#453 From FMCG to AI: Prashant Agrawal on Rethinking Marketing in the Age of LLMs

[00:00:00] 
 

Mehmet: Hello, and welcome back to a new episode of the CTO Show with Mead today. I'm very pleased. Joining me, Prashant Agrawal Prashant. The way I love to do it is I keep it to my guests to introduce themselves and tell us a little more about, you know, their journey, their background, [00:01:00] and what they are currently up to.

Mehmet: But just a hint for the audience and kind of a teaser, we're gonna talk about something, you know, cool, interesting. Which is, you know, AI marketing and everything in between and what Prashant is currently doing in that space. So Prashant, the floor is yours. 

Prashant: Thank you. Thank you so much Mehmet, and thank you so much for having me.

Prashant: Uh, yeah, so it's, it's been quite an evolution for me. Uh, I started my career in 2000 after my MBA in marketing. Joined, uh, uh, joined an Indian FMCG multinational as a brand manager first, and then moved to Dubai in 2004. I. Over the next two decades, uh, you know, spend time building brands across 40 plus countries.

Prashant: Uh, you know, this was, if you may call it the classic FMCG playbook time, right? So, tv, print, radio, in store, uh, eventually digital marketing and uh, e-commerce as well. Uh, so while all this was going on, [00:02:00] so this was while, uh, while this was going on, I was first as a CMO at, uh, Himalaya Herbals, and then I had my own, um, uh, startup where I, which I co-founded in Switzerland called Swiss Image.

Prashant: So this entire two decades period, and this launching and building brands over the 40 plus countries. Was what you might call the classic traditional FMCG playbook. Like I said, uh, during the period, if you see the ai, uh, part was really background noise, you know, we used to hear about AI and we used to get intrigued.

Prashant: Of course, uh, it was more in the space of, you know, Terminator coming in into our lives. Uh, we never thought really, uh, you know, that it's gonna be really in our lifetimes. Uh, anything important or relevant for business. Of course all of this changed with the, uh, with the chat GPT moment as it's called now in late 22, uh, when all of us suddenly said, Hey, what's going on here?

Prashant: And when I saw these, uh, when I saw the [00:03:00] chat GPD coming in and then the rest of the LLMs like everyone else, I started playing around with them more for fun. Really still not very sure of what really business value this could have, not have, but it was more out of intrigue, curiosity, more than anything else.

Prashant: But very soon I realized that this is, for me personally, becoming. Uh, I'm getting more immersed in it with the, you know, started taking, uh, bunch of courses and reading a bunch of fundamental books on ai. First, understanding the fundamentals of how the technology has evolved and developed and what is the technology about.

Prashant: And then eventually starting to get hands-on experience, uh, with these AI tools. And more than anything else, you know, trying to figure out that, okay, so this is the job that I. I've been doing over the last 20, 25 years as a marketer, and we keep talking about these tools that have come out and they can do this fantastic things, but are they really able to solve the problems that I have [00:04:00] encountered in my, in my business work?

Prashant: So it was more, I'm not a, I'm not a technologist trying to figure out what a tool does as an application. It's more that I'm a marketeer who's found technology to, you know, solve. Real life problems as I have encountered them. So that's really where my interest in ai, uh, came in as a business person, as a marketer, uh, more than anything else, 

Mehmet: you know, like a great introduction and, uh, a lot of thought provoking, uh, points you, you shared with us, uh, Prash today.

Mehmet: And of course I will try to deep dive as much as possible. Now, you mentioned about. When you saw, and I, I like the way you called it, the chat GPT moment. Of course it is the chat GPT moment. Indeed. Um, but you know, things have changed very fast. Like, uh, fast forward now, 2025. [00:05:00] Yep. It's been two years and a half almost since we saw the first, uh, you know, chat GPT coming in November, 2022.

Mehmet: To your point, um, so. From your perspective and being. A veteran marketer, I would say. What do you think the most common misconceptions business leaders have about, you know, implementing AI in their marketing functions? To the point, because, you know, you were talking about how at the beginning you were just discovering and, you know, seeing how, what can be done.

Mehmet: So if you can shed, uh, some light on this point. 

Prashant: Sure, sure. You know, what has happened also, mammoth, is that, um, it is the timing of this ai, uh, evolution or the start of the chat GPD moment and from there on has been quite, uh, serendipitous for me. You know, because this was also the period of time that I was actually transitioning from.

Prashant: Operational business, operational leadership to consulting and education. So I found myself with this bandwidth to be able to, you know, [00:06:00] consult, coach, train. I do teach also, uh, marketing at, uh, universities. I. As guest lecturer and as a visiting faculty. So I actually, when I started interacting with the business leaders, because I mean, these tools were really blowing my mind away and, you know, I was saying that, okay, hang on, you know, there's real value here and, and I need to make sure that, uh, my fraternity also understands this the way, you know, maybe they don't have the bandwidth right now to spend the kind of time and.

Prashant: Energy that I have put behind it in terms of understanding the fundamentals. But surely they must understand what, what's going on. So I started talking to these business leaders and market lead marketing leaders and et cetera, especially in this region. And what I found are there are these recurrent themes of misconceptions, uh, and I'm, I'd like to sort of share a few of them.

