​

Personalization at Scale

The promise of the internet was always a world of tailored experiences. Yet true personalization at scale remains elusive. As consumers, we've grown accustomed to search bars and rudimentary recommendations, but the 'perfect' suggestion still feels out of reach. In the realm of online shopping, this gap is particularly evident.

I think personalization at scale is still missing from the internet. Sure, search bar recommendations work exceptionally well with sites like Amazon, Google and YouTube. For example, I’ve been a power user of Google’s Gemini for a specific reason: I use Google for everything. Personally, professionally and everything between. Now Gemini can search based on my personal history and workspace.

However, while Google dominates search, there’s one thing that it doesn’t do well: shopping. In fact, more than 50% of online ecommerce searches happen on Amazon, according to Search Engine Land. Meaning a significant number of consumers are searching with the intent to buy right away. That’s a massive opportunity waiting to be captured by a better delivery model. And I think A.I. chatbots will capitalize on that better delivery model.

The Future of Hyper-Personalization

Hyper-Personalized Product Recommendations only work if the A.I. can analyze your user data. This includes browsing history, purchase patterns, preferences, demographics and much more. Otherwise the software will not be able to tailor products to your individual tastes. This is a tough problem in real-time, especially when you combine consumer security and privacy issues with it. But if personalized recommendations are a success, it will reduce search times and increase the likelihood of purchases. Which is a win for consumers and businesses.

Of course, recommending the right products is only the first step to the marketing funnel. E-commerce businesses are already tough to operate because they have so many moving parts, from logistics to sales & marketing. So integrating hyper-personalization solutions like dynamic pricing can get complicated quite fast. Every store would offer discounts based on user behavior, but collecting data from the supply and demand side makes it challenging.

Right now the simplest form of hyper-personalization I have seen comes from Conversational A.I. Chatbots. The best delivery model has been to provide instant customer support, answer queries, and guide shoppers through the buying process. This has been more cost-effective than investing in a labor-intensive customer support desk. Chatbots have begun to integrate voice and visual inputs that allow shoppers to say, “Show me sneakers like these,” while uploading a photo, with A.I. instantly sourcing similar products.

While customer support is great and necessary, I have noticed some new developments with A.I. Search that will begin to personalize the shopping experience for everyone.

OpenAI’s New Shopping Integration

There’s no question that OpenAI’s ChatGPT has the first mover advantage when it comes to A.I. product launches. With weekly active users surpassing 400 million, OpenAI has enough scale through its own distribution channel, which is the chatbot. Which is why the company is working to integrate shopping as a product feature next into ChatGPT’s Search.

Last week, OpenAI posted the following on X, "We're experimenting with making shopping simpler and faster to find, compare, and buy products in ChatGPT."

OpenAI said Search is becoming one of its most popular features, reaching 1B web searches over the past week. OpenAI added a couple more updates to Search as well, such as improved citations and trending searches. Search is also now available in WhatsApp, which is owned by Meta/Facebook. The new features might help OpenAI compete better with search giant Google and Perplexity.

ChatGPT’s goal will be to offer relevant recommendations based on the user’s search query. This will result in direct links to purchase product recommendations from ad-free results. This means OpenAI’s shopping results are independent and not influenced by paid placements or commissions from sellers.

Overall, this is impressive because refining a user’s search in a conversational manner makes it more personal. And since ChatGPT has so much knowledge about individual users, the interactions will soon feel like speaking with a high-end concierge.

Competing with Online Giants like Amazon

Before OpenAI launched this shopping feature, I tried Perplexity Shopping and wasn’t a fan. Perplexity has an excellent A.I. tool and I recommend checking out their Chrome Extension. However, Perplexity Shopping wasn’t a major success. I think it has potential to produce serious revenue for Perplexity via affiliate links but it’s not there yet. I didn’t find the search process to be very engaging. So then I went to see what Amazon might be offering.

To my surprise, Amazon hasn’t done much when it comes to A.I. yet. Sure Amazon’s Web Services is powering a lot of applications, but their core shopping business hasn’t maximized conversational shopping assistant as you would expect.

Today when you visit Amazon.com, you will find Rufus, a generative AI-powered shopping assistant. It’s trained on Amazon's extensive product catalog, customer reviews, Q&As, and information from the web but I didn’t find it that useful.

Rufus is cool and will have traction, only because Amazon has achieved scale. But I think it will go the way of Amazon Alexa. It will become a cool, new feature but the masses will not adopt it. Even with the supply advantage, Amazon will be able to build a better product than OpenAI in its current state.

Future Vision of A.I. in Online Shopping

In the next few years, A.I. will create seamless, intuitive, and highly personalized shopping experiences. Imagine a scenario where:

• A shopper opens an app, and A.I. instantly curates a homepage with products tailored to their upcoming vacation, factoring in destination weather and activities.
• A consumer uses voice and AR to guide them through a virtual store, suggesting items based on their budget and style, with real-time try-ons.
• The checkout process is frictionless, with A.I. predicting the best payment method and delivery time, and offering a personalized discount to seal the deal.
• Post-purchase, A.I. tracks the order, proactively suggests complementary products (e.g., accessories for a dress), and handles returns via a chatbot.

I think these possible scenarios will require an A.I. first-approach when it comes to personalized shopping. While OpenAI might build a better product, it’s more likely that a niche industry like apparel or electronics will need to build its own specialized A.I. assistant. Having vertical integration will carry more weight when it comes to delivering a personalized shopping experience.