Beauty is no longer in the eye of the beholder; it’s at the fingertips of the generative AI (gen AI) prompter. Gen AI could add $9 billion to $10 billion to the global economy based on its impact on the beauty industry alone, and early movers have already begun testing the technology. But scaling these experiments will be challenging, given the velocity of gen AI innovation.
The gap between the laggards and leaders in the beauty industry will only grow once leaders successfully deploy gen AI at scale. The fast will become faster, more responsive, and better equipped to anticipate and deliver what consumers want, while those left behind may find it harder to hold on to slivers of market share.
Beauty players who focus on priority use cases and customizing gen AI to meet their needs can help realize the technology’s full potential. This article outlines four gen AI use cases that beauty players can prioritize, explains how to bring gen AI to the organization, and lays out a set of imperatives to support the use of gen AI in beauty over the long term.
Four use cases of gen AI in beauty
More than a dozen gen AI use cases that apply to the broader consumer sector also apply to beauty. These use cases span the organization from front to back, including functions from user experience to customer support (table).
Table
Beauty players can use generative AI across their consumer-facing functions and within their internal value chain.
To prioritize the use cases, we consider that the beauty sector relies on speed in bringing products to market and responding to consumer feedback. On that basis, four gen AI use cases are likely to have the greatest impact: hyperpersonalized targeting, experiential product discovery, rapid packaging-concept development, and innovative product development. These employ gen AI tools at various stages of adoption. Some (for example, gen AI customer chatbots) are already in fairly wide use among beauty players, whereas others are nascent but promising.
Hyperpersonalized targeting
One of the most important moves a beauty brand can make to survive in the competitive beauty sector is to develop a unique value proposition. But beauty players must also ensure that the products they have thoughtfully positioned reach the consumers who will be most receptive to them.
Today, most beauty companies can afford to target only a handful of consumer segments because they have limited capabilities to personalize messages on a bigger scale. This broad approach to consumer segmentation leaves much of the market untapped. But with gen AI, beauty brands can create hyperpersonalized marketing messages, which could improve conversion rates by up to 40 percent, based on our observations.
AI can analyze large consumer data sets, detect patterns, and create microsegments based on pattern recognition algorithms. From there, a beauty brand can train its gen AI platform using a variety of inputs, including customer data, inputs that describe the brand voice, and product information. When entering new markets, beauty brands can train gen AI models on internal product data as well as external market research, such as customer surveys. Gen AI can then create and test variations of text and images to see what resonates best with each consumer segment.
Consider the hypothetical automated texts that might be delivered to an imaginary customer named Camille. The beauty brand knows that Camille lives in France, has a low annual spend, and recently purchased a face sunscreen. Camille has responded positively to promotions in the past. Before gen AI, an automated text to Camille might say, “Exciting news! New products are here. Take up to 20 percent off when you shop sale.” After gen AI, the automated text might say, “Bonjour, Camille! Did you know that our special cleansing foam for face sunscreen removal is now 20 percent off? It will pair perfectly with your recent face sunscreen purchase.”
Marketing specialists should review AI-generated messages before they are sent to ensure they reflect the brand’s ethos and value proposition while avoiding plagiarism or potentially harmful connotations. Some messages that seem innocuous can be detrimental to a brand’s image. In the previous example, the gen-AI-created greeting might have said, “Good evening, lovely lady,” instead of “Bonjour.” A customer may find the tone of this message offensive or inappropriate, or the message might be at odds with the brand’s overall ethos. The marketing team should deliver feedback to the gen AI model—perhaps rating its outputs with a thumbs-up or thumbs-down mechanism and entering detailed comments in free-text fields. The gen AI platform can then process the feedback and convert it into new training data.
Beauty brands will also need to integrate their gen AI models with assets from their digital-asset-management (DAM) systems, which serve as the repository for all the digital creative assets a brand uses, as well as integrate the models with the brand’s campaign management tools. Gen AI can categorize the creative assets in the DAM system—a task that would otherwise have to be done manually. This automation frees up time for the marketing team to focus on higher-value tasks.
Even as they continue to work with marketing agencies to develop their brand strategy and deliver specialized campaigns, large beauty enterprises might consider investing in in-house hyperpersonalization capabilities. This would offer two main advantages: companies can use their own consumer data to train gen AI models, and they can create and test personalized communications with greater speed and agility.
