Latest / The DTC Marketing Podcast with Fexingo: Direct-to-Consumer Brands, Shopify, and E-Commerce Growth / How DTC Brands Use AI Chatbots for Personalized Shopping
Transcript
- Lucas: So you walk onto a DTC site, and instead of hunting through drop-down menus, a little chat window pops up and says, 'Hey, I'm here to help you find the perfect product.' That's not a customer service bot anymore. That's a sales assistant — powered by AI — and it's quietly becoming one of the most effective tools for DTC brands to boost conversion and cut returns. Luna: I've definitely seen those. Some are good, some are glorified FAQ pages. But the ones that actually work — what's the difference? Lucas: The difference is whether the brand treats the chatbot as a product recommendation engine rather than a help desk. Let me give you a concrete example. Glossier — the cosmetics brand — launched something they call 'Glossier Bot' about a year ago. It's a conversational AI that asks you about your skin type, your coverage preference, your shade range, and then suggests specific products. Luna: So it's like having a beauty consultant, but automated. How does the tech work under the hood? Lucas: They're using a large language model — think GPT-4 class — fine-tuned on their entire product catalog, plus thousands of customer reviews and Q&A transcripts. The model learns which product attributes correlate with positive outcomes for different customer profiles. So when you tell it you have oily skin and want a light-coverage foundation, it knows that the 'Perfecting Skin Tint' is a better fit than the 'Stretch Concealer', even if neither product description explicitly says 'for oily skin.' Luna: That's clever. But does it actually move the needle on sales? I've seen some stats that chatbots can annoy customers more than help them. Lucas: For Glossier, the numbers are pretty compelling. They shared that customers who interacted with the bot had a 15% higher average order value compared to those who didn't. And more interesting — a 20% reduction in return rates for products recommended by the bot. Because the recommendations are more personalized, people end up buying things that actually work for them. Luna: A 20% drop in returns is huge. That's not just revenue — that's margin improvement, lower logistics costs, better sustainability metrics. Lucas: Exactly. And this is where DTC brands have a real edge over traditional retailers. They own the customer relationship and the data. So they can train these models on first-party data that's incredibly specific — what a customer bought, what they returned, what they browsed, what they asked in live chat. A department store can't do that because the data is fragmented across brands. Luna: So the bot is essentially a conversion layer that also reduces friction. But what about the customer who just wants to talk to a human? Do you risk alienating that segment? Lucas: That's the key design challenge. The best implementations — and Glossier's is one — make it easy to escalate to a human agent. The bot can hand off the conversation context seamlessly. And they track sentiment: if the customer seems frustrated or asks three times for a person, the bot proactively offers to connect them. So it's not all-or-nothing. Luna: I read a study from Juniper Research that said by 2025, chatbots will drive over $100 billion in e-commerce transactions globally. That was before the latest AI wave — the number is probably higher now. Lucas: And the technology is getting cheaper. A few years ago, building a custom conversational AI required a team of NLP engineers and a six-figure budget. Now you can get a decent off-the-shelf solution from platforms like Tidio or Zendesk for a few hundred dollars a month, and fine-tune it with your product data in a week. That's democratizing access for smaller DTC brands. Luna: What about the risk of the bot giving bad advice? If it recommends a product that doesn't fit, the customer gets annoyed and might leave a bad review. Lucas: That's real. And that's why Glossier invested heavily in testing. They ran a two-month beta where the bot's recommendations were reviewed by human stylists before going live. They also built in guardrails: if the model's confidence score for a recommendation is below a threshold, it defaults to saying, 'I'm not sure, let me connect you with a specialist.' So they're not letting the bot run wild. Luna: Sounds like a smart approach. Are there other DTC brands doing this well besides Glossier? Lucas: Warby Parker has a bot that helps you pick frames based on face shape and style preferences. They claim it increased optical sales by 10% per session. And then there's a newer brand, Hims & Hers, that uses a chatbot to guide customers through a health assessment before recommending products. That one's essentially a triage tool — it collects medical history and suggests regimens, then a human doctor reviews it. So the bot handles the tedious data collection, and the doctor focuses on clinical decisions. Luna: That's a great use case — letting AI handle the 80% of the work that's repetitive, and humans handle the 20% that requires judgment. I think that's the sweet spot for a lot of DTC applications. Lucas: Absolutely. And it ties into a broader trend: conversational commerce. The idea that the chat interface becomes the primary way people shop, especially on mobile. Amazon's already testing Rufus, their AI shopping assistant. And Shopify merchants can now embed chatbots that recommend products directly in the chat. Luna: