This article on generative AI ads is part of our Vogue Business Membership package. To enjoy unlimited access to Member-only reporting and insights, our NFT Tracker and TikTok Trend Tracker, weekly Technology and Sustainability Edits and exclusive event invitations, sign up for Membership here.
Advertising is getting the generative AI treatment. If it takes off, it could change how fashion marketers work. Will it also change how we perceive ads?
In May, Meta (the parent company of Facebook and Instagram), announced that in July it would begin rolling out access to what it calls the AI Sandbox, a way of testing new tools and features that include ad tools powered by generative AI. The first of these includes the ability to create background images from text inputs, in the vein of text-to-image tools such as Dall-e and Midjourney. Image outcropping has also been added to automatically generate a range of aspect ratios from the same content to better fit in different formats (such as Stories and Reels), and a tool that generates multiple variations of text, so that advertisers can tailor messages to their audiences.
Soon after, Google introduced similar tools that will begin to be available in the US in late July. Called Product Studio, these include the ability to generate scenes that create lifestyle imagery out of product shots; the ability to remove backgrounds from existing product shots and the ability to increase the resolution of small images. And this week, software company Salesforce announced what it calls MarketingGPT and CommerceGPT (which stands for “generative pre-trained transformer”); capabilities will include generating personalised emails (starting in October), creating “contextual” visual assets for multi-channel campaigns and generated product descriptions that are tailored to buyers (starting next month). Meanwhile, e-commerce firm Shopify is introducing ChatGPT-generated product descriptions, and clients of Amazon Web Services will now be able to build their own generative AI bots.
Generative AI, which can create content in the form of images, videos and text, is a new field for fashion and retail that promises to aid creativity and business processes — but also introduces new considerations.
Unlike the applications of AI in text and video that have surfaced as experiments online — people with seven fingers and blurred faces, or suggested text riddled with clichés, repetition and factual errors — the hope is that, with these new tools, customers who encounter the creative results won’t notice that generative AI has played a role. It offers an early look at why some are so bullish on, or concerned by, the potential for generative AI to upend marketing roles — and it’s just a starting point.
“While it’s still early days, we wanted to start a conversation with advertisers and build new use cases for generative AI together to help them maximise performance,” says Nicola Mendelsohn, head of global business group at Meta. She says the company is intentionally starting the AI conversation with advertisers and the broader industry so it can learn best uses and appropriate disclosures before any AI ads are deployed to customers.
These tools would make it faster and easier to generate enough marketing materials to last across multiple platforms and uses. This could have a major impact on the ad ecosystem, on e-commerce and on personalisation, because it can change the way brands test various assets, and lets them create more variety that is personalised to more customer segments. “The old way was to plan a campaign and stick to it,” said Jones Road Beauty CMO Cody Plofker during Meta’s presentation. “You would spend the majority of time thinking about creative, but with AI you can test and learn and iterate so easily [because it] allows AI to do the learning for you and provide data that you can iterate off of.”
The tools would also let brands test personalised emails and web copy on a scale and speed that wasn’t previously possible. Salesforce’s new CommerceGPT tools use what it calls “harmonised first-party data” (data from a variety of sources that is directly owned by the brand and from the customer) to enable retailers to automatically generate product descriptions tailored to every buyer. If it knows that someone bought a certain handbag and is planning a summer holiday, it might include text that suggests certain pairings.
“Personalisation is becoming an expectation from a majority of people, so tools that help scale and improve this stand to be a key unlock for a number of brands,” says Samantha McCandless, chief merchandising officer at luxury resale marketplace The RealReal, which works with Salesforce and has already been using AI to generate e-commerce product descriptions.
More than half of marketers are already experimenting with generative AI at work, according to a May survey of marketers in the US, UK and Australia commissioned by Salesforce; 71 per cent expect it to increase their productivity, estimating it will save about five hours of work per week. At lingerie brand Adore Me, VP of strategy Ranjan Roy estimates that an internal tool that helps generate product descriptions has saved employees about 30 hours a month on copywriting.
Demand for generative AI products could add about $280 billion in revenue for software tools such as specialised assistants, infrastructure products and copilots that accelerate coding, according to a June report from Bloomberg Intelligence, and big tech companies could be some of the biggest beneficiaries. The second largest driver of incremental revenue will be digital ads driven by the technology, accounting for $192 billion. Overall, the gen AI market is estimated to reach $1.3 trillion in the next 10 years, up from $40 billion last year, per the report.
Like most new technology, generative AI comes with both legal and ethical considerations whose boundaries have yet to be hammered out. The most straightforward concerns are often focused on the risk of computers copying the work of others or making mistakes in material that is ultimately used by brands, or replacing humans in jobs on a mass scale. Strikingly, tech leaders even warned in May of a worst-case scenario that threatens extinction comparable to pandemics or nuclear wars.
In the immediate term, these messages haven’t seemed to meaningfully mitigate enthusiasm. For brands, the siren call of marketing efficiency is more tangible. For the tech companies, their platforms and tools stand to become more valuable. While Adore Me has been using and building its own generative AI tools internally, having them integrated into a core platform like Meta will be better and easier, Roy says, because it will enable the brand to create and launch a variety of ads faster — even 20 at a time — to targeted demographics.