Prashant: I'm sure that you have your idea of many of you know, in your interactions. The first, uh, one that I would like to share is this whole idea of this [00:07:00] spectrum of, you know, two ends. So one end is this thought that it is completely overhyped. Mm-hmm. And, uh, the other end of that spectrum is this magic button.

Prashant: So. People are either at the space saying, this is all, this is all total, totally overhyped. And, and you know, nothing is gonna really, we've heard so many things happening and then really nothing changes or the other extreme where people really feel that AI is almost like a magic button, you know? So I'll press this magic button and suddenly I'll get exponential growth in my business.

Prashant: Now, any of us who's been involved with AI transformation and, and you know, helping out companies and professionals realize that it's neither of these two, it's actually really. Very thoughtful integration into your marketing workflows. I mean, the magic is there. The magic is the exponential benefits are possible, but it requires thoughtful integration into your marketing workflows.

Prashant: There is no magic button. Uh, so it, it, it needs to be really implemented properly. [00:08:00] So that's the first one that I would, I would like to put out there, which, which I think still exists. It's, it's reducing. But it still exists. The other one is that most people really feel that these AI tools and AI is all about cost cutting.

Prashant: So also we'll implement ai, we'll get cost benefits. Uh, yes, this is definitely about bringing cost efficiency, but I think we are missing the point here, especially in the marketing function, if we are really thinking of it only as a cost cutting, uh, tool. It, it is more about being able to improve dramatically your quality of work, uh, in the marketing space about your understanding of your consumers.

Prashant: Uh, dramatically. You can get exponential benefits. There are so many ideas that marketing people have, but they don't have the bandwidth to take all of them to a reasonable point to evaluate. Uh, this allows you that ability to go after many more ideas than what you [00:09:00] pre, previously could pursue, because to take it from, uh, zero to, uh, 60, 70% where you can make sense of an idea.

Prashant: It required earlier, too much effort. Now, with these tools, you can take it to that point where you can actually give it a serious evaluation before you, uh, decide what's really right for your business. So that's another one, you know, cost cutting versus, uh, value addition. The other, I find another one very, very interesting is that, uh, when I speak to some leaders, you know, you, so everybody follows media.

Prashant: I mean, you can't today go through a news day on any news website or a news channel or open your social media feed. Without having something on ai every day there is some news on ai. It's, that's the nature of the beast today. So you keep hearing these buzzwords, right? So sometimes it's ai, sometimes it's ml, sometimes it's a GI.

Prashant: Sometimes it's generative ai. Uh. So everybody keeps hearing these buzzwords, and that's what they're chasing. So I want to [00:10:00] implement a GI, I mean, we, we don't even understand, uh, as many of, uh, these people don't understand what, what these terms mean. Uh, but that's what they're chasing. And I would really, I, I, what I really recommend to my clients and friends and professionals who I interact with don't get swept by these buzzwords.

Prashant: Think about what your. Problems are, where are you finding friction in your existing workflow? Let the technology give you the benefit rather than the technology defining what you are gonna change. So that's, that's another one which I find that interesting. One is, uh, this whole fear about. Job loss, right?

Prashant: So all AI is gonna come in and replace all the managers. So there are no, not gonna be any managers. So I really don't know what's gonna happen in the next 10, 15, 20 years. I mean, it's very, very difficult for any of us to make any predictions. I can really come to this conclusion as I have in the last, uh, a few months, years, is [00:11:00] that what Rob Thomas said, uh, is, is probably the most accurate in this, is that AI is not going to replace managers.

Prashant: But managers who know how to use AI are definitely going to replace managers who don't. So I totally believe in this now that this is something which over the next two, three years, everybody will have to learn. Everybody will have to operate with because it's gonna become part of every single industry.

Prashant: It's gonna become a part of every single function, whether we like it or not. 

Mehmet: Uh, Prashant, what you mentioned resonates a lot, you know, with I, uh, what I see on day-to-day basis, what I hear from, from different people, of course. Um, and no surprise for me, you know, because you mentioned about, uh, people who also are not understanding exactly what's going on, which I can't blame them at the same time.

Mehmet: Sure. Because, because honestly, like even for me, as someone coming from a technology background. Uh, I put myself in the foot [00:12:00] of someone who is not from a technology background, right. And I try to grasp this amount of information and the amount of, you know, changes also as well That's happening. To your point, you know, the, the key thing which, uh, attracted me is about people trying to find use cases.

Mehmet: Rather than trying to see what are the real problems and see, okay, what are the, you know, possible solution for that problem today. Right. Which is, which is, I believe it's something, you know, it's like growing pains, we call it like in the world of startups because we're still in the early days, but early days are passing very fast also as well.

Mehmet: Yes. And, and you know, like no other time before, uh, Prashant, seriously, like I've been in tech. Almost all my career for 20 plus years now. Every day there's something new. Right. And in the marketing space, we are seeing like some [00:13:00] massive amount of tools coming every day. Right. So based on your experience, um, what are like really some of the tools or the platform that you think.

Mehmet: They are really effective in streamlining marketing operations or, you know, maybe they can be beneficial or maybe something you've seen recently. You use it recently and you think it has a huge potential also as well. 

