This article on AI is part of our Vogue Business membership package. To enjoy unlimited access to our weekly Technology Edit, which contains Member-only reporting and analysis and our NFT Tracker, sign up for membership here.
Amid a surge in generative AI and a recalibration to more practical business goals, data science and artificial intelligence are once again hot topics for fashion and retail. While this has been an ongoing effort for a decade, the difference now is that generative AI can create more complex and complete outcomes (such as generating imagery), without requiring as much up-front technical know-how.
This was top of mind during this week’s SXSW conference in Austin, where the bulk of fashion and beauty topics centred on data insights, as well as the new ways that brands are using them in decision-making. Rather than outsourcing decisions to a machine, developers are finding that the new tools are helping expand on human capabilities.
Some of these innovations are behind the scenes, while others put the tech on full display. In February, Google began experimenting with the ability for online shoppers to convert their product search queries into images that can be matched with real-world products, while Stitch Fix announced a tool that helps merchants more accurately buy inventory. Tapestry is looking at how generative AI can be used in forecasting, while in December, Meta began testing a way for people to get style advice via its own AI assistant.
The current interest in AI comes at a time when a new consumer segment — Gen Z — is the first generation to have grown up with smartphones and smart speakers with in-home assistants, offering a potentially more receptive audience to using AI to shop, says Lilian Rincon, Google’s senior director of product for consumer shopping at Google Shopping, while speaking onstage with J Crew CIO Danielle Schmelkin. “We’ve done extensive research with a bunch of Gen Zers… and they really are much more open to a lot of artificial intelligence-type of experiences.”
Here are four new ways that AI is being used to inform products and purchases.
Google’s design-your-own product search
Google Shopping is testing a generative AI shopping tool, called Search Generative Experience (SGE). People looking for a specific product online can use a text prompt to generate an image of something that looks like what they are imagining, then Google will source available products that look similar. For example, someone might enter a prompt for “colourful quilted spring jacket” to generate an image, then tweak their prompt for accuracy, before being served items that visually match the creation.
Google research found that 20 per cent of people need to type five words or more to identify the exact product they are looking for. This is also well suited for those who are looking for a very specific product — especially products that might be described differently by different people. “We call these people ‘vision to reality’,” Rincon says, “They have a vision in their mind.” During an earlier preview with Vogue Business, Rincon demonstrated with a search for an imagined polka-dot puffer coat, which was soon matched with available product. The tool is currently available only to those in the US who have opted into SGE via Search Labs, but already, Rincon says that Google is seeing “good usage” and that people understand how to use it.
Google Shopping is also expanding the virtual try-on tool it launched in July, which lets online shoppers see clothing items virtually fitted on a range of models, using generative AI. Google has now expanded this virtual try-on tech to work with flat-lay product images. This means that smaller retailers who might not have on-model clothing imagery can still take advantage of Google’s on-model fit tool.
Stitch Fix’s future simulation tool
Personal styling service Stitch Fix’s business model is built around algorithms to help scale access to personal stylists. It ingests customer feedback to narrow down and prioritise the potential inventory items that are suggested to stylists, who then curate a final selection of pieces that are sent to the customer. They have recently built on this capability with an internal tool that generates inventory recommendations to its merchandisers. While the tool has been in development for two years, the company is only now sharing how it works. The technology includes a simulator that mimics the business to simulate how clients and stylists will behave in the future. When considering pieces, the simulator factors in details such as client impact (likelihood that a client will purchase the style) and client demand (how much they will want it compared to other options).
The tool can help both speed up and improve the accuracy of inventory decisions, says Sophie Searcy, director of merchant algorithms at Stitch Fix, while speaking about how the company uses AI to predict trends. “We have about three million clients and a few hundred thousand items, so we’re talking about billions of potential matches across all of those clients and items. For each of those clients, we only pick out five items to send. This simulation actually takes what would be an impossible task for humans and allows them to see insights into how things will surface and use that as a demand prediction to make the best decisions on what to buy and what to stock.”
