Episode 165 / Fatima Baz / Unilever / Digital Innovation Manager in North Africa, Gulf and the Middle East
Podcast: How to Make the Most of AI in Marketing
A computer engineering graduate who started working in innovation and digital transformation from the very beginning, Fatima Baz is now the Digital Innovation Manager in North Africa, Gulf and the Middle East at Unilever. She’s worked with various businesses, from start-ups to big organisations, and is an expert in AI and the supply chain, marketing and e-commerce. Her Shiny New Object, it will come as no surprise, is how innovation and AI impact business and marketing.
Although she spends a significant amount of her time in front of screens, Fatima’s best investment, personally and for her career, have been books. She’s an avid reader who wants to stay up to date with the latest innovations, but also be informed of what’s happening in the world, read up on personal development, and be inspired by thought leaders she admires, such as Elon Musk. Fatima recommends reading about alien life and artificial intelligence, looking at how the latter can actually be harnessed in many useful ways for humans.
We talked about xenobots and creating artificial organs for humans, but also about the difficult question of whether AI can have a conscience and true decision-making capabilities beyond what it is programmed to do. In Fatima’s view, AIs should be able to not just make data driven decisions according to their programming, but also make decisions using data and their own judgment, to better provide business solutions.
To ensure that this happens, her advice is to add more data and complexity to the structure of AI, which will help it learn and yield better results. This starts with having AI embedded in every department of an organization, as opposed to the more traditional set-up that would keep it separate. As we provide more complexity for AI, we will provide it with more cognition to be able to better run business figures.
Fatima knows that there are people who argue that using AI in marketing doesn’t quite work. But, in her opinion, this doesn’t have anything to do with the recommendations made by AI, but with the system that companies have built. Yes, there will always be a level of human intervention needed, but brands should test and see what works for them and then might find themselves getting a lot more out of AI.
Fatima recommends having an environment where brands can do A/B testing with ideas issued from AI, before going live at scale. As with any strategy, a shift should be made progressively, not overnight. But, in all situations, combining AI with human intervention and working to use data as much as possible in real time will be the winning recipe for marketing strategy.
Listen to Fatima talk about her passion for books, the future role she sees for AI in organisations, and more advice on how AI can work in marketing, in the latest episode of the podcast here.
Transcript
The following gives you a good idea of what was said, but it’s not 100% accurate.
Tom Ollerton 0:06
Hello, and welcome to the shiny new object podcast. My name is Tom Ollerton. I'm the founder of automated creative. And this is a weekly podcast about the future of advertising. Every week or so what we do is chat to early eventual guests about what their vision is for the future of the industry, and go a bit deeper into the subject that they're most excited about. And I'm on a call with Fatima Baz, who is digital innovation manager in North Africa and Middle East, Fatima for anyone who doesn't know who you are and what you do. Please, could you give me just a bit of a background to help the audience understand your journey and where you are today?
Fatima Baz 0:46
Hello, Tom. Yes, for sure. So my name is Fatima Baz. I'm a Computer Engineering graduate who got into the field of innovation and digital transformation at a very early stage in my career, I have evolved, my career has evolved from working with startups on ecommerce platforms into working on digital transformation programs for big organizations. I touch base on AI and influence on supply chain, marketing and ecommerce.
Tom Ollerton 1:22
So I'm keen to know what has been the best investment of your own time, energy, or money in your career.
Fatima Baz 1:30
You know, my best investment and it's something that I have been working on for the past three years, it's books, buying books, reading a lot of books throughout my every year actually, I have like limit or I have like a goal to read more than 30 books a year. It's amazing how much how much of transformation and and advancement does that offer to every person. As an innovator, I assume it's really crucial for you to stay up to date with the latest innovations and what's happening around the world. These books can can vary from personal development books into books that touch on the economy, and a lot of other issues that are happening in the world. I always choose the books that are recommended by Elon Musk, I find him to be my idol, to be honest, he's a very visionary person. In the latest books I've been touching upon are related to AI, aliens, the presence of aliens in our world and how we can better control artificial intelligence now.
Tom Ollerton 2:42
I've never heard anyone talk about aliens on the podcast. So we've got to dive into that a little bit. Tell me tell me what has Elon Musk being telling you to read about aliens?
