What if AI became our personal shopper?

In the future, the advent of AI in the daily lives of citizens around (almost) the entire world could be remembered as a third industrial revolution. Artificial intelligence has already begun to transform our daily lives, from self-driving cars to the numerous applications of ChatGPT. But now, this new technology is set to revolutionize the world of shopping with the emergence of AI personal shoppers that select, suggest, and present products to end customers. As reported by BoF, these AIs are designed to optimize the shopping experience, making it more “efficient” for both customers and retailers by handling style searches, product and price range comparisons, and even checkout processes. Imagining such a service online, the AI personal shopper could eliminate the tedious task of browsing sites like SSENSE or Farfetch, where users must scroll through hundreds of pages, endless brand lists, and effective yet not entirely specific search filters. All this with the ease and intuitiveness of dialogue: you tell the AI what you're looking for, your style, your budget, and so on, and the assistant takes care of gathering everything into a single list. In-store, the same mechanism could assist physical sales associates in real-time by updating inventory and availability as well as analyzing the spending patterns of a specific customer. Of course, doubts remain about their impartiality, as no one can guarantee the end user that the service won’t end up recommending sponsored products — the autonomy of choice is the price to pay for automation of the process. Moreover, nothing guarantees brands and retailers that such AI won’t be used to find cheaper dupes of their products, indirectly benefiting competitors. But are we really ready to entrust our shopping to an algorithm?

The growing demand for AI assistance in shopping is evident: a report by Salesforce cited by BoF found that U.S. holiday online sales increased by nearly 4% in 2022, partly due to the rising use of AI-powered chatbots offering simple and effective customer support capable of guiding them to desired products through personalized recommendations. Another report by Salesforce reads: «Retailers are betting big on the future of artificial intelligence (AI), with 92% investing in this technology. AI use in retail is not new: 59% of retailers use it to help store employees recommend products, while 55% use digital assistants to help online shoppers». This trend is expected to accelerate: according to Gartner consultants, by 2027 over half of consumers will regularly use AI personal shoppers to make purchase decisions. Meanwhile, a report by the Canadian development company Springs states that 40% of users already use chatbots to navigate inventories, tools that allow companies to improve sales by 67% and increase profits by an average of 20-40%. 

One of the most relevant companies in this new sector is Perplexity, which last November introduced a new feature that allows users to directly request product recommendations from AI, based on blog posts, videos, expert reviews, and social media posts while avoiding sponsored links and other “guided” content. With the Pro version, users can even complete the purchase directly on the platform and, over time, the system learns individual user preferences. Last March in Saudi Arabia, Shahad Geoffrey and her company Taffi introduced Amira, a generative AI stylist designed for the needs of the Middle Eastern market. Unlike many AI systems built with Western aesthetics, Amira understands the nuances of fashion in the Gulf region and has been trained to recommend clothing based on both local inventory and cultural preferences. For example, it avoids suggesting outfits that might be inappropriate or culturally irrelevant for the local audience. The implications of an AI capable of engaging with customers based on the cultural needs of a specific market are immense, as ideally, AI assistants could do much more: manage entire inventories, provide insights into consumer behavior, social media trends, and even analyze individual user patterns, increasing conversion rates, suggesting products based on demand, thus reducing the risk of overproduction or stock shortages. 

Nevertheless, one should not assume these AIs are perfect. Fallibility is the price of complexity, and indeed most of the AIs discussed, including Perplexity, rely on language models that cannot understand real-world context as humans do. They operate according to schemes prone to errors or nonsensical information known technically as “hallucinations”. On the search methodology, strangely, little is yet known. «To be honest, all these aspects still need to be fully understood in terms of ranking mechanisms [and why] AI prefers ranking one over another», explained Aravind Srinivas, CEO of Perplexity, to Fortune last November. «Is it the number of reviews? Is it the exact rating and the origin of the rankings, such as what people say on different platforms about their product? There is a lot of distillation and condensation happening. I think even we today do not fully understand it». Furthermore, these systems function only based on the data they have access to, which becomes problematic in the case of outdated inventories or data the AI cannot access — essentially, the search engine is still less universal than the ambitions of a mega-system spanning the entire Internet would require. This and the suspicion of “guided” choices perhaps represent the biggest challenges that new personal shoppers have to overcome.

Despite these complexities, some companies are still introducing AI assistance on a smaller scale. For instance, in September, The Guardian reported that Marks & Spencer (M&S) uses AI to offer personalized style advice based on customers' body shapes and personal preferences. By analyzing customer data from an online quiz, M&S's AI can suggest outfits from a wide range of options, making personalized shopping more accessible. According to the article, the system is helping M&S boost online clothing sales and increase customer engagement. In October, Who What Wear launched its new assistant ISA (short for Intelligent Shopping Assistant), which appears as a chatbot capable of helping customers find the products they seek. Other examples of existing AI assistants include Sephora's Virtual Artist, Amazon's Rufus, eBay's ShopBot, and IKEA's Room Planner — not to mention H&M's chatbot active since 2022. In the fashion industry, Burberry and Tommy Hilfiger introduced one back in 2016 via Facebook Messenger but not on a continuous basis — similar initiatives occurred with Audemars-Piguet and Jaeger-LeCoultre in the same years, but they did not appear to follow through. However, today, it is certain that many retailers and even many brands could be interested in having such a chatbot on their website. 

According to Perplexity’s chief business officer, Dmitry Shevelenko, the possibilities for AI personal shoppers are immense, and they could soon respond precisely to consumer demand based on a combination of their purchase history, browsing behavior, and social media activity — though such integration raises privacy and personal data usage concerns. Even though many still prefer the physical experience of in-person shopping (in-store purchases remain a must in the luxury world, as demonstrated by Chanel and Hermès), integrating AI systems could significantly eliminate many of the most frustrating aspects of shopping — especially considering how, even today, fashion boutique associates perform much of their work on tablets to consult inventory availability, effectively blending technology with a human approach. On the physical retail front, it will be essential to understand how to integrate such systems. Last year, Zegna introduced the digital tool Zegna X, a comprehensive AI-based system that not only allows customers to choose and find their models, selecting fabrics and colors while visualizing the complete outfit in detail, but also gives the brand's style advisors and artistic director direct access to every sale and client information, with the ability to refine, perfect, and optimize the offering. On the digital front, however, the widespread adoption of AI personal shoppers will require earning customer trust and proving they can genuinely improve the shopping experience — especially at a time when online channels are used to search for the best discounts across different retailers, as seen with Lyst. But will speeding up the process of purchasing a specific product really help boost declining sales? Or will it merely eliminate the human relationship between customer and retailer altogether?