Prashant: Yeah, you know, uh, this tools question is a very, very difficult one. There is almost every single day you are part of this industry, mammoth and then.

Prashant: Uh, you are a technologist. I'm a marketer. I, I mean, I'm trying very difficult, but in a, you know, very, very, uh, uh, I'm finding it very, very difficult to keep up with, uh, you know, with the tools that are coming out every day. But I must tell you, uh, you know, a, a a, a small, uh, side story on this is I was watching, uh, about a year back, almost, uh, you know, maybe about six months back, uh, uh, an interview of, uh, van Jones.

Prashant: Uh, you know, he is. Uh, US [00:14:00] Democratic Party strategist comes very often on CNN and, you know, uh, is, is interviewed a lot about politics. I'm, I'm not sure if some of your listeners have, have looked at him. Uh, so he, I was listening to an interview of him and he mentioned something very interesting. So he said, uh, you know, to take over the world, all you need now is Chad, GPT, and Mrs.

Prashant: G. And, and, uh, the interview said, okay, what is Mrs. G? So it says, midjourney, runway, SUNO and gamma. So, so with the, with the chat, GPD and Mrs. G, you can, uh, really take over the world and I found that quite interest. Interesting. And I, and I use it a lot now in my conversations where I say that as a marketing person, uh, I find that I just expanded that Mrs.

Prashant: GA little bit to Mrs. Peg, where I say Midjourney runway sun. Perplexity, 11 Labs and Gamma, besides Chad, GPT or Claude or one of the LLMs. And, and this, uh, everybody will have their own preferences of tools. I always feel, but I mean, to [00:15:00] me, these are the tools which I use most often. And, and I found the maximum value out of, uh, just a disclaimer, I don't get any commercial, uh, benefit by promoting any of these names.

Prashant: So, so they're just what I use. I, I'm not commercially associated with any of them, except as a user. 

Mehmet: Yeah, definitely. And to your point, uh, yeah, no one is affiliate. But see, look, like, I think, uh, these AI tools became same thing like when I remember when the first, uh, you know, new generation of, of browsers started to appear, Hey, go use Chrome, go use, right?

Mehmet: So it's the same thing, and I think it'll be personal preference, maybe at the end, which AI platform or tool they're gonna use. Um, same thing happened also when, you know, the smartphone started to appear like, Hey, go use Android, go use iOS, and so, so, so on and so forth. Uh, so definitely, [00:16:00] and I like this, you know, like, but, but by the way that I think the list gonna keep growing.

Mehmet: Yeah. Uh, with time. Yeah. I'm not seeing any soon, you know, we gonna have some consolidation because now it became. It became kind of a mandate, I think, on, on these companies because back to basics, and this is maybe audience who are familiar with like how people go from the zero to one thing and people who want to grab the market as much as possible.

Mehmet: So you need to keep your audience with you as much as possible. So social media folks, and maybe Prashant, you know it's better than me. They try to keep you within the application. They will put anything. Algorithm to keep you there and they will even demote you sharing something that would take the audience outside.

Mehmet: So I think the AI tools and platforms they are trying to, to play this. I'm not sure what some of these AI tools would do. I mean, the ones which are [00:17:00] like little bit old in the market, which by the way, they were before chat GPT, they were using open ai API. Funny enough. Yeah. Um, like the Jasper of the world and the writers of the world.

Mehmet: Right. So I was lucky to, to, you know, I felt that something is coming when I start to see the Jasper and you know, these AI tools. And still people didn't, I know, I knew of course about open AI and I knew somehow. They wrapping the, the API over there, but yeah, the tools are, are going more and more, but now, and I, I'm just, you know, like, uh, adding something, maybe your, your point of view.

Mehmet: So Prashant, when it comes to something which are the advanced features, we are looking into these tools. I want to stop on the two little bit. Sure. So we're starting to see something which never been there before. Let's take, again, we're not affiliated, neither you or me. Let's take for example, perplexity, right?

Mehmet: Yeah. [00:18:00] So perplexity has a deep search function, which is, yes. It's not, they didn't invent it, but I'm talking about it's, it's quite good. Yes, absolutely. Same thing. The Elon Musk one. Now from a market perspective like this deep search. And the thinking capabilities, what it means for me as a marketeer. 

Prashant: Yeah.

Prashant: So, you know, it, I, I'll tell you, it's a fantastic one because, see, what earlier was happening was if you were using LLMs, you were essentially, uh, you know, what you were not able to rely on while it was being able to, you know, give you language. It was not being able to think deeply on a particular subject.

Prashant: Right? So what the, what the game has changed probably in the last. Probably as recent as three to six months if you, I mean, you are the technologist, so you'll know probably this more better than me. But what I can see as a user is, is that in the last three months or so, or four months or so, the game has [00:19:00] changed, uh, a lot towards bringing in the thinking feature into these.

Prashant: And that's essentially, uh, what that, what's, uh, what's happening because of that is that if I have a complex marketing issue. Earlier what would happen is that I'd start working with the LLM and what I would be able to rely on the LLM for was to construct language around a particular idea rather than think deeply and.