Fatima Baz 2:52
You know, he's the type of person who believes in the probability of having a foreign life other than life on earth. From a probability perspective, it's actually something that's very probable and can exist. The universe is very fast, and it's very actually big. And yeah, it's something that he's touching upon, is also in some of the books he recommends. When he talks about alien life, he is referring into artificial intelligence and the way artificial intelligence can actually take over a lot of tasks that are happening in the world and be able to have their own conscience. So he talks about the the field of nanobots, which are robots that are first constructed via frog cells, and then have the ability via AI to to multiply on their own without being programmed to do so. So yeah, it's a very interesting field. And it's very futuristic.
Tom Ollerton 3:50
Sorry, go back to the self replicating AI frogs. That sounds fascinating. So just help us understand that a bit better. I'm curious.
Fatima Baz 4:00
So, you know, xenobots, they are kind of AI organisms. They're made of frog cells that are programmed to do certain functions that some of our audience are not able to do. So assume that you have a kidney failure. So this field of the medical field that allows us to replace that organ with certain cells that will be able to do the function of the kidney, for example. And they have noticed, a lot of medical practitioners have noticed that these cells are able to do some functions that they're not programmed to do. And it's something by the way, that has been there for a long period of time. And Mark Zuckerberg has touched upon this stuff. He was talking about certain servers for Facebook, that were able to do certain tasks at a faster pace, and not by following the way they're programmed and not by following the code that they're programmed by. And here comes to the subject of having AI or AI having a certain conscience. And a lot of the books that he recommended for 2022, they touch upon paying attention on how we're developing AI, and always having certain rules to know how to be able to control. It controls their ability to actually perform on their own.
Tom Ollerton 5:21
And how does the consciousness piece come into it? One of the things that always confuses me is that we're all aware that we have a mind, but no one's been able to prove the existence of it yet. So the idea that we assume that something else is going to have a mind that we can't prove seems scientific, but at the same time, like I said, there's a tension there. What's your view on that?
Fatima Baz 5:49
You know, my view, I think it's something that has been embedded for a long period of time in science. You know, science has debated for a long period of time, that certain particles, small particles, they have conscience, there was, I think it's an experiment that was done by Newton, it was regarding a beam of light, I might be wrong in some aspect, like with some names, I'm sure about the experiment, but I'm not sure about who ran it. It's about having a beam of light touching on a certain surface. And they will tell you that when the light particles were aware of somebody watching them, the way they have moved has varied. So it's a scientific sector that has been debated for a very long period of time, has been existing, by the way, since the 90s. It's not something new. But now in practice, it's something that they are seeing happening. I deeply think that, since AI is designed in a way not only to make data that is data driven decision making, that we program them to do, they're able also to do data driven decision making that they think is better, to better provide a business solutions. So since a lot of the decision making that they do is not actually how we program them to do and since we expect them to do some kind of smart decisions. I think this is not something that is very impossible, I think it's something that will happen sooner or later.
Tom Ollerton 7:23
Yeah, this is where I have the uncomfortable problem, trying to bring this conversation back around to marketing and obviously running a business that has automation at its core and uses machine learning and artificial intelligence in our products. But ultimately, the challenge that I talked to a lot of our clients about is the fact that with AI, there isn't any cognition. But really, they can understand that it can do clever things and make moves and suggestions that a human wouldn't but doesn't really understand. So for example, when a GPT three, for example, which is for anyone who doesn't know, it is a open source algorithm that can generate copy, really impressive copy. But it doesn't understand what it's saying. It's a parlor trick still. I mean, it's incredible. Like, you know, if you want to generate free blog post copy, you can easily do it. But it doesn't actually understand, it's not relating to those words, feeling those words. It's not cognizant of what it means and it and so from GBT three hasn't has no feelings, it doesn't actually take on board those words, and the meaning. So therefore, quite often, something like GBT three will fail when it comes to writing copy for a brand because it doesn't understand the context within which it's talking. Right? So sure, I could come up with lots of free copy lines, for quick copy lines for a brand. But then the people or the brand or the agency will have to go through going no, no, yes, no, no, because they apply the cognition to the content. So I'm rambling horribly, I'm supposed to be interviewing you. But you've got me really interested. So I guess the question is, is that it? Does it need to be cognizant, to perform useful marketing and advertising functions?