Experiential product discovery
Despite the tech-powered innovations in consumer product discovery over the past few years, there is ample room for improvement. The first generation of consumer chatbots, for instance, provide relatively rigid answers and can be frustrating for consumers to use. When a consumer asks for a recommendation for a new blush for a darker complexion, for example, a chatbot might give a generic list of products, rather than personalizing the conversation for a specific shopper and engaging them in deeper conversation. Virtual try-ons are helpful but can be glitchy or fail to accurately reflect what a product would look like on a consumer. In these cases, online purchases often can lead to costly returns, since returned beauty products generally cannot be resold.
Gen-AI-powered chatbots can help improve the shopping journey and decrease the likelihood of returns. These large language model (LLM) chatbots, which are trained on product data and consumer preferences, can respond to a wider variety of questions and offer more personalized recommendations, both of which can improve conversion rates. One global lifestyle player developed a gen-AI-powered shopping assistant and saw its conversion rates increase by as much as 20 percent.
The virtual try-on experience—which has already proven successful in other consumer categories, such as accessories and eyeglasses—might also be enhanced with gen AI. Using the same technology that powers image-generating gen AI tools, consumers can see the look of different products on their skin in different settings or see the potential benefits a product could have to their appearance over time. An online shopper who wishes to lighten dark spots, for example, could virtually try on a brand’s spot-lightening serum by uploading a photo on a beauty player’s website and running a simulation of the serum’s possible effect on their skin over several months.
Gen AI could also enhance experiential product discovery in physical stores. Today, interactive touchscreen monitors in stores can show products available both in-store and online, allowing customers to browse through SKUs, select items they want to see in person, or scan QR codes for exclusive offers. Even with their limited functionality, these screens have been shown to improve the in-store shopping experience and conversion rates. Gen AI can boost the effectiveness of these screens. For example, when a shopper who has location services enabled on a beauty player’s app walks into the company’s store, gen AI could generate content personalized to that consumer based on customer profile and purchase history. Given what we know about the effectiveness of personalized content, these principles could translate to the store setting, though large-scale implementation hasn’t happened yet.
Rapid packaging-concept development
When evaluating a beauty product, consumers consider both the product itself and its branding and packaging. Beauty brands typically spend months developing new branding and packaging concepts—a process that typically requires designers, copy editors, strategists, and packaging experts to iterate on ideas.
Gen AI wouldn’t necessarily eliminate this process, but it could dramatically accelerate it. Here’s how it could work. A packaging designer asks a gen AI platform the following prompt: “Show me five packaging options for a nighttime moisturizer, emphasizing skin care benefits and sustainable packaging materials.” The designer then modifies the gen AI platform’s output based on information about customer preferences, which could come from focus groups and customer surveys. Next, an advertising designer uses mockups of the new packaging in digital advertisements to test whether the images appeal to consumers, based on online engagement with the new ads. That data is then used to further refine gen-AI-powered concept creation and prototyping. With this basic approach, one beverage company reduced its concept development time by 60 percent.
Innovative product development
Creating new beauty product formulas is a multiyear process. It requires beauty players to partner with laboratories to research ingredients and experiment with formulas to determine the safety, stability, and efficacy of a new product.
Gen AI can speed up this process. A gen AI model—once it has been trained on a beauty product’s bill of materials, raw material usage, process parameters, internal research data, and other data (such as product patents or previous product trials)—can identify the ingredients that may be best suited for a new product, predict the product’s benefits, and recommend formula recipes.
Returning to the example of a nighttime moisturizer, assume that a formulation scientist could prompt the gen AI tool to create a new formula that emphasizes neuropeptides, a popular skin care ingredient, and prioritizes anti-aging benefits while also reducing formulation costs. Once the tool creates a potential recipe, the scientist would run lab tests to assess the compatibility and stability of ingredients in the formulation, as well as additional safety and consumer testing and clinical trials, if applicable. Formula iteration would continue based on consumer feedback.
While the physical testing process will still take time, McKinsey analysis has found that gen AI tools can reduce the time it takes to research new products from weeks to days. This can help save up to 5 percent on raw materials when developing those products.
Buy, borrow, or build?
The market for gen AI enterprise platforms is growing. But which approach—if any—is best suited for beauty players?
