Latest / The Interline Podcast / Finding The Feeling In AI Campaigns

Transcript

Finding The Feeling In AI Campaigns · The Interline Podcast

• speaker-0: Hey, welcome back to the Insuline Podcast. There's no escaping AI-generated images right now. From some of the world's biggest retailers to mid-sized domestic brands here in the UK, Europe, America, further afield, fashion organizations are turning to AI at massive scale as a stand-in for traditional photography. Both routine, high-volume website and PDP content and full-blown creative storytelling campaigns. In a very real sense, if you pause this podcast and you go shop online for a bit, or even scroll your feeds, I can practically guarantee you'll be seeing images that are either fully AI generated or, at the very least, AI enhanced. This is, to put it pretty mildly, contentious. Every study out there has reported what seems to be a basically universal sentiment that consumers aren't keen on AI being used to sell to them, which is kind of a Corollary and knock-on effect of the general groundswell of anti-AI feeling out there. But this is all also very clearly where the industry is headed. Technology doesn't reach this kind of adoption this quickly in that kind of hostile environment I just talked about without some very clear incentives in place. Fashion companies on balance, but they don't invest purely speculatively, which means that. AI images are everywhere because they and the workflows behind them have some inbuilt advantages over traditional photography. The advantage that springs to everyone's mind first is also the reason behind a lot of the AI pushback. It's potentially a lot faster and easier than photography would be. If you want to take photos of garments or footwear or accessories for your own storefront or to provide them to retail partners, then you need final samples. You need a studio. You need a human model, you need a photographer, you need a stylist, a makeup artist, and so on. And you need to align the calendars to get all of those things in the same place at the same time, unless you're big enough to have them on staff and in-house permanently. If you want to create the same kind of images for the same use case with AI, you need a final sample still. You need a few photos that you can get away with snapping on your phone realistically, and you need a little bit of time with the generative model and some budget to cover the tokens. There are, inarguably, fewer moving parts of a synthetic shutter snap than there are behind a physical one. There's a tricky continuum here now between efficiency and laziness. And AI detractors would argue that the latter is dominating, which is what is the reason behind us having a flood of slop. The advantage people come to second is cost. Take everything I've just said, and if you switch your decision-making matrix from time to budget, And even if the token prices increase as markedly as people suspect that they will when the big AI labs IPO, it seems inevitably cheaper to generate campaigns or PDP images than it is to photograph them. The other advantage people talk about is the potential that can come from being freed to some extent of those time and cost constraints I just mentioned. Now, that might manifest itself as more interesting, more inspiring, more unfettered creative marketing campaigns. Might show up as more comprehensive regionalization of content or more lateral coverage of images and video on storefronts, it might look like the fabled old do more with less. Today we're running through all of these, and my guest to do it is Glazione Rocha, who's the creative director at Lounge by Zalando, which is Zolando's off-price business unit. I reached out to Gladsiani for two reasons. The first is that Zolando itself is at the forefront of Communicating around its intent to scale AI usage in image creation. It also seems to be at the forefront of AI adoption generally, but specifically it's the forefront of ⁓ talking about it openly. You know, this isn't a company that's quietly swapping out a few product images for AI alternatives and taking a read on what happens and then maybe walking it back. This is a big retailer. Listed on the DEX, Germany's Blue Chip Index, doing more than 4 billion euros gross merchandise volume in its last quarter, saying that effectively generative AI is the future of its entire image and video creation operations. The second reason is that Glazione himself has been extremely open about how, where, and why Lounge is using AI to create campaigns. He's also clear-eyed about what he thinks the impacts will be on junior creatives, models, photographers, and more. I wanted this opportunity to quiz Lazioni partly because I think Lounge's most recent AI campaign, the summer one, offers a distilled version of the whole AI images debate. But primarily I wanted to talk to him because, as you'll hear, the workflow behind that campaign doesn't come across at all as lazy or driven by cost cutting. It comes across as a company taking its mission seriously and doing what it thinks it takes to get ahead of its future. All while still recognizing that that commitment is gonna rock the boat in some really profound ways in the present. So with that context out of the way, let me bring on Glacione Rocha, creative director at Lounge by Zolander. Okay, ⁓ Glacioni Rocha, welcome to the Interline Podcast. speaker-1: Thank you very much. ⁓ happy to be here. speaker-0: Pleasure is all on this side of the table, as per usual. Now, we start every show by asking our guests to do two things. Walk us through their day-to-day and then define something. Now, the definition thing, it can sound a bit silly where we say, well, you want we want you to define a really simple low-stakes word. But as you'll see, and as we've seen from some of the episodes we've done before, it can end up revealing quite a lot about people's roles, people's priorities, and about the direction the industry is going. Let's start with your job and what it entails doing day in and day out. Now, it's a cliche