Is There Such a Thing as Ethical AI?

If you follow any of the Estée Lauder Companies’ brands — MAC, Clinique or Deciem — online, you’ve already seen AI’s use in beauty marketing, whether you realised it or not.

The cosmetic giant has been quietly using the technology for tasks from editing images to assisting with copywriting in its digital campaigns, according to Yuri Ezhkov, vice president of the company’s Creative Center of Excellence. But wary of the customer blowback that hit brands early to generative AI such as Levi’s, Selkie and Baggu, Estée Lauder has opted for a slow and cautious approach.

It’s finally ready to come forward. Today, the Estée Lauder Companies announced its partnership with Adobe and its use of the software maker’s generative-AI platform, Firefly, in its digital marketing operations.

“We’ve been working closely with Adobe in partnership with all of our creative teams to identify where we can use the tool effectively, ethically and in line with any sort of local regulation,” said Ezhkov. “It’s been a 16-month journey.”

Beyond AI, the conglomerate is also on a turnaround journey. In October, Estée Lauder Companies announced its new chief executive, Stéphane de La Faverie, who just months on the job has had to deal with continuous disappointing results. In tandem with declining sales, the company announced last month that it plans to cut up to 7,000 employees.

If the company is hoping AI can help it work more efficiently, it isn’t alone. Brands have eyed the technology as a way to quickly produce text, imagery and video for purposes such as design and marketing. Volvo just aired a new commercial that it created entirely with AI (and which features no cars).

But along the way, consumers have voiced growing concerns about AI’s harmful effects, such as the job losses it could cause among human workers and the impact on creatives whose work AI developers have used without consent to train models. The situation raises a dilemma for businesses that are attracted to the technology as a way to streamline their operations but also see a reputational risk that could lose them the trust of shoppers.

ELC thinks it can thread the needle. It requires putting restrictions on what the company will and won’t do with AI.

Ezhkov said its brands do not use AI to generate images of people, for example. Models and other talent are off limits, as are images illustrating a product’s efficacy, in part because the company doesn’t want the AI perpetuating biases like unrealistic beauty standards.

“We haven’t replaced any main campaigns with AI images,” he stated. “And there’s no plan for that.”

Where they do see opportunities to create imagery, however, is when they want to showcase ingredients in their products that might be costly and time consuming to obtain otherwise, such as a plant that grows in a distant region, according to Ezhkov. (One reason it chose Adobe’s AI for image generation at least is because, unlike many other companies, it trained its AI on images it has rights to.)

Similarly, while the company uses an enterprise version of ChatGPT that’s been trained on its brands’ voices and styles to help with marketing copy, it isn’t writing any copy on its own. Instead, it’s meant to help give the copywriters direction and ideas to help with their workload. Ezhkov said employees have generally been eager to use the AI tools. Some feel like knowing how to use them will provide a path forward. Still, he was unclear about who on the creative teams might be affected by greater company cuts.

Making manual processes faster and more efficient is where much of the company’s focus on AI lies. The brands of the Estée Lauder Companies sell in 150 countries and territories globally, and they try to tailor their marketing at a local level, since they can have different positionings even in neighbouring countries, and those countries may have different regulatory guidelines about what claims a beauty product can make.

It’s practically inevitable that companies using generative AI will run into thorny areas in the uncharted frontiers the technology is opening. One subject businesses are weighing is when to label content as being created with AI. (Last month, Meta said it would begin labelling AI-generated images on its platforms, so it could end up doing some of the work for them.) There are no clear rules on what merits a disclosure. Ezhkov gave the example of an Adobe Firefly tool that lets you extend the background of an image. If you expand an image by 20 pixels to fit a banner format, should that be labeled? For the moment, ELC doesn’t think so.

Beauty companies, of course, have also edited photos for years without the benefit of AI. Even with manual tools like Photoshop, brands often erase imperfections and retouch images in all sorts of ways. It’s only become easier to do so, or even to generate entire campaigns with AI.

Though marketing is just one industry experts believe could be dramatically reshaped by the technology, Ezhkov argues companies can’t take a fear-based approach.

“People were afraid of Photoshop when it first came out, and here we are 20 plus years later, and it’s sort of the backbone of the entire industry,” he said.