Variations on a theme
For fashion and beauty brands, the use cases for these AI tools is especially relevant, says Rachel Tipograph, founder and CEO of social commerce analytics platform Mikmak, as creating a few attractive lifestyle images and one version of compelling copy is no longer enough. Brands need an array of options, especially as ad formats diversify and customer segmentation becomes more specific.
“If you want to sell a pair of jeans, you need 36 versions of the creative to drive conversion,” says Tipograh, who was formerly the global director of digital and social media at Gap. “And that’s just on Meta alone.” This is in contrast to a more traditional brand awareness campaign with a renowned photographer and a celebrity, which might only call for a total of six final assets, she points out.
Beauty brand Jones Road was an early tester of Meta’s new AI Sandbox, and used it to add various backgrounds behind a flat product image and to help write copy. Jones Road Beauty’s Plofker says that while there was still “a long way to go” in terms of ideal quality, Meta’s integration can improve the brand’s ability to test what resonates, in addition to saving time creating various approaches to ads. For example, Jones Road quickly found that ads that emphasise a product’s moisturising capabilities didn’t land. He estimates that the time it would have taken to previously produce and test 10 versions now allows for 20 or 30.
Mendelsohn, of Meta, says this can be especially useful early on. Brands can “get a sense of what an ad could look like before investing in a full photoshoot,” she says. “By helping to accelerate the brainstorming process, advertisers can move more quickly towards green-lighting the ads that will drive the biggest impact.”
When you are layering on various texts and images, “now you are talking about the opportunity to generate so many variations of assets and you figure out what works. And, once it works, you can reallocate ad spend,” Tipograph says.
Finding the line in personalisation
Marketers have for years been navigating the tension between personalisation and privacy, and the same type of automated variety that creates, deploys and tests ads can also lead to an eventual one-to-one feel on e-commerce pages.
“Generative AI will help online shopping come closer than ever to replicating the joy of the physical store experience,” says Kelly Thacker, Salesforce SVP of product marketing and CMO of retail and consumer goods. Thacker predicts that this will give fashion and beauty marketers more opportunities to be more creative and personable, and to help shoppers feel more confident about their purchases.
Currently, personalisation in e-commerce is more “cohorted”, says The RealReal’s McCandless, meaning that a site might know that someone likes expensive handbags, so it will prioritise those while letting the customer refine through filters. However, a more unique approach could be informed by customer data to generate more specific imagery, recommendations, descriptions and web copy. Currently, “there are not enough design resources to generate all the content you would want to generate to be as personalised as you can be”. She notes that this approach will naturally introduce more conversations around privacy, and brands and their consumers will have to navigate the level of personalisation at which people are most comfortable.
Automatic doesn’t mean straightforward
While the immediate perk of fast-tracking image edits and copy creativity is obvious, the ethical, legal and longer-term considerations are less so. If a fast-fashion retailer informs a tool to make an ad in the style of Ralph Lauren, for instance, “things would start to get strange”, Mikmak’s Tipograph says. A similar concern comes with the copy conversation, in which a brand might unwittingly publish something that was copyrighted. This is in keeping with the ongoing and much larger conversation that big tech has been grappling with for years: just who is responsible for the content on tech platforms?
Google has applied safety measures such as filters for offensive keywords, and all images created through Product Studio pass through Google’s existing reviews for products, which face content standards, says Matt Madrigal, Google’s VP and GM of merchant shopping. Google is also building models to include watermarking (information that stays even if images are “modestly” edited) and metadata (which enables “additional context”) on AI-generated images.
For now, the brands and their creatives are likely to be responsible for any oversteps, Tipograph and Roy both say. “You have to imagine that tech companies, like most areas, will try to avoid responsibility on these,” Roy says. This is why some big companies are still reluctant to adopt this type of tech, Tipograph adds. She anticipates that the next two years will introduce litigation that ultimately results in more regulations. Still, “the advertiser has to be accountable. You will want to make sure the creative is original.”
Originality also extends to the concepts. The RealReal’s McCandless anticipates that as brands have more access to easily use “optimised” content, brands will have to intentionally tweak and personalise content to maintain differentiation.
That human touch is why she feels that generative AI tools won’t soon entirely obviate creative jobs. “I think — a lot of people believe — it is more augmentation and allows us to be better, smarter and faster and you still need humans to do the job and do it well.” Marketers should embrace it and not be afraid, as it allows them to focus on tasks that are more important, Plofker, of Jones Road, says.
Mikmak’s Tipograph has a more tempered perspective. “It would be my advice that you can’t let software eat our jobs, but rather harness the power of software to create the next generation of jobs. For creatives, it is absolutely a threat. So, the question is: how to become better at your job by adopting these tools?”
To receive the Vogue Business newsletter, sign up here.
Comments, questions or feedback? Email us at feedback@voguebusiness.com.
Generative AI is coming to Google search
The fashion exec’s guide to generative artificial intelligence
With generative AI, luxury hopes to finally crack digital clienteling
