Latest / The CMO Podcast with Fexingo: Marketing Leadership, Budgets, and Executive Strategy / Why CMOs Are Using Generative AI for Ad Creative
Transcript
- Lucas: So the question we are hearing from marketing leaders right now is less 'should we use generative AI for ad creative' and more 'how fast can we scale it without wrecking the brand.' Luna: And that's a real tension — speed versus quality, or maybe speed versus distinctiveness. Lucas: Exactly. I want to anchor this conversation in a couple of specific cases. The first one is Coca-Cola. In early 2025, they launched something called 'Create Real Magic' — a generative AI platform that lets creators and even consumers generate ad assets using Coke's brand assets. They've since rolled it out globally. Luna: And that was a pretty big bet. I remember reading that they cut production time for certain campaigns by something like 60 percent. Lucas: That's right. 60 percent reduction in time from brief to final asset. And the cost savings are significant too — one estimate I saw was around 30 to 40 percent reduction in production costs for digital ads. But here's the thing: they didn't just hand the keys to an AI model. They trained it on decades of Coke advertising, so the output feels like Coke, not generic AI slop. Luna: Right, so the brand consistency is baked in from the start. But that requires a huge upfront investment in training data and guardrails. Lucas: It does. And not every CMO has that luxury. But the ones who are succeeding are the ones who treat generative AI as a creative partner, not a replacement. Another example is Mattel. They used generative AI to create concept art for new toy packaging and even some ad visuals. They reported a 50 percent reduction in concept to production time for packaging. Luna: Fascinating. And that's a very controlled use case — packaging is highly regulated, needs specific legal copy, safety warnings. AI can handle the visual generation, but humans still approve every final file. Lucas: Exactly. The human-in-the-loop model is table stakes right now. What's interesting is how CMOs are rethinking their agency relationships because of this. If you can generate 50 ad variations in an hour internally, do you still need a full-service agency for production? Or do you shift to a strategy-only retainer? Luna: That's a painful conversation for a lot of agency CEOs. But I've talked to a few CMOs who say they're now hiring 'prompt engineers' as contractors instead of paying agencies for production. It's a real shift. Lucas: It is. And I think it ties into something we've discussed before — the move toward in-house creative studios. Generative AI accelerates that trend because the production barrier drops so much. One CMO at a mid-size consumer brand told me they went from 15 external agency partners to 5 in the last 18 months, largely because AI handles the grunt work. Luna: But there's a risk, right? If everyone uses similar AI models — like dall e, Midjourney, or Stable Diffusion — won't all the ads start to look the same? Lucas: That's the big concern. And we're already seeing it. I saw a study from earlier this year — I think it was from the Association of National Advertisers — that found 40 percent of consumers can now tell when an ad image was generated by AI, and about half of those said it made them trust the brand less. Luna: So there's a trust penalty. That's a serious risk for CMOs who chase efficiency without thinking about brand equity. Lucas: Yeah. And that's why the best practitioners are using AI for what it's good at — generating lots of options, A-B testing variations, personalizing at scale — but then layering on human creative direction to add the unexpected, the emotional, the culturally relevant element that makes an ad memorable. Luna: So the CMO's job becomes more about curating and less about originating? More about taste than raw creation? Lucas: I think that's exactly right. The CMO becomes the chief taste officer, essentially. They set the creative brief, train the model on brand parameters, and then select from ai generated options. It's a different skill set. And that's causing some tension in leadership teams — not everyone is comfortable ceding creative control to a statistical model. Luna: If this conversation has sparked something you've actually started thinking about in your own marketing — maybe you're evaluating AI tools or rethinking your agency roster — the way we keep the show ad-free and focused on these kinds of practical topics is listener support. Lucas: Yeah, that's right. It's entirely funded by people who find value in it. You can chip in at buy me a coffee dot com slash fexingo. No pressure, just if it's useful to you. Luna: And we genuinely mean that. It helps us keep doing deep dives without commercial interruptions. So back to the creative side — I think there's another angle here around measurement. Lucas: Great point. Because one of the promises of generative AI is that you can create hundreds of ad variations and then test them in real time. But most brands don't have the measurement infrastructure to actually learn from that. You need a robust attribution system to know which ai generated variant drove the conversion. Luna: Right. Otherwise you're just generating noise. I've seen brands that created 500 versions of a Facebook ad and then only measured clicks — they saw no lift because they weren't controlling for audience segmentation. Lucas: Exactly. The CMOs who are getting real ROI are the ones who combine generative AI with a proper experimentation framework. They run controlled tests, measure incrementality, and then feed those insights back into the model. It's a learning loop, not a one-shot production line. Luna: So it's not just about the technology — it's about the process around it. And that probably means internal teams need new roles. I've seen job postings for 'AI Creative Strategist' and 'Generative AI Production Lead'. Lucas: Yeah, those titles are popping up more and more. And they're often hybrid roles — part creative, part data analyst, part technologist. That's hard to recruit for. But some CMOs are upskilling existing staff. One example: a large CPG company I talked to trained their junior art directors on Midjourney and saw a big jump in their output quality within three months. Luna: That's smart. You keep the institutional knowledge and just add the tool. But what about legal? If the AI generates something that inadvertently copies a copyrighted image, who's liable? The brand, the agency, the software vendor? Lucas: That's the million-dollar question. And it's still largely unsettled. Most major brands are using indemnified enterprise versions of these tools — like Adobe Firefly, which is trained on licensed data — to reduce risk. But if you're using open-source models, you're on your own. I've heard of at least two brands that had to pull campaigns because the ai generated background included a recognizable building that they didn't have rights to. Luna: Ouch. So the legal department needs to be looped in early. That's another layer of complexity. Lucas: Absolutely. And it's not just copyright. There's also brand safety. If the AI generates a version of your ad that puts your product next to something offensive in a user-generated context, that's a reputational risk. CMOs are building in human review checkpoints for that reason. Luna: So we've talked about cost, speed, quality, legal, and measurement. Is there a consensus yet on whether generative AI actually drives better business outcomes, or is it still mostly efficiency play? Lucas: I think the honest answer is that it depends on the category and the maturity of the brand. For direct-response e-commerce brands, the efficiency gains are clear — lower cost per acquisition, faster testing cycles. For luxury or heritage brands, the risk of diluting brand equity is higher, so they're moving more cautiously. But the overall trend is unmistakable. A recent Gartner survey said 70 percent of CMOs plan to increase investment in generative AI for creative production in the next 12 months. Luna: That's a huge number. And it's not slowing down. So what's the one piece of advice you'd give a CMO who's just starting to explore this? Lucas: Start with a narrow, high-volume use case where you can measure the impact. Don't try to reinvent your entire creative process overnight. Pick one channel — maybe programmatic display or social — and run a three-month pilot. Measure time, cost, and creative performance against a control group. Then scale from there. Luna: That's practical. And it avoids the 'shiny object' trap. Let's keep following this — maybe we can check back in a few months and see how those pilots turned out. Lucas: I'd love that. For now, I think the takeaway is that generative AI is not a fad in marketing. It's a structural shift in how creative work gets done. The CMOs who treat it as a tool to augment human creativity — not replace it — will be the ones who win.