbot is having a separate function in that business where it actually lives in the inbox, the Gmail inbox of the business. And any new website inquiries that come in, they're assessed by the AI with the knowledge of the website. And also that AI has the personal writing style of the owner. So it actually takes whatever comes in, whatever inquiry it is. It drafts a simple email and it pings the owner to to look over it. So that's cut down the amount of time he has to spend on just replying to generally the same questions people ask about the different services that he offers. And that's a massive thing. So that that's gone from him spending eleven plus hours a week doing that to only a couple of hours usually just editing a little bit of that email before it's sent. So I'm going to include some basic templates of that and we'll have a case study on that whole project. up on our website as well. So should be interesting. But that's Sinéad Walsh: Great. Denis Jastrzebski: today. That's everything for for today, folks. Really appreciate your time, Elish and Sinead. It's been a really interesting conversation. We have a lot of great resources that we're going to be able to share on the notes, on the episode notes. and yeah of course you'll be able to find the podcast on all the main podcasting platforms. and once again appreciate your time Elish and Sinead and we'll see everyone in the next episode. Sinéad Walsh: Thank you. Yeah, absolutely. Thanks, Dennis. Cheers. Denis Jastrzebski: Thanks everyone. Elish Bul-Godley: Thanks, mind yourselves, guys.