to say that your job description is changing as a result of AI because everyone's job description is changing. But yours specifically, Creative Director, is going through some seriously deep practical and philosophical shifts, I think. It might not always be clear to people outside the industry, but people within it, I think your job would have been fairly straightforward to understand, not necessarily do. But you it would have been straightly fairly straightforward for people in the industry to say, I understand what a creative director does if you and I had been having this conversation five years ago. Now I'm gonna guess that your day-to-day might be pretty unrecognizable to those people, ⁓ and pretty unrecognizable to what it used to be. I know you know this because your history is in traditional advertising, agency, big brand, in-house. And now you're working on direction and campaigns for lounge at Salando, and you're working with Internal teams who look like they've upskilled really fast into AI native workflows. You're working with external agencies who have even deeper generative expertise. ⁓ long story short, your your job looks really different, I imagine, to what it looked like. Tell me what your day is made up of right now, and tell me how much of that would look real and familiar and intelligible to the Glazioni of five or ten years ago. speaker-1: Course, I mean my day my day to day is basically trying to kind of like keep the team motivated and ⁓ having kind of like a really structured process that they can follow up on. we have multiple campaigns ⁓ that's running in parallel and each one have their own complexity. So my role ⁓ now is actually to ⁓ make the whole process a little bit more structured for them. we we fill a few months ago we created like a or we shaped a creative vision for the for the department or for the whole company. So I think like my my goal now on daily basis is that like they ⁓ they understand and they can follow this creative vision and the process that they go into the ⁓ in the making of the campaigns is clear and structured for them. and this this of course affect the end result. speaker-0: And that's and that's independent of the tools, right? So the way the way that you're framing that is that the yeah, so the creative vision lives and then the tools that you use to execute on it are their tools. They're they're what get used thereafter. ⁓ one thing just to clarify, because you you brought it up there, where does lounge sit in the whole Zalando structure? So you mentioned multiple companies and things. Just walk me through what the structure looks like. speaker-1: Yeah, I mean Zalando Group is ⁓ composed by different companies or different ⁓ different call it we call propositions. So we are the off-price one. So we are the ⁓ the outlet. in basically everything that actually go doesn't sell at at the big Zalando, we call it like fashion store, ⁓ our team acquire these products and sell it at discount. speaker-0: D do you also design products for outlet for off price? Do you design your own products or or is it solely made up of intake from the other brands? speaker-1: No, it's totally take from other brands. So we don't we don't design anything. We just actually have a different type of like a products that ⁓ the main Zalando we call a fashion store, ⁓ they don't don't have, which is ⁓ homeware. So our our ⁓ units specifically work with homeware, with actually the Zalando main brand doesn't work with homeware, only fashion and accessories. speaker-0: I think this really interesting because it puts so much emphasis on creative direction and storytelling, because you are not you're not creating product to a brief. What you're doing is you are communicating product to a specific audience. And the communication and the storytelling is all of it. Like it's not, it's not wrapped up in new product innovation and introductions and everything else. It is exclusively a storytelling and creative exercise, which I think puts a lot of importance on your role. speaker-1: Absolutely. I mean, I think this needs to be very fleshed out. ⁓ the Creative Direction actually it's it's ⁓ it's what makes us recognizable and then creates differentiation from other brands. Because the product itself, I mean, we we we cannot focus on on the technical part of each product or different products. We need to we work in a more category base. ⁓ and then that's why in in Creative Direction needs to create a this ⁓ this layer of ⁓ that bring attention to to our brand speaker-0: Perfect. So for the second of the question, so we've done your day to day for the definition. I want to dip straight into the latest campaign you did for Lounge, the latest at the time of this recording, at least, which was that was the reason I reached out to you. That was the catalyst for this conversation. The strap line for that campaign is quote unquote summer to feel. And I want you to tell me what you think it means to feel something when you're creating or looking at or interacting with. Campaign that is, depending on your workflow, which we'll get to, partially or wholly generated by AI. Now, you can spin that summer feeling as a vibe. I g I get that. ⁓ but no generative model has ever felt that poolside vibe for itself. You can spin it as the tactility of the fabrics, the feel there. But again, the model has never touched linen or so on. So when you're directing or running evals or r retrospectives on this kind of campaign, tell me tell me how it felt and what you think. what feeling means when you're working with partially or wholly synthetic media. speaker-1: Yeah, I mean I think the whole the whole challenge of like the summer to few is is because the our campaign, our previous campaigns, they were they were very entertaining. ⁓ they were, I mean, we did like two campaigns like the with AI before, and then they were more focused on the storytelling and focused on the entertainment part. This one, our challenge was like actually to really focus on the how AI could interpret the product, the