Luxury brands are turning to generative AI to improve their personal shopping capabilities. On Kering’s experimental KNXT platform, users meet Madeline, a ChatGPT-based luxury personal shopper. “Ask me anything and I will do my best to surprise you with luxury pieces,” it says, and “give me as much context as possible,” it adds, so that it can pull recommendations from a product list provided by Kering.

Since OpenAI’s ChatGPT tool opened up to developers last month, luxury brands are testing generative AI to improve their personal shopping capabilities in a way that can scale. Brands can integrate these into their websites and apps, while specifying parameters such as which product assortments to pull from. German e-tailer Zalando has just announced that it is testing a “fashion assistant” powered by ChatGPT on its app and website, where customers can ask similar prompts, in an effort to make the process more “intuitive and natural”. Peer-to-peer marketplace Mercari just launched a shopping assistant called Merchat AI, using ChatGPT tech, which follows up on prompts with questions that further dial in to customer preferences and needs, such as brand, colour and style, before making a series of recommendations. E-commerce personalisation provider Bloomreach, used by Neiman Marcus and Benefit Cosmetics, just integrated OpenAI into the company's content management solution (CMS).

What’s particularly groundbreaking when it comes to personalised style recommendations, Stark adds, is the ability to tailor results as if it were a real-time conversation. If KNXT’s Madeline initially serves up a white caftan for a beach holiday, the customer can respond with something as simple as, “No, I need something colourful and floral,” and the tool will come back with a more appropriate suggestion.

AI and machine learning tools have become widespread across industries, and AI is being increasingly leveraged to improve the customer experience, especially with better chatbots. Personalized style recommendations are especially groundbreaking, as they tailor results as if it were a real-time conversation.