Prashant: Almost like a thought partner battled with me in terms of the, of the, uh, you know, the process. What is the, what is the reasoning behind a particular approach or not? And we were, uh, I mean many of us were really using, uh, you know, the prompt engineering, uh, like chain of thought reasoning pattern. To, to, you know, sort of get the LLM actually to think a little bit more before responding, but with these reasoning models which have come out and, and, and you are so right about this, uh, the models, I mean, look at what's happened just in 2025.

Prashant: I mean, we started the [00:20:00] year with this big bang of deep seek, and then you've grok three, and then you plot 3.7. And you know, I mean, it's, it's extremely difficult now. It's almost like a race happening every week. Uh, or for a new model from one of the guys to come out and, and then that's better than the other one.

Prashant: So the way I, as a marketer use this is essentially when I have a marketing problem and I have a thinking issue, I actually, I. Create a prompt and I put it across on all the three or four main LLMs that I use. I see the kind of responses, and I do this a few times. Okay? So not, not just with one prompt. I do this maybe two, three times with uh, uh, with different prompts.

Prashant: I see the kind of output I'm getting. If I'm finding one particular LLM giving me better reasoning, better thinking process on a particular topic, then I actually end up using that for that particular, uh, marketing problem than the other one. So this is probably how we are making the most sense of it as a marketing user.

Prashant: Uh, you know, [00:21:00] in terms of, in terms of the improvement on these, you spoke about perplexity. It's amazing, you know, uh, today if, if you want to make a very detailed reporting, what would happen earlier is if I had to do research. Typical start point would be Google, right? I mean, you Google something and you start getting some data and inputs in and, and then you had to click on these different, uh, uh, different links.

Prashant: Uh, and then now you can actually pretty much from Xero. So I had a very interesting case, uh, which I was discussing with somebody in the US where they were looking at, uh, expanding retail. Uh, they, they have a business in, uh, automotive, uh, repair and maintenance, and they're looking at locations, uh, you know, how to decide on locations on this.

Prashant: And this was a friend actually, so I was almost, you know, sort of trying to explain to this person how to use AI and, and this, and I have absolutely no understanding of the automotive maintenance retail store market in the [00:22:00] us. And we start sat together on perplexity and we started out from zeros building slowly the thing.

Prashant: And by the end of it, this person who is from that industry, they said, it's amazing, you know, I'm in this industry sitting to do this kind of work and you, and I'm doing this alongside a person. Who actually has not much of an understanding of that, uh, region's or business. But because I understand how to prompt the ai, we in, in few minutes or in, in half an hour, we were able to reach something which this person found of immense value in the presentation that that person was gonna make to the board.

Mehmet: Right. So pr on, on this point, like, uh. It's like kind of philosophical. I would say for me, um, I think, you know, the first phase of knowledge, uh, sharing was the internet, of course. And Google came, of course there were like search engines before, but their way of, you know, organizing [00:23:00] the internet data was, it was not.

Mehmet: Yeah. Right. But now, because you just mentioned this and I was thinking this is why, like. Took me like a couple of seconds thinking about what you just mentioned, like how abundant is knowledge now with these LLM tools, right? So I imagine, and the nice thing you don't have to read again, like this is a, this philosophical discussion, you know, because imagine now, because even you can ask perplexity, grog even, you know, chat GPT, they give the open ai, they give the access for their, uh, deep search reasoning also as well.

Mehmet: And the nice thing you don't even have to read, you can listen to it. Right. So yeah. Now with the voice 

Prashant: enabled, uh, features, I mean, you don't even need to read. 

Mehmet: Yeah. It's, it's, so, it's fantastic. Now coming back to reality, um, from your perspective, you, you've done this as a practitioner, you've done this as an educator.

Mehmet: What do you think? Let's say I'm new to this [00:24:00] field, or maybe I've been sometime, so what are like the new. I would say skills, critical skills that I need to develop to make sure that I'm working effectively with these AI systems and tools. 

Prashant: Oh yeah, this is, this is a memoth is something which, uh, you know, anybody who was in the, I mean, why only people who are like me in the, in, in the education consulting field, or I mean, any of us who are involved with the next generation, whether as parents, whether as, you know, seniors in a family, trying to guide the younger generation on what they should be scaling up on because, uh, I mean, you know.

Prashant: Uh, you know what this reminds me is, uh, a, a quote from, uh, one of these, uh, noble laureates, right? Uh, Nils Bore, if I'm not wrong. Who said that? Uh, uh, prediction is very difficult, especially about the future, right? So, so. Essentially what you're, what you're saying right now is we [00:25:00] are looking to the future and trying to identify what are these skills that probably people will require over the next few years.

Prashant: Uh, very difficult to predict what will be required in the next, again, in the 10, 15, 20 years. But what you require now today, and even that is something which I'm not sure that a lot of people are paying attention to, and they must is, uh, what I would call data literacy. Uh, mm-hmm. You don't need to be a data scientist.

Prashant: Uh, no. Nobody's asking you to become a data scientist overnight, but you need to understand data, uh, you, because a lot of this value that we can generate from these tools comes from the kind of data that we provided, the kind of information that we are wanting analyzed. Why are we wanting that analyzed?