The tool doesn’t just identify popular products — this is something that the team can already observe — but rather items that might have untapped opportunities, she says. For example, a certain style of graphic tee was performing moderately well, but the simulator identified that for a certain segment of customers, this was filling an underserved need, and the customer wanted more; even though the merchandising team might not have initially acquired more. The prediction panned out. Now, about 70 per cent of inventory “re-buys” have been algorithmically informed, and there are plans to expand that, though Searcy notes that human instinct remains a key part of the buying process.
Stitch Fix is also experimenting with a way to “cold start” new styles via its gamified “Style Shuffle” tool, which invites customers to rate outfits and pieces with a Tinder-like vote. Stitch Fix can surface items that aren’t currently in its inventory to get an early read on consumer appetite, says Loretta Choy, chief merchandising and client services officer.
In this month’s earnings call, new Stitch Fix CEO Matt Baer shared that the company is working to deepen the relationship that customers have with stylists, as it practises “financial discipline” to drive profitability. Part of that includes extending its private label offering. “Extensive ongoing client feedback enables us to offer private brands that perform better and more profitably than our national brands,” Baer said, adding that the company plans to introduce more private labels.
Previewing those items through Style Shuffle, and then testing them via the new simulator tool, offers a way to gain feedback and insights efficiently. “I think that this is going to be an incredible way for us to be looking ahead,” Choy says. “It would be amazing for us to continue to use this tool to make better and better decisions. When we talk about productivity, that is one of the ultimate goals for this. We never want to overbuy.”
Meta’s AI stylist
Meta’s second generation of smart glasses, produced in partnership with Ray-Ban, now come with a multimodal AI assistant that is available in beta testing for some users. Wearers can ask the assistant for information on what they are seeing — including styling advice.
For example, someone wearing the glasses can look into the mirror and say, “Hey Meta, look and tell me what goes with this outfit.” The glasses will then capture an image and analyse it using computer vision, while the Meta AI offers audio feedback on the look and advice on how to pair it. (Other features include identifying items and other contextual information, like those from the smart assistants from Google and Apple.)
Forecasting and customer insights using GPT
Tapestry Group’s consumer insights team is experimenting with using ChatGPT and similar tools to supplement its more analogue, qualitative research in trend forecasting, says Alice Yu, VP of consumer insights at Tapestry. For example, this might include using a generative AI tool to uncover key terms related to emerging trends, or to build off of existing trends (such as “Y2K” or “Barbiecore”). This type of research would normally require a heavy lift from research teams, she adds. Researchers could then later use generative AI tools to prototype mood boards using those insights, which can then be shared with focus groups to gather further data.
Employees from Tapestry brands Coach, Kate Spade and Stuart Weitzman now have wide access to a data dashboard that shares relevant insights into their function. A store associate might use the dashboard to better understand what is most popular, for example, and be able to merchandise or recommend this item to in-store shoppers.
Yu cautions that in her work, the “why” is often just as important as the “what”, and while data science might offer facts and figures on key points, this can free up more capacity for the valuable in-person, one-to-one research that often uncovers even more valuable insights, she says. For example, during a closet walk-through, one customer’s interpretation of “bright and floral” garments was revealed to manifest a small pattern on a muted top; while the customer self-described as preferring bright and floral pieces, the actual items purchased did not fit that assumed description.
Tension between AI and human intuition remains, but those experimenting with the technology have found that often, the final decisions come down to more personal insights. Human empathy is important for researchers, Yu says onstage during a SXSW panel. In one interview, a young consumer revealed that they had been teased about wearing a vibrant sweater, which ended up being the inspiration behind Coach’s “Wear Your Shine” campaign. “At the heart of it, we are talking about people. We are not talking about data points or factoids; it’s about people, and their stories are so important.”
Sign up to receive the Vogue Business newsletter for the latest luxury news and insights, plus exclusive membership discounts.
Comments, questions or feedback? Email us at feedback@voguebusiness.com.