Fatima Baz 9:19
Very interesting question. I think the more you bring data and complexity to the structure of AI, the more you will be granted better results. That's known. For example, if you look at all the strategy of all the companies, organizations peer, they always used to have one team that is into AI and digital innovation. And this team will give their feedback to the other teams, sales and marketing to better manage their campaigns, but now the structure has been different. Now the organizational structure to actually have good results or a good business strategy, the structure is to have AI embedded in every department. And that means that we provide more complexity for AI, and in a way or another, it will be provided with more cognition to be able to better run business figures. Before AI was more of an accessory, now AI is actually a necessity for the sustainability of any organization. That's, that's what I believe in. Take, for example, the real time bidding structure, right? Like they're deploying machine learning to control bid price. They're deploying machine learning to maintain liquidity to monetize the content, we're talking about advertising, right, and to make sure that there's safety on the platform, because they will be able to filter, any bot any what actions. So I think the more you rely on or the more complex your AI, your AI development or the way you're embedding AI in your business processes, the better results you will have, however, there's something I always stress, it's very important to make sure that you have enough data. So first of all, there's a lot of work that needs to be put in the structure of the organization, you have to have a structure where you understand you know, the data that you are having you understand all your systems, you understand this data, you analyze properly, to be able to develop the right AI platforms that will fit with the strategy. So AI is not a strategy on its own AI is an enabler for business strategy to be able to better achieve results.
Tom Ollerton 11:29
So do you have a best marketing tip that is related to this or unrelated to this, like you read a lot? So do you have a marketing tip that has really helped you with your career, something that you share quite often.
Fatima Baz 11:48
My marketing tip would always be to base everything on data consolidated and analyze it, I would say there's a lot of strategies that you can follow like predictive analytics, embed AI in a lot of your strategies, have a strategy to improve your CRO conversion rate. These are all strategies, but the main tip in in setting a marketing strategy and knowing what's best to do next is to rely on data and analyze, right? It will give you the ability to actually learn from what worked in previously, it will give you also a vision of what are the trends that you have seen. And it will give you a better ability to actually reduce your costs, but improve your revenue by implementing the right thing. So you will not be able to actually improve your strategy if you have not learned what works and what doesn't work from before.
Tom Ollerton 12:47
So we're at the halfway stage now. And normally here, I would reveal what your shiny new object is. But we've kind of covered it really, certainly in title and your shiny object is how innovation and AI impact business and marketing. So let's let's carry on that conversation. So I think it's clear to the audience what what your view is and how important it is. But one of the things that I've been battling with is that is, for example, this is a completely hypothetical case. But what I share often, but imagine you're scrolling through Instagram, and everything you see is red. And then you see an advert that's blue, okay, an AI would tell you that your ad will always be blue. But in reality, it should have been anything that isn't red could have been yellow, or pink or spotty, or transparent or black and white, it doesn't matter whether it's in an advertising context, the issue we have is that you can get the data back from Facebook or Google or wherever you're serving your ads that always it was the it was the blue ad that worked. But no, it wasn't it was the fact it wasn't red that mattered. So how do you as a brand deal with that kind of conundrum where, yes, you can get a lot of data, so much data or an audience, but unless you've got the data that provides the context, how can you be sure that you're making your data-based decisions on the right recommendations from something like an AI?