So where do you see this going in the next year or two? Is the chatbot going to replace the product page entirely? Lucas: I don't think it replaces the product page, but it becomes the front door. You land on the site, the bot says, 'What are you looking for today?' You tell it in natural language, and it serves up a curated set of products. That's a much lower friction experience than browsing categories and filtering by size, color, price. Especially on mobile, where screen real estate is limited. Luna: That makes me think of the 'headless commerce' trend, where the frontend is decoupled from the backend. A chatbot could be one of many touchpoints — like a voice assistant or a social media DM — that all pull from the same product data. Lucas: Exactly. And that's where DTC brands have an advantage, because they're often built on modern stacks like Shopify Plus or BigCommerce, which have APIs that make it easy to plug in conversational AI. Legacy retailers with monolithic systems struggle to do that. Luna: Speaking of which, if these marketing conversations have sparked something you've actually used — maybe a chatbot idea for your own store — we just wanted to say that the reason we can keep doing deep dives like this without ads is because of a small group of listeners who chip in monthly. Lucas: Yeah, it's a handful of people — but it makes a real difference. If you've found the show useful and want to help keep it ad-free, you can find us at buy me a coffee dot com slash fexingo. No pressure, genuinely. Just a way to say thanks if it's added value. Luna: Totally. And now back to the bots. Lucas, you mentioned that some brands are using chatbots not just for recommendations but for post-purchase follow-up. How does that work? Lucas: Right. So after a customer buys a product, the bot can check in a few days later and ask, 'How's the moisturizer working for you? Do you have any questions about application?' That's a retention play. It builds a relationship, and it also surfaces issues early — if someone says, 'It made my skin break out,' the bot can offer a return or an alternative product before they write a negative review. Luna: So it's almost like a customer success function. That's smart because it reduces churn and can increase lifetime value. Lucas: Exactly. And the data from those conversations is gold. You learn which products cause issues, which ones delight, and you can feed that back into product development and marketing. Glossier actually uses chatbot interactions to identify trending concerns — like 'puffiness' or 'redness' — and then creates content around those topics. Luna: That's a virtuous cycle: chatbot collects data, data informs content, content drives traffic, traffic engages with chatbot. Very DTC. Lucas: It's the flywheel effect applied to conversational AI. And the more customers interact, the better the model gets. That's why brands that start early have a compounding advantage. Luna: One concern I hear from marketers is about privacy. Chatbots collect a lot of personal data — skin type, health conditions, preferences. How are brands handling that? Lucas: It's a legitimate concern. The good players are transparent about what data they collect and how it's used. They offer opt-in, they anonymize data where possible, and they make sure the chatbot doesn't store sensitive information like credit card numbers. Some are even building their own models on-premise to avoid sending data to third-party APIs. Luna: So it's not just a plug and play solution. Brands need to think through data governance from the start. Lucas: 100 percent. But for those that do it right, the payoff is significant. We're seeing early adopters report 10 to 20 percent lifts in conversion, 15 to 25 percent reductions in returns, and better customer satisfaction scores. It's still early days, but the direction is clear. Luna: What about smaller DTC brands with limited budgets? Can they still benefit from this technology? Lucas: Yes, but with caveats. There are affordable SaaS options like Chatfuel or ManyChat that integrate with Shopify and offer basic recommendation flows without custom AI. They won't be as sophisticated as Glossier's bot, but they can still handle simple product finders. The key is to start simple — maybe just five or six questions — and iterate based on customer feedback. Luna: So the barrier to entry is lower than people think. It's not all about building a custom LLM. Lucas: Exactly. You can get a lot of value from a well-designed rules-based bot, as long as the questions are smart and the recommendations are relevant. The AI part becomes more valuable as you scale and have more data to train on. Luna: I think a lot of listeners will be interested in trying this. Any final advice for someone considering adding a chatbot to their DTC site? Lucas: Start with a clear goal. Do you want to increase average order value? Reduce returns? Improve customer satisfaction? Pick one metric, build a bot that targets it, and measure. Don't try to do everything at once. And always, always make it easy to talk to a human. The bot should be a helper, not a gatekeeper. Luna: Good advice. I think we'll see a lot more of these bots in the next year, especially as the technology gets cheaper and customers get more comfortable chatting with AI. Lucas: Definitely. And for the brands that get it right — combining personalization with genuine helpfulness — it's a real competitive advantage. Alright, that's all for this episode. We'll be back next week with another deep dive.