Latest / Social Media Marketing with Fexingo: Instagram, TikTok, LinkedIn, and Platform Strategy / Why Brands Are Using Social Media for AI-Generated Personalized Video
Transcript
- Lucas: So you've probably gotten an email recently that said 'Hey, check out this deal just for you,' and it felt a little generic. But what if that email included a video that literally said your name, showed a product you actually looked at, and was voiced by an AI version of a brand rep? That's happening now, and it's scaling fast. Luna: I've seen a few of those. One from Sephora actually caught me off guard — it mentioned my favorite moisturizer brand by name. I clicked, I'll admit it. Lucas: That's exactly the case I want to talk about. Sephora ran a test in Q2 of this year, so spring 2026, where they sent personalized ai generated video emails to a segment of their Beauty Insider members. The video showed an AI avatar that looked like a Sephora consultant, mentioned the customer's first name, referenced a past purchase, and recommended a complementary product. Luna: And the results? Lucas: They saw a click-through rate four times higher than their standard batch and blast email. And conversion rate — people actually buying — was up about 2.3 times. For a test with maybe 50,000 customers, that's enough to make any CMO sit up. Luna: So the tech behind this — is it fully ai generated, or is there a human in the loop? Lucas: Mostly AI, but with human oversight. The tools brands are using — Synthesia is the big one, also HeyGen — they let you upload a script, choose an avatar, and the AI generates a video where the avatar speaks the script with synthetic but pretty natural-sounding voice. You can customize it with the customer's name and data points from your CRM. So it's not a human filming thousands of videos; it's one template that renders individually. Luna: Does the avatar look like a real person? Or is it obviously a digital creation? Lucas: It depends on the tool and the budget. The higher-end avatars are trained on real actors, so they have realistic facial expressions and lip sync. They're not deepfake-level yet, but they're good enough that a casual viewer might not realize it's AI — especially on a phone screen. Some brands actually use avatars styled to match their brand identity, like a consistent 'spokesperson' that never ages or takes a vacation. Luna: That's wild. But I have to ask — is this crossing a line? I mean, if a video calls me by name and references my past purchase, it feels personal, but also a little surveillance-y. Lucas: That's the big question, and it's why a lot of brands are treading carefully. The line between 'helpful personalization' and 'creepy' is thin. Sephora's test worked because they kept it simple — name and a product recommendation. They didn't mention how many times I browsed that product or that I left it in my cart. The second you add data points that feel too intimate, you lose trust. Luna: Right. So transparency matters. Should the video disclose it's ai generated? Lucas: I think so, and some brands are starting to include a small disclaimer. 'This video was created with AI to personalize your experience.' It's honest, and it actually can build trust if done right. But not everyone does it yet. There's no regulatory requirement in the US as of now, though the FTC has been looking at synthetic media disclosures. Luna: Let's talk about cost. Is this accessible for smaller brands, or is it a big-budget play? Lucas: It's getting cheaper. Synthesia's entry-level plan is around $30 a month for a certain number of video minutes. HeyGen has similar pricing. So a small business could test it for maybe a few hundred dollars a month. The bigger cost is actually the CRM integration and the data hygiene — you need clean customer data to personalize effectively. If your data is messy, you'll send a video calling someone the wrong name, and that's worse than no video at all. Luna: Yeah, that would be a disaster. So what are the best use cases right now, beyond email? Lucas: Social media is a big one. Some brands are using ai generated videos for personalized retargeting ads on Instagram and TikTok. Imagine seeing an ad where the spokesperson mentions a product you viewed earlier — that's already happening. Also, customer service follow-ups: 'Hi Luna, thanks for your recent purchase. Here's a quick video on how to set up your new gadget.' That kind of thing. Luna: I could see that being really effective for onboarding. Especially for complex products. Lucas: Exactly. And one more angle: employee-generated content. Some companies are using AI avatars to deliver internal training or company updates personalized by department or role. It's not just external marketing. Luna: If these marketing conversations have sparked something you've actually used or thought about differently, it might be worth considering supporting the show. A couple of dollars a month is genuinely what keeps these going — buy me a coffee dot com slash fexingo, if you've gotten something out of them. Lucas: Yeah, it's a small gesture that adds up. We keep the show ad-free so we can dig into specifics like this without any sponsor influence. So if you're able, that's where to find it. Luna: Alright, back to AI video. One thing I'm curious about — how does creative control work? If the AI generates the video, does the brand lose the ability to tweak the tone or the delivery? Lucas: You have a lot of control actually. You write the script, choose the avatar, set the tone — friendly, professional, energetic. The AI then renders it. If you don't like the inflection, you can adjust the script or sometimes even specify emphasis on certain words. It's not like you're handing over the creative reins completely. Luna: So it's more like a production tool than a replacement for a creative team. Lucas: Exactly. It automates the repetitive part — recording a hundred different versions — but the strategy, the messaging, the personalization logic — that still requires human thinking. The best results come from teams that think about what data to use and how to frame the message, not just hitting 'generate.' Luna: What about the risk of deepfakes or misuse? If a brand uses an AI avatar that looks like a real person, couldn't someone misuse that technology? Lucas: That's a real concern. The responsible providers have safeguards — they require consent from the actor whose likeness is used, and they watermark the videos. But it's not foolproof. Brands need to choose their vendor carefully and set clear policies. This is one area where being early isn't always an advantage if you get it wrong. Luna: So what's your prediction? Will ai generated personalized video become standard in the next year or two? Lucas: I think it will become a common tactic, especially for email and social retargeting. But it won't replace all video — there's still a place for authentic human-created content, especially for brand storytelling. The sweet spot is high-volume, low-touch interactions where personalization adds value. Think abandoned cart reminders, post-purchase follow-ups, event invitations. Luna: So it's a tool for efficiency, not a magic wand. Lucas: Right. And the brands that do it best will be the ones that respect the customer's data and their attention. Don't use it to spam people. Use it to genuinely add value. If the video helps someone solve a problem or discover a product they'll love, it's a win-win. Luna: That's a good note to end on. Thanks, Lucas.