feeling of the product. So summer campaign, it's also something that is kind of like it's all about feelings and and ⁓ in a more abstract way. But our goal was actually to create a process or to focus on the tactile or the the feeling of ⁓ to really that AI to bring the the details of the product a little bit more to the forefront and let the entertainment a little bit in the in the ⁓ passenger seat and then have the the textile feeling in the driver's seat. So we we work with different categories, ⁓ of course swimwear, ⁓ other types of textile and then we we ⁓ briefed our agency to actually take advantage of the latest kind of like models and then try to to really focus on bringing this this tactile feeling to the images. So that's what's actually our goal. speaker-0: And do you was that tactility, was that that change from entertainment to tactility? Did that come about just for internal reasons because you thought, Okay, it's time to try something different? Or did that come about because you think the generative models are more capable of capturing and conveying that tactility than they used to be? speaker-1: I think both reasons. I think the I think for us was actually to to see what actually we could change in the process to bring this ⁓ to at the challenge of changing or tweaking the process to bring this more tactileity. And in the same way is to push where the models could go in terms of bringing almost close to reality. So actually both ⁓ both ways we're we're where we are trying to are trying to go. speaker-0: Now, I'd like you to walk us through the process kind of step by step and tool by tool that went into creating this campaign. Now, you mentioned you partnered with an external agency, which I think is called Copy Lab. tell me how the work was broken apart across two human teams, your internal one and your external partners. And what I'm imagining is a spectrum of different generative image models across the full funnel. So what models are you using? How are you balancing the work? What workspaces are you all collaborating in? How much physical photography did you start with? either on location or of the products. Did you start with any? Just walk walk me through how this worked. speaker-1: Yeah. So we started like by actually concepting campaign, the concepting the campaign internally in with our in-house team. So then I mean it's the normal process of concepting. So we use use Gemini ⁓ for for concepting here because it's the tool that actually we have a partnership at Salando. So all the Google products are available to us. So we use Gemini for for concepting. So of course I I briefed ⁓ I briefed my team and then they kind of like came up with like different ideas in in how how we could kind of run the campaign. So after the generation of the idea and the approved of the main concept, which was like summer to feel, we won we we went to kind of like mood boarding. So the mood boarding we we were using kind of a classical photography, but the goal we wanted to to to to to reach is the is to detectality, right? Like the that to the summer to feel. So then we we create a mood board that actually could express that. And this mood board could be AI generated images that we we found to be interesting or could be normal photography. So we're a mix. And then with that we created the concept, we created the mood board and first we aligned with our in-house stylist. ⁓ Why? Because to give an idea that like okay, it's possible that we can kind of ⁓ we can reach to to that concept. And after the f the in-house style is vetted, they'll say, okay, we can it's something that we can find, these products we can find in in internally at Salomba, right? Because does it make sense that like a create a mood board with products that to doesn't exist here for us. So we we actually source these products in internally ⁓ as ⁓ SKU, right? So then after we have ⁓ the concept, the mood board, and we have we usually the the styling process also we do creating the looks and cre and ⁓ let's say prototyping the looks in an intern in an avatar. ⁓ in a lo-fi in a lo-fi way just to see if the looks that if the the SKUs that we we we want we picked ⁓ inside the website they actually make sense or they look cool. And with this we create actually this is an interesting process that we learned afterwards. We create a actual fitting ⁓ a fitting real fitting with ⁓ in ⁓ with the in-house with our in-house studio here. That we that we ⁓ invite like one model, ⁓ a simple model or a man or a w a man and a woman to actually try this real product. So we actually source, we we do a little bit of a hybrid process here. And why we why why we we do that? Because in this process of fitting, real fitting, ⁓ which is kind of like quite cheap. Because we have a studio here, we have the models available because they the is part of like the the studio here. We take pictures of 360 ⁓ of the model for each look, and we take pictures ⁓ with phone, okay, with with the close-ups in apart from apart from the SKUs, these pictures. are handing or handed over to the agency. And the agency will get from us the concept, will get from us the mood board, we get from us the SC use of the look, and we get from us the fitting pictures. Okay? The fitting pi so we we discovered that the fitting pictures helped in the prompt to make the prompt better, right? adding these pictures, adding these inf extra information apart from the SQUs, because it gives the the models actually how the ⁓ the information, how the product will blend and fit the the ⁓ into the body of of of the the avatar that we'll create afterwards. So then this we hand over to the to the agency or to the studio AI Studio and then the I the AI Studio come back to us after one, two weeks with some proposals. ⁓ that's that's a little bit the process that we have. speaker-0: I have one follow-on question for that, which why work with an agency? Why hand off that last part? Why not do that last part internally? speaker-1: ⁓ because they we don't have capacity internally to to do