Prashant: Uh, so having a, a sense of some kind of data literacy, you know, to be able to question insight. So if AI throws something at you. Uh, you can't just take it at face value. You should be able to question that data point and you should be able to, you [00:26:00] know, interpret that in a certain way. So that kinda data literacy is something which I feel is a critical skill, uh, going forward.

Prashant: One thing which has emerged definitely in terms of usage of AI tools, and I, and I feel that everybody will be using AI tools sooner than later. Is what I call prompt engineering competence. So there is a way to deal with AI system. So I, when I work on a chat, GPD, you may be having a chat. GPDI may be having a chat, GPD.

Prashant: And I've had these conversations with multiple clients, multiple people, professionals, they say, but I also use chat, GPD, my chat GPD doesn't give me the same kind of response that your chat g PT gives. I said the model is the same. It's not that I'm using some different model and your. Using a different model, there is a way to prompt and play and, and, you know, guide the AI model in a certain way to be able to generate the best value out of it.

Prashant: So I think more and more what will be, what will sort of discriminate between somebody who's good at this job and not good at at the job [00:27:00] is going to be the ability to deal with these AI systems and to be able to guide the AI system. Towards the kind of output that we are looking for. Uh, one very other important thing I really feel is.

Prashant: You spoke about knowledge, right? So knowledge is becoming abundant. It's becoming cheap, it's becoming available. It's no longer about, uh, knowledge. It's about, or even, not even about problem solutions. I mean, we are now talking about, you know, you give a problem, you'll get a solution. Uh, probably, you know, again, sooner than later, it's about figuring out what problems to solve, right?

Prashant: So. As a business person, if I'm not clear, I can keep throwing problems at the AI tool and lo and behold, it'll keep solving it, but it'll, it'll, if there was information overload earlier and we used to speak about information overload earlier, we are gonna have solution overload now where you have so many solutions, but you don't know [00:28:00] why did you ask for so many solutions?

Prashant: Was it really important for your business? Have you taken the effort of identifying. Uh, the problems that you want to solve or you must solve, and addressing those through the AI solutions rather than just because the AI solution. Uh, very interesting. Side note, again, I I often mention we, we try and compare.

Prashant: We, we have this habit of, uh, and anthropomorphizing ai, right? So we try and talk to it like a human and, and, you know, we start imagining it like a human. There's one big, big difference between AI and a human. After a point, a human will tell you no. Uh, I mean, whoever it is, it's your assistant, it's your colleague, it's your supplier.

Prashant: AI doesn't know how to say no, so it'll keep answering whatever you keep asking. It'll just keep answering that, right, and only you will be left with all this information overload to sort through. So, so the, again, you know, very important [00:29:00] skill now is to actually being able to identify the problems to solve and then use the AI tool to actually solve those problems.

Mehmet: Right Now, you mentioned something Prashant about people a couple of times, and I've seen it by the way, but maybe you can shed some light. Uh, and this is for maybe the leaders. I would say this question. Sure. Uh, because what you mentioned at the very beginning about like some people they think it's, uh, you know, super high and some people, you know, they have also there.

Mehmet: Doubts also as well. So we see companies, you know, getting this dilemma of implementing. By the way, it's not only for the ai, it happens for all the technology, it happens 

Prashant: for all tech. Yeah. Right. Absolutely. Any new thing comes and 

Mehmet: yes. So, but now for the AI specifically, it's, it's a journey. They're gonna start from somewhere.

Mehmet: What do you think would be the right approach to take [00:30:00] so they can. You know, strategically, you know, futureproof their team that, Hey guys, look, I understand you are afraid maybe that this is gonna take your place. Maybe you are thinking it doesn't do a great job, whatever it is. So what do you think the leaders strategically should do so their organization is really protected in the future?

Mehmet: Because I think you would agree with me, Prashan. That I don't imagine a future without these technologies implemented. 

Prashant: Yes. Uh, you're, you're absolutely right ma. AI is going to come in into every industry, every function, whether we like it or not. Some companies are gonna be early adopters, get the benefit of it, uh, faster than their competitors.

Prashant: The others are going to be laggards. It's always, it always the case. Sorry about that. Any new technology, any new thing which comes, there are people who are early adopters, they get some [00:31:00] additional benefits versus the competition and eventually the technology, if it is, uh, like, uh, Andrew Wing says, you know, AI is the new electricity, so we, we don't think about electricity.

Prashant: It's not, it's not equal to the internet. So people think of AI as the internet, you know, where, oh, internet came in and then, you know, it's not that it's actually electricity. So we don't think about electricity. Electricity runs the entire world right now. And we don't think about it. And AI is gonna be like that.

Prashant: AI is gonna run everything in the background and we will not even talk about it. So it's, it's gonna be one of those kind of things. So if it is gonna come in and it's gonna come in and pervade every single function and every single industry, you're absolutely right. That sooner or later everybody will have to implement.

Prashant: And, and what probably are the ways to overcome this initial. Inertia resistance, whatever you wanna call it. Uh, misconceptions. Uh, one of the, one of the classic, uh, playbook mod, you know, uh, modules that I have on this is pick up one case. [00:32:00] Don't try and revolutionize your marketing department overnight by saying, oh, we are going to just have a totally different workflow process than what we used to earlier haven.

