Latest / Marketing Made Simple with Fexingo: Branding, SEO, and Digital Strategy for Beginners / How Brands Are Using AI-Powered Personalization in 2026
Transcript
- Lucas: Luna, I want to talk about something that is quietly reshaping how brands talk to customers. It is not a new social platform or a viral ad format. It is personalization at scale, driven by AI, and the bar has moved dramatically just in the last eighteen months. Luna: You mean the kind where a brand actually knows what I want before I type it? Or is it still just 'Hi first name' in an email? Lucas: Way beyond that. We are now at a point where machine learning models can process a user's past purchases, browsing time, click paths, even mouse hover duration, and serve a unique homepage, a unique product recommendation, and a unique email sequence — all in real time. Luna: That is a big leap from the 'customers who bought this also bought' era. But do consumers actually want that level of intimacy from a brand? Lucas: The data says yes, but only if it feels helpful, not invasive. A McKinsey study from last year found that 71 percent of consumers expect companies to deliver personalized interactions, and 76 percent get frustrated when they don't. The tricky part is execution — too many brands cross the line into creepy. Luna: Let's talk about a brand that does it well. Spotify's AI DJ feature — I have been using it for months, and it is spookily good at picking songs I actually want to hear. Lucas: Spotify's AI DJ is a textbook example. It uses a combination of your listening history, your playlist curation habits, and real-time context like time of day and day of week. It even speaks to you in a synthetic voice that sounds like a real radio host. The result is a personalized radio station that updates constantly. Lucas: Spotify reported that users who engage with the AI DJ stream 25 percent more music per session on average. That is a massive engagement lift, and it directly impacts retention and subscription revenue. Luna: Twenty-five percent is huge. But Spotify has the data and the engineering muscle. What about a brand with fewer resources? Can a mid-size e-commerce company pull off similar personalization? Lucas: Yes, and the barrier has dropped significantly. Tools like Dynamic Yield, Bloomreach, and even Shopify's built-in AI personalization features allow smaller brands to segment audiences into hundreds of micro-cohorts and serve tailored content without a dedicated data science team. Lucas: Take a brand like Brooklinen. They use AI to personalize product recommendations based on browsing behavior and past purchases. If you looked at linen sheets but bought percale, the next time you visit, they show you percale accessories, not linen. Simple tweak, but they saw a 12 percent increase in average order value. Luna: That makes sense. But I think the line between helpful and creepy is really easy to cross. I once searched for a vacuum cleaner on one device and got vacuum ads on every single platform for two weeks. That is not personalization — that is harassment. Lucas: Exactly. And that is where the 'retargeting fatigue' comes from. The best personalization is contextual, not just sequential. It understands intent, not just identity. Sephora's virtual artist is a great example — you upload a selfie, and the AI applies makeup looks to your actual face in real time. It feels like a service, not a sales pitch. Luna: And Sephora has the numbers to back it up. I read that customers who use the virtual artist are 2.5 times more likely to purchase, and they spend 10 percent more per transaction. Lucas: That is exactly the kind of metric that makes a CFO pay attention. But here is the tension: personalization at this level requires data. Lots of it. And with privacy regulations tightening — GDPR, CCPA, and now the new federal privacy law that just passed in the US — brands have to be transparent about what they collect and why. Luna: So how do you balance personalization with privacy? Because if you ask for too much permission upfront, users bounce. If you ask for too little, you risk legal trouble or backlash. Lucas: The smartest brands use a progressive profiling approach. They start with basic data — email, maybe a preference — and then layer on additional data over time as trust builds. And they always offer a clear value exchange. 'Give us your birthday, get a free coffee.' Not 'Give us your location and contacts so we can spam you.' Lucas: Another tactic is zero-party data. That is data the customer voluntarily shares, like a style quiz or a product preference survey. It is highly accurate and feels consensual. Brands like Stitch Fix built their entire model on it. Luna: Zero-party data — I love that term. It is the gold standard because the customer is literally telling you what they want. No guesswork, no privacy concerns. Lucas: And that is where the future is heading. In 2026, the brands winning at personalization are the ones that make the customer feel understood without feeling surveilled. It is a subtle craft, but the payoff is real. According to a recent Boston Consulting Group report, companies that excel at personalization generate 40 percent more revenue from their marketing efforts. Luna: Forty percent is hard to ignore. But I wonder — is there a downside? Can you over-personalize to the point where it becomes a bubble? Where a customer only ever sees products they already like and never discovers anything new? Lucas: That is a real risk, and it is called the filter bubble. If your algorithm only reinforces existing preferences, you lose serendipity. Some brands deliberately introduce randomness — like Amazon's 'Customers who viewed this also viewed' section, which is less precise but more exploratory. Or Netflix's 'Trending Now' row, which is not personalized at all. Lucas: The best personalization systems blend relevance with discovery. They show you what you want, but also something you did not know you wanted. That is the art. Luna: It reminds me of how good physical retailers work. A great salesperson remembers your size and style but also suggests something a little outside your comfort zone. AI is trying to replicate that intuition. Lucas: Exactly. And speaking of great recommendations, if these kinds of marketing conversations have sparked something you have actually used at work or even just thought about differently, a small way to keep the show going is to support it at buy me a coffee dot com slash fexingo. No pressure, but listener support is what keeps this ad-free. Luna: Yeah, it really helps us keep digging into topics like this without any sponsorship strings attached. We appreciate everyone who's chipped in. Lucas: Anyway, back to personalization — let's talk about how to avoid the creep factor. One key principle is frequency capping. If you show a user the same personalized offer three times in one day, it stops feeling personal and starts feeling desperate. Luna: Right. And context matters too. A personalized email on a Tuesday morning is fine. A personalized push notification at 11 PM about a sale feels intrusive. Lucas: Another mistake is assuming personalization means 'more is better.' Sometimes the best personalization is knowing when not to market. For example, if a customer just bought a laptop, do not email them about laptops for at least 90 days. That is simple, but most brands do not do it. Luna: That seems like a no-brainer, but I bet most CRM systems are set up to blast the entire list with whatever promotion is running. Lucas: Exactly. And that is why personalization at scale is still a competitive advantage. The tools are getting cheaper, but the strategy is still rare. The brands that invest in clean data, thoughtful segmentation, and empathetic timing will win. Luna: So what is the one piece of advice you would give to a marketer listening who wants to start using AI personalization but does not know where to begin? Lucas: Start with one channel and one segment. Pick your highest-value customer group — maybe your repeat purchasers — and personalize their email experience. Test subject lines, product recommendations, or send times. Measure the lift in open rates and click-through rates. Once you see a win, expand to a second channel. Lucas: Do not try to do everything at once. Personalization is a journey, not a switch you flip. And remember, the goal is not to be clever — it is to be helpful. Luna: Helpful, not creepy. That is a good tagline for this whole episode. Lucas: It is. And I think that is where the smartest brands will separate themselves in the next few years. Not by having the most data, but by using the data they have in the most human way.