Latest / The CMO Podcast with Fexingo: Marketing Leadership, Budgets, and Executive Strategy / How CMOs Use Generative AI for Personalized Video at Scale
Transcript
- Lucas: So earlier this week, I was watching a TV ad for a major airline — I won't name which — and the whole spot was clearly generated with AI. The lighting was too perfect, the voice had that slight unnatural cadence, and I realized: we've crossed a threshold where generative AI is no longer just a tool for static images or text. It's now producing full-motion video ads that brands are running on national television. Luna: And consumers are starting to notice. There was a survey from earlier this year — I think it was from the Interactive Advertising Bureau — that found 62% of viewers could correctly identify an ai generated video ad when prompted. So brands are walking a tightrope between efficiency and authenticity. Lucas: Exactly. And today I want to focus on one specific application that's really taking off in the back half of 2026: personalized video at scale. Think about this: Coca-Cola, this summer, launched a platform called 'Create Real Magic' where consumers could upload their own photos and use generative AI to create personalized video clips featuring Coca-Cola's brand characters — the polar bears, Santa, the contour bottle. Luna: Right. And that's B2C, consumer engagement. But I've also seen B2B companies doing something similar. HubSpot, for instance, started using ai generated video avatars that look and sound like their sales reps to send personalized demo follow-ups to thousands of prospects. Each video is unique to the recipient's company name, industry, and pain points. Lucas: That's the key shift — personalization used to mean inserting a customer's first name into an email subject line. Now it means generating a 60-second video where the actor on screen speaks directly to that customer about their specific problem. And the cost differential is staggering. Luna: What are we talking? Like order of magnitude? Lucas: A professional branded video with an actor, crew, editing — you're looking at $50,000 to $100,000 per finished minute. With generative AI, you can produce a high-quality personalized video for under $500 per version. And once the model is trained, generating 10,000 variations costs almost nothing beyond compute. Luna: And the metrics back it up. I saw a case study from a company called Synthesia — which is one of the leading AI video platforms — where a financial services client used personalized AI videos for their outbound sales. They saw a 3x higher click-through rate compared to static emails, and a 40% increase in meeting bookings. Lucas: So the unit economics are compelling. But there's a creative tension here. If everyone uses the same few models — Synthesia, Runway, Pika — don't all these videos start looking the same? There's a risk of what you might call 'AI homogenization' of brand creative. Luna: Yeah, I think that's the number one concern I hear from CMOs. They're excited about the efficiency, but they worry about losing brand distinctiveness. How do you solve for that? Lucas: The smartest approach I've seen is using generative AI as a starting canvas, not the finished product. Some brands are training custom models on their own brand assets — logos, color palettes, approved typography — so the output stays on-brand. Then a human creative director oversees a final pass, maybe adds a live-action element or a custom soundtrack. Luna: That hybrid model seems like the sweet spot. You get the scale and speed of AI, but you preserve the human touch that builds trust. And you avoid the uncanny valley problem. Lucas: Let's talk about the uncanny valley for a second. It's real. A recent academic study from Stanford's Virtual Human Interaction Lab found that when viewers detected even subtle AI artifacts in a video — like slightly asynchronous lip movements — their purchase intent dropped by 18% compared to a human-created video. Luna: So authenticity isn't just a nice to have; it directly impacts conversion. That's a powerful data point for any CMO trying to justify a larger budget for human oversight. Lucas: And then there's the legal side. Copyright law hasn't caught up. If a model generates a video that inadvertently resembles a copyrighted character or uses a voice similar to a celebrity, who's liable? The brand, the platform, the agency? The SEC even issued guidance earlier this year reminding companies that ai generated content in investor communications still falls under fair disclosure rules. Luna: That's a huge risk. I know several CMOs who have held back on deploying generative video precisely because their legal teams can't get comfortable with the liability. Lucas: So what's the pragmatic path forward? I'd argue that 2026 is the year of the pilot. Don't bet the whole Q4 budget on generative video. But run a controlled test on one campaign, with clear guardrails: only use approved training data, include a disclaimer that the content is ai generated, and measure the brand lift versus a control group. Luna: That makes sense. And speaking of pilots, I love what the team at Unilever did with their Dove brand. They used generative AI to create thousands of variations of their 'Real Beauty' campaign — but they trained the model exclusively on images of real women, not synthetic faces. So they upheld their brand purpose while getting the efficiency gains. Lucas: That's a brilliant example because it shows that the technology itself is neutral — it's how you apply it that matters. Dove turned a potential threat to authenticity into a reinforcement of their core message. Luna: If these marketing conversations have sparked something you've actually used in your own strategy, you're exactly the kind of listener who keeps this show going. A handful of people chip in monthly through buy me a coffee dot com slash fexingo, and that's what lets us keep producing these deep dives without any ads. Lucas: Yeah, it's a small group, but it makes a real difference. And we're grateful for that support — it means we can stay focused on the substance rather than chasing ad revenue. Luna: Exactly. So to bring it back to generative video: the CMO's role is evolving from chief creative director to chief curator and quality controller. You don't have to make every frame yourself, but you have to set the standards for what 'good' and 'on-brand' look like. Lucas: And that's actually a more strategic job. If you're spending less time in production meetings and more time defining creative guardrails and interpreting performance data, you're probably adding more value to the business. The question is whether your team has the skills to manage that shift. Luna: I think that's where we'll see the biggest talent gap over the next twelve months. Not data scientists — prompt engineers and creative technologists who can bridge the gap between the algorithm and the brand voice. Lucas: So to sum up: generative AI video is real, it's scalable, and the economics are undeniable. But the winners won't be the brands that adopt it fastest. They'll be the ones that adopt it smartest — with training data discipline, human oversight, and a clear metric for brand equity. Luna: It's a tool, not a strategy. And the best CMOs I know are already thinking about it that way.