that. This is one question. I mean, also because there's ⁓ I mean this ⁓ the my R directors ⁓ that ⁓ they are also following different campaigns, right? ⁓ this is ⁓ usually our R directors work in t two or three campaigns simultaneously. There's not enough time in our timeline. To have like an art director or a ⁓ director prompting in back and forth ⁓ all this all this ⁓ alone. so my art director actually managed two two AI campaigns at the same time, at least, ⁓ with different with different ⁓ AI studios. Plus we have also ⁓ a work stream of like non-AI campaigns that also they need to kind of like maybe support or something like that. So then the why is that because of like the timelines, right? Like so ⁓ I have like the the art directors, they are following in two or three different projects. So does it make sense to kind of stop one art director to prompt that? and I mean there will be a lot of back and forth and then so on. So this requires a lot of time. So we prefer to be able to kind of like execute this in the highest quality possible. We give this specifically prompt execution to an agency, and then they ⁓ they can come back with ⁓ with ⁓ with some the execution for us. However, we give It's not that like we di their role is just prompting what actually we already previously execute or or created. We give them some freedom to interpretate our briefing, right? So in terms of c in terms of casting, in terms of location, in terms of light. I think of course we have as a brand we have a specific creative direction that we share with them as well. Our brand book, how a light how light looks like, ⁓ how our casting looks like. So this we share as well, our brand book for them to give an idea. But we we actually reinforce to them that would be great that this actually ⁓ prompter or or this AI artist ⁓ also kind of like come with some suggestions for us. speaker-0: I think that's interesting because I think externally, for people kind of outside of this workflow, there will be the perception that AI is the more hands-off, easier, more streamlined sort of route. ⁓ and I think you the way that you're describing the workflow, the major thing that impresses upon me about it is that it still involves a lot of people. It still involves a lot of work. It still involves it is not just prompting. It is not just basic prompting. What you are describing is a traditional campaign workflow with multiple stakeholders, interpretation, people's different perspectives. The tools are different, but that's it, as far as I can tell. speaker-1: Exactly. I think that's that's I'm I'm really our team really wanted to go in this direction because I think the collaboration of different creatives is what makes the work grow and my what makes the work better. So if I l if I just drop if I would just drop this in the table of the art directory and then the art director and have to kind of figure out casting, have to figure out ⁓ prompting the background and have to figure out the the the the ⁓ the styling. And this will be just one person vision and then that's that's that's I think is ⁓ in my opinion limited, right? And what actually the what actually we try is to Right? So is actually involving and then in this in this feedbacks, the copywriter is involved as well, the ⁓ the the ⁓ the stylist is involved as well, ⁓ me and the art director, and then we ⁓ and then the there's ⁓ also the agency. So we we just kind of like in in this check-ins that we have with agency. We brainstorm in the meeting and say, okay, this is going in the right direction. This maybe is not going the right direction. Or the the c the the the closing is not looking good, or this looks a little bit kind of so not not real in or the mood is not right. So there's a constant there's a constant collaboration in between the check-ins that we have ⁓ following the timeline. It's not only one person dictating the the the the ⁓ the way, right? So in and that's the that's was our choice, deliberate our choice. speaker-0: Fascinating. Now, how do you feel and what's your perspective on disclosure and labeling here? So the final deliverables after the workflow we've just talked about are touched by AI to a greater or lesser extent. They are generated or at the very least enhanced. You've been clear, this is an AI-native campaign. And I think that's right. I applaud you for coming on, talking about it, being transparent about the workflow. You're doing this at a time when the market feels Quite conflicted, I think, about having either AI generated or AI enhanced images put in front of them from a consumer point of view. Trust is is hard won. And there's this perspective that people do not trust AI images. And there's also a few different looming kind of potential regulations around disclosure and labeling. So what what's your perspective on what it means to put AI content out into the market right now? speaker-1: I think ⁓ I mean for any AI contact ⁓ I think is if it's valuable, if if if it's valuable for the consumer, if it's still inspirational or entertaining, I think honestly, I don't think that they care. If it's something that ⁓ actually gives some information, if we are curating something from our platform to them in a way that they can take advantage of it, or they it's something that we are educating them, or it's something that we inspire in them or or entertaining them. I think if he's AI, if it's not AI, I think honestly they don't care. I think what actually is is is they might be annoyed if is kind of like something that is not thought through, right? It's not intentional. Or ⁓ something that is is is j is just kind of content for the content's sake, you know? And then instead of like, okay, is this something that it's interesting for them? Is something new? And then AI, I mean, if we go if we if we take this as a as a as a ⁓ as a goal, I think I'm sure that like our our audience will like it and then they will appreciate it. So and even if we labing ⁓ labeling and that AI, I think that would be fine for them. As soon as we we always try