Organizations can bring in gen AI tools in any of three ways—what we call the taker, shaper, and maker approaches. Most beauty players likely won’t take the maker approach, where companies build their own LLM models from scratch. That would require capital expenditures and talent investments greater than most beauty companies can justify; it could also unhelpfully dilute a beauty player’s focus on its core competencies. However, beauty players can still get value out of the two other approaches:
Taker approach. The taker approach entails integrating off-the-shelf third-party gen AI solutions into a business’s workflows, with little to no customization. This is the least costly and resource-intensive of the three approaches, so it is an attractive option for beauty brands that rely on retailers for distribution (and therefore have less consumer data with which to customize models), have less tech talent, or have less cash for investments.
In evaluating a gen AI tool or platform, beauty players should ask questions such as the following: What are the data privacy and encryption protocols in place at the vendor? Will the vendor use the brands’ data to train third-party or first-party proprietary models? Who owns the copyrights to the outputs? How easy is the integration with the beauty player’s internal systems? (For example, does the vendor have an Application Programming Interface? Are they integrated with players like Google Analytics to enable broader use cases?)
Piloting the tool is crucial, of course. Most reputable gen AI vendors offer a low-cost pilot for a limited time—usually around one month.
Shaper approach. Being a shaper means training third-party gen AI models on the company’s own data and insights related to specific geographic, sector, organization, and business case needs. For example, for hyperpersonalized targeting, the data may include information about a brand’s voice, customer demographics and preferences, or successful campaigns. For innovative product development, raw data from clinical test results could help train models.
Larger beauty brands or retailers with a wealth of consumer data may choose the shaper approach. They will need a bench of tech talent that can add new components to the gen AI tool, integrate it into existing workflows, and deploy it across the organization.
Beauty players can use a mix of the taker and shaper approaches to gen AI, depending on their specific needs and use cases. Speed—in getting to market and responding to consumer demand—is particularly important for beauty players. For this reason, beauty organizations should consider modular gen AI components, which make switching between LLM providers easier to do, so scaling is easier. Gen AI may enable streamlining and automation in beauty, but the industry is as much science as it is art; it will be critical to keep a human in the loop to check for risks and inject uniquely human creativity into, say, marketing and packaging design.
How to implement gen AI at scale
To outcompete in digital and AI, consumer-packaged-goods players should consider critical questions such as “Where is the value?” and “Are leaders from the business side actively part of the transformation?” In addition, beauty players can take four steps to truly integrate gen AI into the business:
Align leadership on the vision, value, and road map. To move from experimenting to scaling, beauty players should identify which of the four use cases described earlier in this article will yield the greatest revenue lift, time and cost savings, and customer experience impact. To calculate this potential and then shape the road map accordingly, executives must bring together leaders from across various functions, such as marketing, customer service, and product development.
Bolster capabilities. As promising as gen AI is, using it effectively over the long run requires that beauty leaders assess how it fits into and is supported by the organization’s capabilities, including its operating model, data and technology practices, and talent. Companies should set up cross-functional teams to evaluate the organization’s existing capabilities and requirements for additional capabilities. These teams should deploy upskilling programs that help address capability gaps within their ranks.
Test, learn, refine, repeat. Beauty players should test gen AI’s output in controlled settings to determine what’s working. For example, in marketing use cases, a beauty player would select a channel—such as email, SMS, or paid media—and use A/B testing to measure the effectiveness of a gen-AI-created advertisement, both quantitatively (using metrics such as sales impact or click-through rates) and qualitatively (by asking questions such as “Did the ad feel true to the brand?”). The gen AI platform could then be trained on these learnings to produce better results in subsequent tests.
Adopt a risk framework. Beauty products often resonate with consumers on an emotional level. Because of this and because of the highly social nature of the category, beauty players must institute firm guardrails to prevent and contain risks involved in using gen AI. This risk framework should consider the interpretability and reliability of gen AI outputs, security threats, impaired fairness or bias, infringement on intellectual property, risks associated with using third-party AI tools, and privacy concerns. Gen AI should augment—not replace—the work done by a beauty player’s marketing or product development teams.
While much of the beauty industry’s products are cosmetic, gen AI applications in beauty are more than skin-deep. Integrating the technology alongside other digital and AI tools and boosting organizational capabilities can differentiate leaders in beauty for years to come.