How Companies Use AI to Predict Fashion Trends

  • Artificial intelligence is helping fashion brands understand consumer behaviors and demands.
  • The startups Spate and Fashable use AI insights to identify style trends for clients.
  • Fashable’s cofounder told BI he envisions brands and customers using AI to cocreate in the future.
  • This article is part of “Build IT,” a series about digital tech and innovation trends that are disrupting industries.

Artificial intelligence looks poised to change fashion at blistering speed. A 2023 report from the management consultant McKinsey & Company estimated that generative AI could add up to $275 billion to the industry’s operating profits over the next three to five years.

Industry leaders seem similarly optimistic. The Business of Fashion and McKinsey said that in a survey they conducted, about 73% of fashion executives said they planned to prioritize generative AI in 2024.

Though the potential is clear, the path forward is still being built. The opportunity and challenge of AI is in its multifaceted nature. AI can be used to improve efficiency and sustainability to both identify trends and create new ones — but it requires careful implementation.

“Fashion inherently has a tremendous focus on creativity, and AI is going to completely transform the creative process,” said Benjamin Bond, the principal at the business consultancy Kearney. “When we talk transformation, we’re talking about creating speed, creating efficiency, creating relevance.”

What’s trending?

Yarden Horwitz, a former fashion-industry analyst at Google who cofounded the data-science startup Spate, saw the opportunity to bring AI to the industry over a decade ago.

“At the time, fashion brands were still going store to store down Fifth Avenue, seeing what was in the windows versus what was in the sale racks,” Horwitz said. “It was a very manual process and very flawed.”

Horwitz, alongside her colleague Oliver Zimmer, built the first Google Fashion Trends report, which used machine learning to identify trends in Google search data. Jogging pants, rompers, and midi skirts were on the rise; string bikinis and skinny jeans were on the decline.

Headshot of Yarden Horwitz

Yarden Horwitz, a cofounder of Spate.
Courtesy of Yarden Horwitz

More reports followed, covering more topics and industries. Horwitz and Zimmer left Google to found Spate in 2018. Today, the company uses generative AI to augment and expand its analytics.

“We have used AI to identify trends before,” Horwitz said. “But using generative AI, you can also have it review those patterns. The AI can generate its own patterns and come up with recommendations.”

Spate launched an AI-generated consumer-trends report in 2023. The company used GPT4 to analyze over 2 billion search signals across 10 industries, including fashion. It found that shoppers wanted authenticity and a sense that a product would improve them — but, perhaps paradoxically, celebrities remained a force in setting consumer trends.

A trend analysis of popular fashion styles and aesthetics

Spate’s trend report highlighted popular aesthetics such as “Barbiecore” and “old money.”
Spate

Finding fashion’s secret sauce

Using AI for automated, customized, and targeted reports holds promise for many industries but could prove especially useful for fashion. Bond said fashion’s focus on creativity can lead to catalogs with thousands or tens of thousands of products, the details of which can be difficult to quantify.

Headshot of Benjamin Bond

Benjamin Bond, the principal at Kearney.
Courtesy of Benjamin Bond

“One of the X factors is the endless permutations of how you can create a specific garment. Is it a T-shirt, or is it long-sleeved? Does it have a pocket? Is it drop hem or even hem?” Bond said. “Identifying those attributes manually across tens of thousands of products is impossible.”

He added that AI would allow this “attribution at scale.” A fashion brand could use computer vision to identify product attributes across its catalog. Then it could use that data to identify trending attributes and find correlations in its catalog.

AI can also assist in the more technical aspects of producing fashion at scale, like choosing a supplier that can offer the best thread for a particular garment or finding tweaks to a garment’s design that lower its cost. Bond said these attributes are often invisible to customers but can make or break a garment on the market.

“At the very early design stage, we’re going to see a ton of generative AI there,” Bond said. He expects brands with a history of designing and debuting products digitally, such as Nike, to tap the power of generative AI for fashion design.

Is cocreation the future?

The fashion startup Fashable is among those building this new breed of AI-powered tools. Born in 2021 out of XNFY Lab, a research organization that worked with Microsoft to foster innovation in retail, Fashable creates AI-generated images for designers, brands, marketplaces, and more.

Fashable says its team of 12 is training AI models to generate realistic photos of garments for clients based on customer insights. The goal is to use those images to create styles that are more likely to sell, reducing overproduction and unsold inventory.

Fashable’s AI approach also addresses a challenge with creativity. While it’s possible to ask ChatGPT to create a design, the reality isn’t so simple. Fashion companies want generative AI trained on their data, not data from competitors, to ensure the results are true to their style. Similarly, they don’t want their data to end up in competitors’ hands.

“Part of our value is that we train a model for a brand, and the model will only be for that brand,” said Orlando Ribas Fernandes, a Fashable cofounder. “We train on their data, and that data, and the results of that data, will only be for them.” This method, Fernandes added, doesn’t “contaminate other AI models,” helping prevent infringement of brands’ copyrights.

Fashable cofounder Orlando Ribas Fernandes speaks on stage at an event highlighting Fashable.

Fashable’s Orlando Ribas Fernandes showcasing an AI-generated dress at a retail event.
Fashable

Fernandes argued that the goal of using AI in fashion isn’t necessarily to reduce design costs. “Everyone thinks AI will be very cheap, and it’s not cheap,” he said. “It’s very expensive.” Fernandes expects the industry to instead adopt AI to fix its inefficiencies.

Many fashion products are sold well below retail value, while some end up in landfills without ever being worn. It’s a massive problem Fernandes knows is difficult to solve. Still, he hopes generative AI will lead to fashion that meets people’s specific desires, in turn reducing the need to spawn endless varieties in hopes that one will be a hit.

“My vision is that brands will start cocreating with the customer,” Fernandes said. “In the future, I might prefer to have one blazer that was custom-tailored for me, instead of something from fast fashion. I can start having my own digital wardrobe, and that is produced only for me.”

Related Posts