to surprise them with the content that we have. And then it just it's just came up being AI. But ⁓ you know, I think it's it's the thought before that actually matters. speaker-0: Yeah, and I think, you know, again, to your credit, ⁓ talking about the workflow and all of the different people involved and all of the different stages here. I think the reason people dislike AI is because they think it's used in a lazy way. Like I I feel like I feel like that's the reason people rebel against it. ⁓ there there are other you know, there are other ethical and environmental concerns or what have you. But generally speaking, people will say, I don't want to see an AI campaign because I feel like it's inauthentic or it hasn't doesn't have enough people thinking about it. It's just a shortcut. ⁓ and that kind of brings me to the next question, which is I think the reason that people have pushed back against AI campaigns in the last year or so, which we've definitely seen, is that concern that it's either a quick fix or that it's definitionally only able to repeat and recombine what's gone before. Right. So the AI output is not going to be innovative, it's not going to be new, it's not going to be, to your point, inspiring because it's based on what it's trained on. Now you spoke at The TED talks in Berlin a couple of weeks ago, ⁓ and you said that AI is only as good as the brand framework behind it, which to me sounds like an answer to this, right? It sounds like saying, well, you can get good AI output if you build the right brand framework to create that. Tell me what that means and tell me if you think AI is capable of creating something fresh despite being pre-trained on existing data. speaker-1: ⁓ yeah, so I think in terms of brand, so I think I think the big foundation for any any brand that wants to work with AI is actually to have to have a good brand framework already done. I think we were lucky to kind of like create to to be build these campaigns when we already have our brand framework very well rounded, very well shaped and with an idea of kind of like trying to surprise and entertain our our customers in terms of ⁓ ⁓ creative direction. So ⁓ with that that being said, so I think if you it it depends how intentionally you want to use the tool. Right. If for example, if you for us, we already know where we want to go because our brand framework already tells us where we want to go. So one example, for example, we have one ⁓ One word that define our brand characteristic, which is called ⁓ soft surrealism. What actually that is actually soft surrealism is actually you get ⁓ we use like we get a scene, we want to build a scene, but we we tweak that scene to be something there that is a little bit surreal. This is a this is a framework that we created as a brand. So the When we when we we create the campaign, we try to in to ⁓ infuse this this DNA of the brand into the prompt, right? So why? Because we we try to create every every prompt that we infuse this type of like creative direction there, or we ⁓ we have this as a goal, we try to escape of the of the sameness of like what actually if you prompt there. A very vague, a very vague kind of like prompt that like just to create a scene, it will we'll kind of like just come up with something that very maybe every brand can can have that same image. But if we if we if we add this specific brand DNA, this specific brand traits into the prompt, which is unique to our brand, then there is a chance that. like this image is becoming more unique. It's coming, it's coming. So the question is like, what is the unique ingredient of your brand that can be added to the prompt? Right? What actually the specific traits that intentionally I can brief an agency and then they can kind of add this ingredient is a special sauce, right? That can be added in this recipe, right? And then that's that's actually where ⁓ I mean what actually different I can create, right? So if I if I'm an entrepreneur now, listen to this to this talk, what actually special I can create for me that no one can have? And then how I can infuse this in the process that I use to create AI images or AI videos. This is a distinction. So and then this is this is ⁓ help AI to create your unique your unique signature, right? And then this with this you can escape the the sameness that like they can they can generate in most of the cases. speaker-0: But okay. Question with this one. So your your remit with this campaign and others is global. These are global campaigns. Do you do any regionalization for them? ⁓ so I mean you o you obviously have a single hero direction ⁓ and that gets manifested across the products and the human models that make up the assortment. Do you do any like country by country variation in how you present this? speaker-1: Absolutely. So this is actually something that the AI is can help us a lot, not in terms of content creation, generation, but in for example, we have 21 countries that we have to localize. ⁓ so ⁓ what actually AI helped a lot in voiceover generation. So we have a copywriter that is following the whole campaign, and then ⁓ the copywriter needs to create ⁓ they create the master copy. And with Gemini, we localize this copy for 21 countries. Then with ⁓ AI voiceover 2, I mean we I can name here, like we use 11 labs, and with 11 labs we create a process that with a copy for each, we create a session ⁓ for each ⁓ each language. So We go we ⁓ and then in this session we we actually simulate a voiceover recording, right? So the art director, the copywriter, and a local expert from our company here meet for half an hour and they craft in that half an hour the voiceover that needs to be executed. So ⁓ we have more or less ⁓ we take a week to do this voiceovers. And then we we record like more or less ⁓ I don't know fifteen to sixteen different languages because some countries they have the same language, right? So we do 15 sessions of half an hour each. And then in that session, we prompt the voiceover and the voiceover and the the voice expert validate if