Prashant: Seeing marketing, uh, because I personally deal more in marketing space. But it is true probably for any function that you start implementing AI in, is that don't pick up, don't try and revolutionize overnight, uh, you know, pump in half a million dollars into completely revamping without even starting.

Prashant: Don't do that. My, my, uh, recommendation always is start from a simple use case. Where you get a quick win, so you start. See nothing. Nothing succeeds like success. So you get those quick wins, you will start having those naysayers, whether they are people who are, uh, trying to put down the technology by saying, oh, it doesn't do anything, or there are people who are afraid of it, or whatever.

Prashant: Once you start using it and you start getting the wins in your day-to-day workflow is when [00:33:00] the buy-in is much easier. So it's one of the ways to bring in, uh, bring in change in an organization is actually to start small. Create one use case, which is important to you, but is not everything, and probably, uh, you will, you will get a quick win in it, and therefore, this, the other very important one I feel, which is not being, which is not being looked at very carefully right now, is get your people into training.

Prashant: People right now are still, companies are not spending enough time and energy at a company level. So there are individuals who are always interested in learning and trying to figure out sometimes half baked stuff, some, sometimes, you know, some, uh, stuff, which is not really, like I said, there are enough.

Prashant: Uh. People out there in the social media today claiming to be experts and, and, you know, giving some wrong directions to people. So organize proper training workshops with people who are from the industry. There are people you know, who are technology experts, who are business experts, who can be brought in, [00:34:00] uh, organize simple training programs, get your people started.

Prashant: Uh, do that as an exercise, uh, as a, in, in a company. It'll go a long way in making sure that, because see, the people who are in the business, and I've been a business person for 22, 23 years, right? So people who are in business, they have what we call as institutional knowledge. You can't have software tools replace anything of this.

Prashant: You need to augment, you need to start changing your mindset from replacement to augmentation. Saying that, how do I make sure that if I had a guy who, uh, who understands this particular space so well, but is overwhelmed with the, with the kind of work that this person has, I want to improve their bandwidth by providing them the support of the AI tools.

Prashant: Let me augment some of the repetitive tasks, some of the tasks which, uh, are easier for the AI to do than for the human being. So that, that's my playbook really saying start small, pick up one use [00:35:00] case. Uh, go after it in a nice way. Probably give yourself 90 days to implement it and get your quick win.

Prashant: Hire people. Uh, it's not very expensive. You can get people to train your entire teams in a few thousand dollars. You know, it's not something. Uh, which is really expensive. Get that going. Uh, engage more, uh, on an ongoing basis. Continuous learning is, I mean, AI is changing so far that you, it's not about doing it once, it's about actually starting the continuous learning process somewhere.

Prashant: So that's my 

Mehmet: take. 

Mehmet: Absolutely. And I always tell people what you're missing right is. Going to any of these LLM tools? For me, it's not about charge GPT, honestly. I mean, anymore I use, I think I use them as equal, you know, in time perspective and, uh, amount of, uh, time I use.

Mehmet: But of course, charge GPT was my first encounter, let's say, with these LLM tools. So I say to people, the thing that you are missing [00:36:00] is going and brainstorming with these LLM tools. If you try once. Instead of saying, Hey, even if you give, by the way, the right prompt, hey, charge GPT, like, uh, act like, act as a, uh, you know, experienced marketeer in this niche.

Mehmet: Uh, I'm trying to do X, Y, and Z, so I need, for example, I don't know, a post for LinkedIn. Okay. Still kind of. Give you something, but what I discovered, like if you make kind of a conversation first, because the more data you feed. So actually, and back to your point of understanding the T-Mobile and you know 

Prashant: Absolutely, absolutely.

Prashant: So what you, what you just mentioned is actually a, a very, uh, you know, important technique to deal with ai. Uh, I mean, it might be useful for your listeners, I'm just gonna put it out there. It's called the few short example. Uh, basically you [00:37:00] are giving the LLM. Examples of what kind of tone, what kind of language that you want so you're not just leaving it out there.

Prashant: Every brand has a brand voice. Every brand has a brand personality. So if you try and so let's say for example I tell, uh, LLM and I give it the persona pattern, like you mentioned about giving it. Uh, some kind of, you know, saying, oh, you are an expert in, uh, you know, in marketing and, and or creative development, et cetera, and I ask you to develop, let's say, five slogans for an ad.

Prashant: The kind of output that you will see from that versus if you one, provide examples to it saying that, okay, this is the kind of. Uh, stuff that we do on this brand, and now I want to think about new examples. It'll completely change the, the kind of output that you will get. The second, I think, which you've mentioned a very, very important point, and and thank you for bringing it up because I missed it, is this whole [00:38:00] iterative part, so I didn't mention it enough.

Prashant: You mentioned it very correctly, saying that it's about the conversation. It's not about one time output, so don't take the single output that you get once and and say that, oh, this is the final. So one prompt, one output. That's not the way to work with them. You get an output, you put your input into it, then you get the next output, then you put your input into it.

