This article on Estée Lauder and AI is part of our Vogue Business membership package. To enjoy unlimited access to our weekly Technology Edit, which contains Member-only reporting and analysis and our NFT Tracker, sign up for membership here.
Are generative AI tools more hype or help for brands? The Estée Lauder Companies (ELC), owner of the namesake beauty brand, alongside Tom Ford Beauty, Clinique and Mac, has found a number of practical business uses for the technology, using it to help measure customer sentiment, aid in writing copy and to potentially make product recommendations, says Gibu Thomas, EVP of ELC online.
Generative AI, which can “read”, interpret and communicate in natural, human-like language or imagery, has quickly captured the attention of fashion and beauty executives as new consumer-facing tools such as ChatGPT, Midjourney and Dall-e expand access to the average consumer. Brands have begun evaluating uses of the technology to aid in designing products, communicating with customers, and creating text and images. Tech companies, including ChatGPT owner OpenAi, Google, Microsoft, Salesforce, Amazon, and Shopify, have subsequently released tools for brands and retailers.
When Thomas joined ELC in November 2020, he identified AI generally as a capability to invest in. ELC has been exploring uses of generative AI for the past couple of years and has developed a set of responsible standards to serve as a guideline for how the company structures its initiatives. The company was able to build on an existing relationship with Google Cloud, which gave ELC early access to some of its tools. (For example, Google Cloud’s BigQuery helps companies build a data platform, Vertex AI enables retailers to build their own AI applications and models, and Imagen is a text-to-image foundation model.)
ELC was also able to recruit academics and scientists who are integrated throughout the business rather than siloed. “We had the luxury of building our environment from the ground up and not being constrained by legacy challenges,” Thomas says. Still, the company has largely kept these experiments quiet. “We want to make sure that the things that we’re talking about have real substance. It’s easy to get caught up in hype and want to be part of the conversation, but we want to talk about things when we have delivered business value.”
Reading and writing text
Google Cloud is seeing increased demand for and interest in generative AI from global retailers and consumer goods companies of all sizes, says its VP of strategic industries, Carrie Tharp. She identifies four potential priority uses: creating bespoke images and creative content (which can extend to one-to-one personalisation); conversational commerce and search (which can enable virtual try-on, customer support and more functions via multimodal interactions, meaning those that can process both images and text); customer service automation (like routing a complex query to an operator or scheduling an appointment); and new product development (including helping retailers prototype concepts).
ELC has been using generative AI to improve its customer service. After testing generative AI’s ability — specifically Google’s large language model called Palm2 — to analyse customer care calls and categorise the nature of the customer’s call or concern, it found the AI was twice as accurate at classifying calls than humans were, Thomas says. Now, the company plans to expand these capabilities to social media and other channels.
This improved capability allows the brands to resolve issues fast and glean insights that can then feed into functions including digital and physical product development, site merchandising copy and more, Thomas says. He adds that because of the existing data science foundation that ELC had invested in, the company was able to quickly test a proof of concept and then put this into production in less than a week.
“Ultimately, we’re seeing that businesses who are able to maximise the capabilities of AI and generative AI are those that have built a strong data foundation beforehand,” Tharp says. “If a brand has been lax or slow in finding, organising and centralising their corporate data in reliable ways, we encourage them to start now if they don’t want to be left behind.”
ELC has also worked with generative AI to help generate copy for use in functions such as search engine optimisation and website copy. Because nuanced details related to brand equity and consumer trust are so important, Thomas emphasises that ELC brands view this type of content as human augmentation rather than autonomous generation, with the tools being able to generate versions of existing text with different tones or lengths. “The technology is incredibly powerful, and it makes our talent much, much more productive,” he says. “And our brands obviously love it.”
Generating marketing copy was one of the most common early examples of how brands have used generative AI. Lingerie brand Adore Me found that it could aid in SEO-optimised product descriptions, saving up to 30 hours a month on copywriting and improved website traffic. Similarly, the brand enables human oversight by giving copywriters three choices and the ability to make changes before publishing.
Scaling personalisation while mitigating risk
Thomas says that ELC brands are now beginning to explore how generative AI tools can replicate or create a high-touch service online that mimics the experience provided by beauty advisors or beauty artists. “With generative AI, we feel that we can really bring that very high touch with high tech to reinvent that competitive advantage. We’re obviously excited about the possibilities, and we think that this has a potential to really redefine how we do business and how we engage with consumers,” he says.
Other brands have begun experimenting with AI-informed advice that more closely resembles natural speech. Kering’s experimental site KNXT tested a personal shopping chatbot called Madeline using ChatGPT to recommend products from Kering-owned brands. German e-tailer Zalando, multi-brand retailer Ssense, startup Sociate and payments company Klarna have also developed tools that use genAI in product recommendations.
Thomas notes that potential uses that are a progression of product discovery are more natural to how customers engage with humans. “If you think about how people search to discover our products [online] today, it’s very keyword-driven with product names and that sort of stuff. But, if you think about how a consumer engages with a beauty advisor, you’ll describe the issues you’re having.” Eventually, online shoppers might use natural language to find what they are looking for. A customer might be able to say, “I saw this incredible look in this movie character that I want to replicate for myself. How would I do that?” Thomas says.
While developments like these are currently technically feasible, he adds that the company takes a cautious approach, prioritising brand equity, brand voice and consumer trust over speed or experimentation. “We want to be very deliberate in the way we go after big opportunities. It’s not because we don’t have the ambition. It’s more because we want to calibrate the risk versus value equation.” In beauty, for example, if a consumer-facing tool that was trained on the data on the internet acted independently without human intervention, it could end up promoting harmful standards of beauty.
Part of the challenge of internal decision-making, Thomas has found, is having the discipline to anchor potential use-cases within the business’s values, integrating them throughout the business in a consistent way and even exhibiting restraint. “It doesn’t take judgment to say ‘yes’ to everything. But having the ability to say ‘no’ or ‘not yet’ to the right things? That’s really where that judgment comes in.” This also requires a hefty appetite for “fomo”, he acknowledges. “People feel like they just need to do something and put out a press release. Well, you know, if that was the standard for success, then we would all be living in the metaverse.”
Comments, questions or feedback? Email us at feedback@voguebusiness.com.