that if that ⁓ copy is is is it's it's true, right? If that voice is ⁓ it's it's it's sounding ⁓ genuine. And then with that we we ⁓ we pass to our internal team that design that then ⁓ create the videos ⁓ infuse the overlay copy in the the voiceovers in the videos and then we generate a batch of videos in different formats ⁓ also with AI. And then we end up with multiple multiple ⁓ executions for all the twenty one countries that we have. speaker-0: Okay. And again, that sounds like it's an intentional process with multiple people involved. It's not a turnkey, push a butt push a button, give me this in French or give me this in Italian. No, okay, fine. tied to that previous question, what what's your perspective on the impact of this on the human modeling industry and on the talent in front of the camera? Because if you're doing this sort of regionalization, you're you're doing campaigns that appeal to multiple demographics and multiple regions, you're also looking at representation and inclusivity in casting. and you're also looking at kind of diversity in front of the camera, everything else. What impact do you think this workflow has on that? So you mentioned you have one kind of beginning fit model for this, a human fit model who's part of the content studio at Salando, who comes in for this. Then presumably the models that we see in the campaign are all generated rather than real people. Is that the correct read? speaker-1: Yeah, is that speaker-0: Okay, and what do you think is the future for kind of it it's a difficult market right now for anybody whose job is to stand in front of a camera and to embody a collection? speaker-1: Yeah, I mean the big disruption that we see ⁓ I mean it would be in the modeling in and modeling industry or the agencies. In for example, in this process, for example, in voiceover recording. So we are ⁓ we ⁓ I mean in the campaigns, this this particular two two classic ⁓ processes, they they will be replaced, you know, and that's the truth. What actually I mean what actually ⁓ opportunities have is like if we if we if we want to create something special or I think something that requires acting. I think ⁓ I think this is something that that still ⁓ will be would be required to have models. ⁓ I mean we have some campaigns that we do here for that is a little bit more into feeling into into kind of a more ⁓ shaping a little bit the brand feeling, then we don't use AIs for these campaigns. Because I think like these campaigns require some nuances on interpretation that is at least now, is pretty hard to get into with the with the with AI. Mostly in speaker-0: I was I was gonna say though, do you do you see that changing over time as the capabilities of the models increase? speaker-1: I think so. I think so. I think I think also is is is also the expertise of like the the AI artists on the prompter as well into how to trans transmit this feeling, right? but these campaigns that ⁓ that I just mentioned now, the campaigns that I just mentioned now, they are campaigns that are a little bit more focused on the product. They are more focused on the fashion, ⁓ fashion inspiration, right? When when at least At least now when we are going to talk about kind of like brand campaigns where we we require a little bit more acting, these ones we are not using AI because ⁓ we don't think that like ⁓ is there yet. So but maybe in the future we'll change. speaker-0: I think that's an honest answer. when you there's there's two different kinds of frontiers for generative images here. One is the campaign stuff that you're talking about, which is your remit. And then the other is the broader scope of e-commerce content. So what you would class as like product detail page images and video and so on. a lot of brands are pursuing this for unit economic and efficiency reasons. In your case, because these are pre-existing Zelando products, have they already had kind of PDP page content done using either traditional photography or generative photography. So you don't need to revisit that. speaker-1: ⁓ I mean, particularly my team doesn't doesn't ⁓ is not involved on the in in in commerce because it's the at least here is separated. ⁓ we have a team that that ⁓ take care of like everything that is ⁓ we call it on-site, right? PDP images, banners and so on. There's a specific team that takes care of that. And I think it's well known in the market that like Zalando is kind of like go moving into into into make almost 100% of this content with AI. Okay, so I think this is already announced in the in the market. And then I mean ⁓ the studios that we have internally here are closing and then this will be generated. Not everything yet, but I think is they are working on this on this way. ⁓ So this is already happening, right? This is already happening. In particular our team is a little bit more when in this process that I just mentioned to you. We are we are ⁓ in charge more on the customer acquisition phase, which is kind of marketing, is kind of like AI that actually bring bring customers to the to the to the tool. So working closely with TikTok, with the Meta, with Instagram, or ⁓ Snapchat. So we created this AI campaigns that are going outside and bringing customer to us. speaker-0: So your KPI then is driving down the cost of customer acquisition, essentially, with something with something like this. Okay. And and I think again, that's that's I appreciate the honesty and the transparency because what you're trying to do in both functions, whether it's you know, e-com on-site images or these kind of campaign things. is to cut the time and the cost. In one case it's the time and the cost of the conversion and the purchase. In the other, it's the customer acquisition. And I think in your case you talked about cutting campaign time from weeks to to days and so on. speaker-1: Yeah, I mean I think we we need to ⁓ well as a lit need to be careful about I mean just to talking about costs because it's not the only variable that we are looking. Zalondo has a kind of like good reputation to kind