Prashant: Then you get the next output. Very, very quickly, uh, because that's what we used to do in, in, in without the tools as well, right? When we sitting in a, when we're sitting in a room and doing brainstorming, we don't look at the first idea and say, okay, done. You know, let's, let's, uh, go and start executing this idea.

Prashant: We, we do it iteratively. We come up with some ideas. Then we say, okay, but somebody else adds something to it. Then we refine that idea. Then we go back and again, you know, change the idea a little bit. Think of the AI tool as one of the participants in that, uh, in that brainstorming exercise. And, and suddenly you'll see the, the value that is completely [00:39:00] changing.

Prashant: So thank you for bringing that up, ma. Very, very important point, uh, over there. 

Mehmet: Yeah. So just then the matter of, uh. Sharing also something with the audience. So the reason it's not like, because I was smart enough to, to, to discover it this way because when the first time open AI decided to open chat GPT for everyone.

Mehmet: So I was like, kind of little bit, I would say, Hmm, let's see. And you know, my computer scientist or engineer brain was saying, okay, let's go back to the Turing test and just try to find out. Like how good it is, right? Yeah. And it didn't take me more than two to three minutes because, you know, after a couple of in and out with Chad G PT at that time, I said, okay, yeah, I shouldn't have been like that.

Mehmet: Uh, you know, uh, kind of. Looking at a, like, a little bit, uh, doubting way, [00:40:00] and I said, no, these guys are up to something. It's not like yet another one of these famous, you know, chatbots that were coming so hundred percent good right now, Prashant, like as we almost coming to, to an end. So, and you know, almost like we finished three months as in, in 2025.

Mehmet: Right? So what are like some of the developments and breakthroughs that. You know, are, you are excited about 

Prashant: the the latest ones that, that you see coming out now? Uh, see, I, I believe one of the things which I've been at least hearing, and again, you are the technologist, you can probably add to your, uh, to, to this from your perspective.

Prashant: What I've been seeing is this idea of shifting from, uh, answers, question answers, or a chat bot kind of a scenario. To what we would call agentic behavior. And I'm being very careful using a buzzword agentic because I've heard this being thrown around. Uh, but I'm not using it honestly like a buzzword.

Prashant: I'm using it more from a very practical [00:41:00] point of view, saying that, uh, is it something which, you know, where I can automate certain tasks to the ai, which currently was not possible. So I was currently doing it more like a. Chat bot. But if I can get, uh, AI to actually, uh, do step one, trigger the next action and trigger the third action and then come back and report to me, uh, and close the loop in that way as a marketer, that's, that's.

Prashant: Tremendous value. So, uh, I, I'm sort of looking forward because a lot of talk is happening about, you know, every, every of these major companies is right now going full steam on the agent, uh, AI space. So I'm quite excited about, uh, you know, this year, uh, something on that front that we will probably see, uh, one thing, which I really feel.

Prashant: Has not happened enough. And, and we touched upon this a little bit in our conversation is on the hallucination front. So [00:42:00] a lot has happened. I, I must say that, uh, the technology has evolved dramatically from where we were one and a half years back. Where really one was struggling with hallucinations, every single prompt, every single, uh, task that you would give the AI today, it's, it's dramatically improved.

Prashant: But if you ask me, I'm still as a, as a user, I'm still hopeful that I. We will get further dramatic breakthrough on this because still the time that happens. See, business is about reliability. Uh, it's not just about output, right? It's also how reliable that output is. So I, when I have a team, I'm, I'm asking for input from people.

Prashant: There are people in the team, you know, they know their stuff and you know when they're saying something because it's impossible for you to go out and check every single thing of them. So we keep. Warning people when they're using AI that please check AI's output because it can hallucinate. We keep saying this as a disclaimer, but really in terms of usage, uh, the trust and the [00:43:00] reliability from business, uh, would come in a lot.

Prashant: If the, if the, if the companies which are working behind this are able to further do, uh, you know, good work on this and, and improve the software and the back, the backend, uh, enough in terms of the. Whether it's the number of billion parameters, uh, that they're trained on, or whether it's, uh, faster chips or whatever goes behind the, you know, under the hood to, to get these, uh, uh, models to, uh, to be more reliable.

Prashant: So I, I'm sort of looking forward to that in 2025. Uh, motion. 

Mehmet: Right. So just, um, because you asked me from my, uh, technology, you know, uh, aspect. Agent ai or AI agents as people likes to call them. I think, and you know, I was lucky enough to predict that this will be the next wave, even in 2023, by the way. Uh, because I was following a pro, multiple open source [00:44:00] projects trying, so their aim was trying to let.

Mehmet: A couple of LLMs talk to each others, and this is where the whole idea come up from. But of course, we, we saw like breakthroughs and I'm believing we're gonna see more breakthroughs in, in that domain. Uh, and it's going in parallel with what some people are calling. We are very close to. A GI, which is artificial journal intelligence, which is like, meaning a machine can have the same intelligence as a human.

