How AI Is Making Over Fashion in the New Machine Age

Louis Vuitton is using it. So is Walmart. And seemingly everybody in between.

Artificial intelligence hasn’t just come to fashion, it’s threatening to take over. Bots are styling looks on websites, finding shoppers the right fit, answering customer questions and much more, from front-of-house clienteling to the supply chain and even human resources.

The promise from major tech companies to purpose-built platforms is that AI can make shopping better, smarter and more enjoyable for consumers while driving profits for brands.

The list of brands taking up AI includes household names like H&M, Zara, Tommy Hilfiger, Valentino, Moncler, Collina Strada, Adidas, Louis Vuitton, Zegna and Kering, to name a few.

It’s a tech frenzy that hasn’t been seen since the dot-com boom started in the late ’90s. But it might be nothing compared to what comes next in the world of shopping bots.

In a December report, Polaris Market Research projected the global virtual shopping assistant market will balloon to $6.9 billion by 2032 from $516 million in 2022. Regionally, North America will “have the fastest growth due to advancements in AI and natural language processing technologies,” it said.

Big tech is driving the momentum.

Polaris highlighted fashion examples powered by ChatGPT, like German e-commerce player Zalando, which released a test for its AI fashion assistant in November, and Mercari, which introduced Merchat AI just over a year ago.

It’s indicative of how apparel brands and retailers are joining forces with tech companies to pursue AI, and it’s triggering a wave of fashion-tech power couples. Think Prada and Adobe. Gucci and Salesforce. LVMH Moët Hennessy Louis Vuitton and Google Cloud. Also LVMH and Alibaba, whose AI partnership just expanded. Embodying this dynamic, OpenAI tech made a cameo at the Met Gala, with a custom bot that conjured 1930s-era New York socialite Natalie Potter at the 2024 event, where her wedding dress was on display.

Now the big players are swooping in to shake up the world of bots again, perhaps even faster than researchers anticipated.

The latest ChatGPT updates this May, introduced the world to a new bot that feels more human than ever in how it speaks, interacts, understands and sees the world around it.

These are profound changes for society — from employment to security — and they’re poised to bring what could have been decades of transformation to the high-touch worlds of fashion and retail in just a few short years.

The near future looks destined to be populated by a whole new class of emotive shopping, fitting and styling bots that can see and respond to what clients are wearing and answer their style questions in real time.

Fashion brands will have to move just as fast, or at least try their best to keep up in a market accelerating at light speed.

And not just with AI generally, but with specific applications.

Several venture capitalists and investors have told WWD that they’re over the general AI craze. Now they’re homing in on focused applications of the tech.

“What everyone just realizes [is that] it’s only the Amazons, Googles, Microsofts and OpenAIs that can afford to own the underlying infrastructure,” Corazon Capital’s Sam Yagan told WWD. “The savviest investors said, ‘OK, we’re all going to be building on top of these people’s work.’” Yagan, an investor and former chief executive officer of Shoprunner, exemplified the trend by backing Laws of Motion, a fashion brand turned fit tech provider that just raised seed funding.

“There’s so much room for that shopping experience to be made more fun…there’s a huge opportunity for AI in fashion,” he said.

Here, the AI areas to stay smart on now.

Multimodal AI and the New Fashion Bots

Customers often judge shopping AI features by the chatbots that present them.

That was terrible news for retail, and the debut of ChatGPT highlighted it.

According to a Capterra poll, published just months after OpenAI released its bot, respondents said that only 25 percent of retail chatbots understood their questions of needs, while ChatGPT got a thumbs up from 67 percent of people.

ChatGPT supercharged the space, leading Google to launch conversational commerce tools for retail earlier this year and Amazon to release its own AI shopping assistant, Rufus, as a test. Reportedly, it’s off to a rough start, churning out inconsistent results and easily getting confused.

Now OpenAI has fascinated the world again with a more humanlike version of its bot, called GPT-4o (as in the letter “o” for “omni”).

GPT-4o is a fluid conversationalist that can read and respond to text, voice and photos quickly, translate spoken language in real time and handle interruptions, even revising its chatter midstream. The default voice can also emote, sounding particularly human. Just ask Scarlett Johansson, who turned OpenAI down and then spoke out when the company released a chatbot voice that sounded too close to her portrayal of the Samantha chatbot in the Spike Jonze film “Her.”

OpenAI pulled the voice, but still got the message out: ChatGPT is growing up. Fast. Soon, it will gain the power of sight. Using the video mode on a camera, GPT-4o will be able to “see” the world in real time, including the user’s facial expressions and appearance.

This ability to take in and respond across multiple formats is called “multimodal.” It’s the new black of chatbot tech, with Google and Microsoft pushing their own versions.

These bots could bring back a human-ish touch to shopping assistance, styling and customer service, and real-time sight alone could drive features like on-demand styling and live AI recommendations, even in brick-and-mortar stores.

Version 4o is free to use and developer tools are already available — which might mean that companies like Mercari and Zalando, whose bots are powered by ChatGPT, could give shoppers their first look at multimodal fashion bots.

Size Guides

There’s no shortage of companies that specialize in decrypting sizing guides, those confounding tables that so often vary from one brand to another, and analyze shoppers’ measurements against them to find the perfect match.

It’s both an art and a science. Size is a sensitive topic, and skewed or aspirational views can throw off systems and data models — all of which helps explain why different methods exist to capture size info, from games to surveys to returns-history analysis.