of like deliver good quality and good content in the market. So it's not deliver cut costs at not cutting costs at all costs. So this kind of when we evolve a studio ⁓ that are very specialized in in AI. ⁓ is is is not cheap. We could do this process in a completely different way, as you mentioned before, just kind of having a ⁓ our director doing internally. But then I don't think that the result will be as good as we are having now with many people collaborating. So that's so we of we are ch ⁓ we are more interested in in change the way that we do campaigns to make it a little bit more technologic and in infusing AI in the process and in in in terms of curiosity, then actually cut costs at all costs. So we end up with less costs. But this could be even done cheaper, but we decided to not go in that route. Because then is what actually happened is to create the AI slop that actually will our customer will start rejecting what actually we put out. So we are trying to find a sweet spot where we reduce costs, but at some point we keep the quality high in in the process in a in a doable ⁓ in a doable way that we we can still deliver something that people appreciate. speaker-0: Do you think do you think there's a world where that same kind of cost reduction quality up? Is there a world where you you do more campaigns? Is there 'cause this is this is the thing that people would usually say is they would say, Well, okay, well yes, we use AI to reduce cost to some extent to reduce headcount, but we also use it to enable a multifunctional, busy team to do more. Do you feel that? speaker-1: ⁓ I think I mean what I what I the philosophy I think like here, I mean, ⁓ is that I would not move my team to kind of actually just replicate what they are doing, but it's to give it's to free them time to make our campaigns even more impactful. So ⁓ is not that to duplicate the output, is basically actually free them time to actually think the co to make the concepts even more impactful. And to have the mindset to say, okay, what actually could be doing that our competitor is not doing? Could we do some interactive things out of this? Can we can we do something different in this campaign? For example, instead of just throwing a campaign there in in in in our ⁓ partners like ⁓ you know ⁓ or channels is like could we do something interactive in this campaign that actually will engage more or ⁓ so the way that I would see freeing time or or ⁓ this this the AI campaigns will will will come is like how to make them actually even more powerful instead of just duplicating them to to make it more because also I think we we We have to be careful with the the fatigue of like actually do the with the content that we're putting out, right? Okay. And then I think is I would do I will with this free time that we we have now with that like ⁓ that ⁓ with the digital campaigns that they are maybe instead of three months doing a normal campaign, we are doing now in one month and a half and two months, is actually to have more time to think them through a little bit better and then add things that are unexpected. speaker-0: So if we take somebody with your time of role, you know, senior creative director and so on and see senior folks, ⁓ I think that all makes sense. You recently said, and I forget whether this was, but you said the best thing we can do for junior creatives is to give them space to experiment with AI before the rules are written. So if we think about somebody who might be joining your team, who might be joining some kind of equivalent ⁓ function within a similar-sized organization, what does that look like in practice? ⁓ 'cause I know there are there are students and people in education or people just entering industry who listen to this show. What would should they be doing and what should they turn to if they wanna start to prepare themselves to work in, you know, campaign focused functions that feel like they are very, very different to what they've been before? speaker-1: Yeah, so I think like I mean we something that we ⁓ that we do here with the juniors is like try to to make them part of the process that we are doing, but like ⁓ but more not ⁓ not actively executing the campaign by themselves, but maybe as a supporting role in some campaigns. And by understanding the process and what it takes through the process of the campaign, they are learning how the how what actually the process looks like. So this is one thing that like they understand ⁓ looking at the senior director and how the senior director is is having to kind of like collaborate with the copywriter, collaborate with the stylist, collabor collaborate with the creative director, collaborate with the agency. So they understand what it takes just by being the meetings, right? MB is supporting in the R directoring, collecting some mood boards and so on. This is would be one thing. So, but hands-on, what actually they they can be they can be is to be curious and then propose something that could be an add-on on this campaign. Or ⁓ I mean, we have also different projects that are not only campaigns here that we need to to to support internal stakeholders. For example, the CRM team, how we can kind of like create multiple images ⁓ from a comp from an existing campaign, prompting multiple images that can be that can be, for example, changing the background. ⁓ the CRM team here, for example, segment the segment the the campaigns that they have by region. How how I can ⁓ usually ⁓ a task that the juniors have is to from our already already created campaign, how can then they prompt multiple campaigns that have a different backgrounds? That for example, the same image can have a background that is made in London, that that brings a London vibe, ⁓ or a Madrid vibe, or like a Berlin vibe. So we can actually extend that image or build on that image and create different images that are from that could help the ⁓ the segmentation of the CRM. Not ever so not everything is campaign. ⁓ there's a lot of things here that like we could support stakeholders with