This week, a deep dive on generative AI, including what IP issues are involved, and how it is changing the way fashion businesses are approaching imagery, stock management, marketing and product discovery. Scroll down to use Glossy+ Comments, giving the Glossy+ community the opportunity to join discussions around industry topics.
Generative AI is becoming a big efficiency play for fashion brands. Generative AI is an artificial intelligence algorithm trained on large data sets to create content, text, audio and videos. Going beyond the advanced analytics that brands have been using for years, generative AI is allowing brands to create content quicker and, most of the time, at a lower cost.
“I view artificial intelligence as a subcomponent of advanced analytics, where the steps are automated,” said Sarah Willersdorf, global head of luxury at Boston Consulting Group. Willersdorf said that distinct features of AI are machine learning, or the ability to self-learn, and natural language processing, which is the ability to understand or reproduce human language versus just code. One of the benefits of using generative AI is that it can come from unstructured data sets, as opposed to ordered data sets, she said.
With data being integral to AI content, there are growing legal issues around where that data is coming from and who owns it. Brands keen to avoid litigation are exclusively training and using generative AI using their own first-party data.
Although exact IP laws have not yet been created specifically for AI use, visual media company Getty has already filed a lawsuit against the creators of Stable Diffusion, an AI image generator, in February. It alleged that the involved photos were used without permission, violating its copyright and trademark rights. Owners of large amounts of data like Getty are set to be important decision-makers in how quickly AI use will become commonplace and what imagery is usable for generative AI.
Announced in April, German e-tailer Zalando is using ChatGPT, which is a generative AI text tool, as a consumer-facing virtual assistant. Customer data protection is particularly important for Zalando, according to a company spokesperson. “We aim to create a trusted companion for our customers, which includes privacy and AI protection in line with Zalando’s commitment to offering an experience of trust,” they said.
“In order for our Fashion Assistant to provide the relevant answers, customers’ searches are processed by ChatGPT on our behalf,” they said. “However, this happens under a service provider agreement that includes the protections required under the General Data Protection Regulation, including a data processing agreement. This means that OpenAI will not process our customer’s data for their own purposes at all, but only under our instructions and control.”
For the early fashion adopters that are training AI on their own data, the results have been very good. Sheep Inc, a British luxury knitwear brand founded in 2018, used it to develop a campaign that will be released on May 19. The AI integration, which leveraged the brand’s past campaign photography, cut down the brand’s typical campaign timeline from eight weeks to a couple of days, according to the company.
Sheep Inc used its own photography and AI tools to create photo-realistic campaign imagery featuring its Half Zip knitwear, sheep and models. The creative team worked with the tools to create the desired images. Company representatives declined to specify the specific tools used and the investment required.
Sheep Inc is already using NFC tags to track its garments from its sustainable wool source in New Zealand to steps including processing and manufacturing. For the team of 10, leveraging an AI-based campaign has multiple benefits, including saving time, costs and carbon impact.
“We’re always looking forward to what’s next and how we can further improve our impact,” said Edzard Van der Wyck, co-founder of Sheep Inc. “With all of these generative AI tools coming out, we understand quite rightly that there’s a lot of tech fear around it. We are still embracing this with some trepidation, but from a content creation point of view, it’s a real game changer.”
Typical campaigns for a luxury brand like Sheep Inc take between 4-8 weeks to produce. Steps include sourcing models, putting together mood boards, finding the right photographer and location, shooting the photos and editing them. It’s an expensive process, and if the final imagery is not right or does not receive positive feedback from customers, it often must be started again.
“We’re still a growing business, and we can’t compete with the marketing budgets of some of these huge fashion houses,” said Van der Wyck. “We want to create these exciting, aesthetic worlds that people want to engage with, but often, that is actually unaffordable. It’s very difficult to get that level of result if you’re doing it through traditional photoshoot means, but with AI, it’s significantly easier.”
Fashion photoshoots also have a large carbon footprint. “If we want to do anything remotely exotic, there is the carbon footprint of traveling,” said Van der Wyck. “We’ve also wanted to bring in sheep, but we’ve been hesitant to do so, because then issues of animal welfare come into play.” For the new campaign, AI sheep imagery was used.
Van der Wyck said that AI campaign costs are minimal, especially compared to the brand’s traditional campaigns, plus they reduce burdens on a small team. “In this current environment, to survive, you need to be running really leanly, because the margins everywhere are so tight and marketing costs are massively going up.” To make sure its campaign is legally compliant, Sheep Inc only used its own first-party data to train the AI, he said.
Van der Wyck is also leveraging AI for other areas of its business. That includes sales processes, stock demand forecasting and customer service. “One of the members of our customer experience team is also running our B-Corp process,” he said. “With AI, and she’ll have more time to [focus] on the [B-Corp] side of things.”
Meanwhile, for Revolve, AI image generation opened up new potential around marketing creativity. The e-tailer chose AI creative agency Maison.Meta — the host of AI Fashion Week — to produce and execute its latest billboard campaign “Best Trip” that ran in multiple locations in Palm Springs last month. The creatives leveraged image data, creating a prompt that factored Revolve signatures, including influencers, bright colors and festival connections.
“Revolve started as a technology company first, so we felt … AI reflected our history of innovation and the next phase of evolution,” said Michael Mente, founder and CEO of Revolve. “Our goal in this campaign wasn’t to reduce costs, but to create a campaign leveraging AI technology to produce something that couldn’t be done through traditional production.” At the moment, Revolve is resorting to using image data scraped from various sources by Stable Diffusion.
“We believe AI has the potential to touch every aspect of our business in one way or another, and are aggressively exploring the latest technologies and tools in every zone,” said Mente.
AI was first used by Revolve in its fraud department six years ago, and the company has since continued to roll out applications across the organization. Revolve, which is a publicly-traded company, is heavily investing in technology, which it’s mentioned multiple times on earnings calls. Overall, investors have long been keen on companies that prioritize innovation.