Data is key for fit predictions, and it can unlock other features like visualizations of a garment’s fit on a lookalike model or avatar, among other things.

Players like Bold Metrics, BodiData, SizeWize and Fit Analytics have been capturing and predicting fit for years, long before Google dove into apparel visualizations last year. Zeekit apparently got so good at it, Walmart bought the business in 2021 and powered virtual fitting rooms in 2022 with the tech. Other multibrand online purveyors, from Stitch Fix to Amazon, are also eager to solve it.

They’re not the only ones.

The big name in this game is True Fit. The company usually works with large brands, but last year it brought the Fashion Genome, its AI fit platform, to all kinds of merchants with Shopify stores.

Since January, the companies reportedly clocked 900 percent growth in adoption, with True Fit users accounting for as much as 27 percent of order volume on Shopify sites.

The fit challenge continues to attract new contenders, such as Carly Bigi, founder of Laws of Motion. The company just opened up for business-to-business partnerships with a bold claim.

Based on a quiz or a couple of selfies, Bigi told WWD, “the predicted body measurements that we have for you, as a customer, is going to be within 99 percent accuracy of the actual body.”

Virtual Fitting

Sizing data and virtual fitting are deeply connected, since one often leads to the other.

“Virtual fitting” can mean a lot of things. Some brands offer fit predictions just to personalize styling or product recommendations, while others use AI to power visual, first-person experiences, where users can try digital looks on themselves with augmented reality.

AR, which adds 3D graphics to a camera’s live view, works well in certain retail situations, like digital beauty makeovers, and is often applied to items such as bags, footwear, watches and eyeglasses.

In January, Walmart’s Zeekit moved into virtual eyewear try-ons, and Cartier, a longtime Snap partner, released a virtual version of its Trinity ring in April, so the public can admire it on their own fingers.

Virtual clothes are another matter. Creating digital garments isn’t hard, especially with generative AI, a form of tech that can create text, audio or images from text prompts. But turning them into a believably accurate AR filter for fitting purposes is notoriously difficult. Snap, for instance, has been working on the problem for years.

But that hasn’t stopped other tech makers, both big and small, from pursuing it. Body Labs’ work in 3D body scanning prompted Amazon to acquire it in 2017, while other body-measuring platforms like 3DLook drive AI fit deeper into the fashion business.

The company’s YourFit platform blends fit recommendations with photorealistic AR apparel try-ons. Its AI chooses items based on shoppers’ shape, fit preference and bestsellers, then lets shoppers see what the products look like on them in a virtual fitting room powered by AR. In November, 3DLook signed a new brand, Inditex Group’s Bershka, as a partner.

Product Recommendations

Fit tech platforms often turn into recommendation engines, a logical evolution that ties together what fits best with what customers like. That may give them an edge over competitors that use shopping behaviors alone.

Relying on purchase data is all-too common now, according to Stella Rousou, business development and partnerships director at Style DNA.

“Most solutions would recommend outfits based on an individual’s previous shopping pattern,” she said. Although it doesn’t specialize in fit predictions, the Style DNA app bases selections on “image consulting techniques,” such as color analysis and body type, so it can custom-tailor picks to the user’s individual attributes.

It also catalogues the user’s wardrobe, so it can suggest ways to pair existing pieces.

“We did years of research to identify the problems that consumers face when shopping online or in store, and we found that 64 percent of online shoppers don’t actually know how to style items,” she added.

Naturally there’s an educational aspect, as users learn what works and what doesn’t. That’s one of the major draws for consumers who flock to styling platforms like Style DNA, Stylitics, Intelistyle, Style Genius, Stylebook, Dressme, Fashionadvice.ai and others, which entices a growing lineup of start-ups to jump into the market.

It’s a crowded space filled with giants like Google, Amazon and others, including some unexpected players. Meta is dabbling in styling with its Meta AI bot, which offers fashion advice over chat or via its camera-equipped Ray-Bans.

EBay isn’t known for fashion, but it figures algorithms can change that. The company expanded its “Shop the Look” feature for furniture to clothes, and its new AI-based click-to-resell tool, which auto-populates product listings, could ensure plenty of product to choose from. Italian outerwear brand Save the Duck is the first brand to take up the recommerce feature.

The Complicated Business of Getting Dressed

AI is everywhere, doing a little bit of everything — but not all together.

Amazon, Adobe, Walmart, Google, Shopify, Salesforce, Klarna, Adobe, eBay, Stitch Fix and the like represent billions of dollars in transactions, and their deep investments in shopping AI are influential, driving waves of data-fueled features and experiences.

It’s easy to get lost in all the platforms from personalized recommendations to fit predictions, not to mention tools for marketing, communications, product design and other behind-the-scenes scenarios. That’s especially true for shoppers.

Even when AI features work well — which isn’t always the case — the scenarios are still confusing. Some tools overlap or work together, others don’t. Most don’t account for gift shopping or even allow customers to manually flag when they’re buying for others, which influences their preference data.

Meanwhile, a single shopper can have multiple profiles at various brands that don’t connect together or to an existing wardrobe. This is why connected closet apps were born, and the group, which includes Style DNA, along with Cladwell, Acloset and many more, remains popular among fashion consumers.

“AI in the fashion industry, particularly in the realm of shopping, we all know is rapidly expanding,” Rousou said of Style DNA. “It can be really overwhelming.”

And that at least, is something everyone can agree with.