AI executions that they that are they are less heavy for for or l less risky, and then the juniors they can practice in prompting and workflow in those projects. speaker-0: But you you do see that there is still a pathway for junior creatives. You don't see the door closing to them. speaker-1: I don't think I mean and and it's all about I mean it's it's it's a different ⁓ I ⁓ it's a different it's a different positioning. I think like there is there is room for them. I think it there's but also like it's it's more about them to be hungry into deep diving to the stools and then ⁓ and try to and try to to play around with them and support the team in an intentional way as well. ⁓ but in my opinion, I think they are still very valuable and then I think like we should keep I mean we should keep the stream of like of the creatives and bring the creatives across across this this this workflow. Otherwise ⁓ we'll not have ⁓ more creatives to work ⁓ with AI, right? I mean we need to to have some projects that ⁓ give them a little bit of like a chance somehow. speaker-0: Yeah. No, I think that makes sense. Let's finish with your own role, creative director. Give me two quick perspectives on what you think your job description might look like in twenty thirty, so four years from now. One of them, where let's say a hypothetical company doesn't have to be Zelando, has really figured out how to turn AI to its advantage. So if there is an AI-native creative director position by 2030, what does that look like? And then on the flip side, what does it look like for a company that doesn't engage with AI or doesn't learn the right lessons? speaker-1: I think like a the creative director I think needs to really understand ⁓ in my opinion what the process what the in this new way of like creating campaigns with AI, what actually the processes are valuable to take and which ones are actually not beneficial to the to the impact of the campaign. Is to is to really deep dive into the process and and and try to learn and then try to identify, okay, this piece here is we we can do faster, but is this make it is this make the campaign better? Right? Is this make the campaign so is is it really identify where where where are the pieces on this workflow that actually you you should skip or remove or change? And then in the end, you you are responsible for how impactful this campaign is, right? in how in and how ⁓ how optimize how how how you can optimize the process in a way that makes it faster and maybe even cheaper but still keep like the quality in a in a in a good in a good level. ⁓ This is one, right? Yeah. And then I think the second one, the other, the other ⁓ I think like in the other question, I think Brands, I think they should be jumping as soon as possible on finding their own way into create campaigns with AI or create content with AI. It's super important because it's like a really it's something that really can can scale and then scale in a in a good way. But I think, as I mentioned before, it's really important to for you to define your own vision because AI can help you, but If you don't have your own vision setup, your secret sauce, you start to kind of create content like that are not unique for your brand, and you're not creating differentiation between your competitors. So you I think both sides need to be hand in hand. Finding your own voice and jumping and jumping into putting this in practice as soon as possible. to to optimize your your campaign process or your content process. speaker-0: Yeah, the way I framed it before across disciplines is AI can make you better, faster, potentially cheaper at something you already know how to do and you already know how to discern good output from bad output. It's not gonna make you magically capable of doing something you didn't already know how to do before. ⁓ and I think that's key. So engaging with it gives you the technical grounding to do it. Figuring out what your vision and your secret source is is what distinguishes you from somebody who doesn't know how to do the job and is just trying to use AI as a shortcut. Exactly. Perfect. Glaciane, thank you so much for your time. I've really enjoyed this conversation. It's been honest, it's been authentic. ⁓ and I think there's a there's a lot to get out of it. So I really appreciate you joining me. speaker-1: It was my pleasure. Thank you very much for the opportunity. speaker-0: And that's the end of my conversation with Glacioni. A few times there, I do admire his honesty and his willingness to engage with every side of this conversation. You can tell, you know, he's a he's a passionate creator who sees a step change in the tools, and he's someone who's already internalized what that means. Now, you're entitled to feel however you feel about AI. The point of these kinds of conversations, our AI report, any of our sort of AI-focused coverage. Is not to change anyone's minds. But hopefully you come away from this episode with at least an appreciation for how some of the foremost practitioners of AI native creative processes are approaching it. And the fact that there's possibly more to it than you think. It's not just sticking in a prompt. Now, anyone who's listened to enough of these shows knows that I'm pretty down the middle when it comes to AI. Part of that is because my job requires some measure of objectivity, but part is that I can really see the value of generative AI and creative design and content and storytelling, but I can just as keenly see the losses being felt by the communities that have shepherded those disciplines for a century or more. You can say that makes me a fence sitter. I'm I'm fine with that. I'm open to that criticism. But given everything we've covered today and given just how much AI content you're going to see the next time you go online, I think having someone perched in the middle, sympathetic to both arguments, is a it's a useful thing. So for now, thanks for listening. A very different topic coming up next week, and I'll talk to you again really soon. speaker-1: Yeah.