Mehmet: We are close. I, I believe we are close also as well. I believe we are close. Yeah. Um, and, and for the hallucination. So I think a lot of things happen, but I think, yeah, to your point, there is a lot of breakthroughs expected. Um, so the first, so this is the nice thing about the ai. It was deployed people immediately, the people who deployed it, and I mean, I mean by open AI and Google and Entropic and all these companies, they figured out fast that, [00:45:00] yeah, so there's a problem with hallucination.

Mehmet: So they start implementing the tri augmented reality, uh, reality, you know, so this is the, what they call it, uh, rag. Uh, so rack log 

Prashant: generation, correct. I that's what, yeah, exactly. 

Mehmet: Yeah. So, yeah, sorry. It's Ament generation, so which is make sure that m whatever it spit out, it's also like, you know, checked for, uh, fact, right?

Mehmet: But again, we are going into spaces, which even me as a technologist, you know, like now, and it's funny enough. I remember when I started the podcast, JGPT just came couple of weeks ago, right? So not many people still knew about it. And you know, when I asked one of my guests, Hey, like, what's your expectation in this field in the next decade?

Mehmet: Imagine I'm asking next decade now. I don't dare to, I don't dare to ask more than six months. 

Prashant: Absolutely. I mean, track everyday news. Uh, on AI because [00:46:00] you have one, one model coming out and it, it claims to be the best and, and it is the best, uh, because people check it. And in the, in five days time, you have a better model from another company coming out, which is better than this one.

Prashant: So, so you're absolutely right that, you know, uh, to, to imagine something beyond six months today is really, uh, you know, I don't know what, how to, how to even do that. Yeah, 

Mehmet: some people, the skeptics once, of course they say, yeah, but you know, like, uh, a lot of companies appeared and disappeared very fast. I say, this is good.

Mehmet: This is not bad because. You know, what made Silicon Valley good? Some people they agree, some people they don't agree, you know, fail often, fail fast, right? So you have to fail very fast. And we, we saw the example of this human, you know, the, the pin thing that, uh, HP bought them. I said, okay, so what's wrong?

Mehmet: These guys tried, it didn't work, and we're living in the AI age, which everything should happen fast. So either we have a fast, uh. [00:47:00] What I call it, uh, valuation, uh, sorry. Uh, validation of ideas. Why we need to wait two, three years until we understand if something gonna work or not. We should understand much faster than this.

Mehmet: But anyways, it's a discussion for another time. Prashan, I really enjoyed the discussion. Before I let you go, I want you to share with the audience where they can get in touch and find about. 

Prashant: So, uh, thank you very much, member. That was very interesting conversation for me as well. I, I mean, you know, I, like I said, I'm a marketer who looks as tools as solving real world business problems.

Prashant: You are sitting on the other side of the table as a technologist who's. Helping us get these, uh, different tools, uh, through your career in, in technology? Uh, I think it's a beautiful conversation because the more we understand each other, I think the better it is for the kind of output, the kind of real world value that gets generated because we need to understand what are the technological limitations As marketers, as users, we can't expect the tools to do everything for us, uh, uh, [00:48:00] we, we need to understand that it is evolving and we need to start making use of the current capabilities.

Prashant: Of that, uh, that the technologists like yourselves have provided us. And, and similarly at your end, maybe hearing out from a user, you know, and a marketeer that what, what really are we valuing? What are we looking at? You know, because we need to solve those, uh, real world business problems, uh, to, to, you know, using these tools.

Prashant: So. Great, amazing experience talking to you as well. Uh, really enjoyed it. Uh, people can, the best way to connect with me is on LinkedIn, so maybe if you can, I don't know if you have my LinkedIn, uh, ticketer there. Just put my LinkedIn, uh, profile there, and then that's the best way for people to reach out to me, drop me a dm.

Prashant: Uh, I'll, I'll I come back very, very quick. I'm, I'm quite responsive that way, so, uh, please drop in your line and, and I'm very, I would be very happy to engage with any of your listeners. 

Mehmet: Great. Thank you very much pr and yes, as you said, I will make sure that. The link to your LinkedIn profile will be [00:49:00] available in the show notes.

Mehmet: If you're listening on your favorite podcasting app, you'll find that in the show notes. If you're watching this on YouTube, you'll find it in the description area. Again, Prashant, really, I really enjoyed the discussion with you and it was very beneficial. I learned also a lot of insights from you also as well, and this is how usually I end, uh, my episodes.

Mehmet: This is for the audience. If you just discovered this podcast by luck, thank you for passing by. I hope you enjoyed it. If you did, so give us a thumb up, subscribe and share it with your friends and colleagues. And if you are one of the loyal fans who keep coming again and again, thank you for doing so and thank you for all the support that you have provided to the CTO show very recently.

Mehmet: As I'm sharing in the recent episodes, you helped the CTO show to really go to the next phase of success, having the podcast. You know, trending in multiple countries at the same time, in the top 200 charts. So this is, have been unseen before also for having the CTO show in Mead [00:50:00] listed as, uh, one of the top 40 Must Listen business podcast in Dubai by, uh, feed Spott.

Mehmet: So thank you very much for your support and I really appreciate all, all, all your suggestions, you know. Questions and you know, any, uh, thing that you, you sent to me. So I really appreciate it. And then again, as I say, always stay tuned for a new episode very soon. Thank you. Bye-bye.