Fatima Baz 14:16
Hmm, very good question. You know that there are a lot of people who have argued that the way AI is implemented in marketing is not actually granting results. And a lot of people who have argued also that 94% of these recommendations were inaccurate recommendations, as per application. I think it's not about the AI recommendation. It's about the system that you have built. I'll just give you like, I'll give you a description, something that would actually describe the scenario. It's like when you want to do like innovation in the AI and you build it on legacy systems that you do not know what is there in it, right. So what's built on... We say - "trash in and trash out". So if you're building something or something that you are not quite understanding, that's when actually you will not get the result that you are seeking. So in the example that you are providing, you're saying, in a lot of cases, Facebook does not give you the right recommendation, it has a lot to actually do with, what are your target audience on Facebook, who are your target audience, it has a lot to do with the way you're seeing data that Facebook is providing. It's something also that's likely related to I'll give you an example, certain a lot of the ads that people or that certain corporates show, or in affiliate marketing even, you will see that these ads have a lot of click through rate. But the conversion rate is not working. That doesn't mean that my ad is not working, it actually means that the landing page that I'm converting them to might not be appealing, the price might not be working. So there are a lot of scenarios that you need to take into consideration. And that's where analysis is very important. As I said, the way you view this data, and this analytical ability to be able to distinguish is quite important. Building the right AI is also necessary. And I stress here also the fact that we always assume that especially in marketing, because AI implementation in marketing is not quite as complex as in other fields. It's very important, we shouldn't forget that the human intervention here is important. So I will not replace actually a human intervention here, you will have to have a human who will better analyze all this data and better do these recommendations. In a lot of cases, it's a flaw in the strategy or flaw in seeing certain data, or the way that we're classifying data. In other cases, you have to always test, see what works for you. Like I'll give you an example. Also, in Rocket Internet, I'm talking about Rocket Internet, the first company I used to work with, when we wanted to do any, implement any change on the landing page, or creative on the website or anything, we didn't directly go or implement the change, you always have to have this space, if I might say, where you are doing what we call A/B testing and should be implemented on everything. And a lot of double click for advertising platforms do it, you do not directly shift into the new creative that you have, what you do is you do this gradual shift because you want to make sure that this shift is successful. The only way to grant it is to compare what was done before to what's being done now. So you direct half of the traffic into some, some strategy and the other half into another creative or another version of your website, for example. And you see which traffic is converting more. But to do that, for example, you have to make sure that all the rest of the variables are constant, because only then will you know that the creator is the one who's driving the change. So what I'm trying to explain here is like the way of thinking about any strategy or about any methodology that should be followed, you have to always make sure that you're not directly shifting into any strategy, you have to test first, only when it's successful, you can completely do the shift.
Tom Ollerton 18:04
Yeah, it's interesting, that point and I like the idea of the gradual shift across and the point you made is, it's one that we say all the time, right? If you haven't isolated the variables, then you can't be sure of the outcome. Whereas, the reality is, you can only control the variables that you can control, right. And so I strongly believe that the context within which your ads are seen is the important thing, and I think that there's a horrible reliance in the industry about making predictions based on past performance. So say, for example, I'm just going to make this up that you look at lots of burger advertising from the last two years across lots of different markets and lots of different regions that an AI will tell you, you know what to put in a burger ads. But if, I don't know, there's a giant documentary on Netflix about how bad burgers are for the environment that was really huge and was number one in that chart, then that data would be misleading. Because you, as the advertiser wouldn't be able to control so our methodology is very much making the ads react to the moment, react to the context within which they're being seen, as opposed to creating a learning model that looks at the vapor trails of yesterday to predict what happens next. And so, yeah, you've got me on my favorite subjects. I'm rambling a bit, but I'm keen to know what you think about that.
Fatima Baz 19:36
No, I think real time is always better and everything even now, by the way, relying on past data doesn't require a complex AI. It's not, you cannot even call it AI because it's just if I might say it's a strategy, it's discovering past trends and that does not require AI. AI is basically developing an intelligence that is able to predict the future and tell you right now, what is the best position to be taken. And that doesn't only rely on past data, that will rely on a lot of factors that are happening right now. Like you cannot... a simple example is you cannot target the audience the way they were acting in Corona like they used to act based on past trends, right? The trends in Corona have been completely different. So I think this is a very great strategy. It's actually how it should be done right now to target people st this moment of time. What did they search? That's why in Google ads and Facebook ads, sometimes you feel as if they're listening to you. And actually, in a lot of cases they are, so you will see ads that are very relevant to what you were talking about a moment ago, or what you were, who you saw even, the day before, and I think that's very important, not only past trends. Past trends is a strategy that is very old. It's like more than 10 years old.
Tom Ollerton 20:52
Well, I would love to carry on this conversation, but we're at time unfortunately, Fatima. So if anyone wanted to reach out to you to discuss this topic, and I'm sure there will be many, could you let them know where to get in contact with you and what makes a good outreach message to you?
Fatima Baz 21:07
For sure, they can get in contact with me via LinkedIn on my profile, Fatima Baz. Also, they can can get in contact with me on my personal email. Fattima.baz@gmail.com.
Tom Ollerton 21:22
Fatima, thank you so much for your time today. I really enjoyed that.
Fatima Baz 21:25
Thank you, Tom. Have a great day.
Subscribe to the ‘Shiny New Object’ Podcast on Apple Podcasts, Spotify, YouTube and Soundcloud.