Welcome to Beauty Run by Robots, a Vogue Business mini-series exploring the role and effects artificial intelligence (AI) will have on the beauty industry.
Skincare brand SmartSkn’s pop-up lab in Los Angeles now features artificial intelligence-driven skincare robots by Korean tech firm Lillycover, promising personalised skincare in three minutes. At the Consumer Electronics Show’s 2025 tech showcase, agentic AI (a form of AI capable of independent action and learning) dazzled execs with its potential to reshape everything from robotic surgery to automated marketing analysis. I’ve gotten a 20-minute gel manicure from Umia’s salon robot — a fraction of the usual time with results lasting two weeks.
It’s clear that AI is coming for beauty. Yet, these innovations remain isolated flashes of the future rather than industry standard. Salon treatments still take hours, hyper-personalisation remains limited by mass-market formulations and retail layouts dictate shade range availability. Most brands rely on traditional ad campaigns rather than AI-driven, one-to-one customer interactions. AI bias remains a major industry concern — many models struggle to thoroughly analyse diverse skin tones, leading to flawed recommendations accurately.
But those that embrace AI early will be better positioned to lead. It also can pad out a brand’s bottom line in the long run. Global management consultancy McKinsey reports personalisation can reduce acquisition costs by as much as 50 per cent, lift revenues by 5 to 15 per cent and increase marketing ROI by 10 to 30 per cent. Plus, companies with faster growth rates derive 40 per cent more of their revenues from personalisation than their slower-growing counterparts.
The key question: how does the industry move from A to Z, from promise to execution, ensuring the future flashes become real life? “The industry isn’t ready yet,” says Sampo Parkkinen, CEO of AI firm Revieve, thanks to existing bias and the many manufacturing and logistical challenges at play.
Still, progress is underway. Companies like Haut.AI are refining skin analysis with its system that runs 15 different skin models; in parallel to analysing various facial features combined in one report for greater skin-condition accuracy. L’Oréal has a dedicated generative AI committee, while Renude, an AI data science firm, is working on more inclusive data sets to help mitigate AI bias. But for AI to revolutionise beauty, scalable solutions must be in place.
All down to data
Artificial intelligence is only as good as the data it’s trained on. Skin tone detection must be flawless before virtual mirrors, smart glasses and skin analysis tools hit the masses. However, gaps remain, especially in skin tone analysis, which pose a potential minefield. “There are a lot of companies in the space today doing virtual try-ons or skin analysis but many are regionally focused, which increases the risk of AI bias — brands need to watch out for this,” says Parkkinen.
“AI bias skin tone analysis is largely due to imbalances in training data sets. Many AI models have been developed using data sets that skew towards lighter skin, which can lead to inaccuracy,” says Majad Hussain, co-founder of AI-powered skincare brand MiQuest. It all starts with imaging. Dermatology reports from the National Institute of Health, Oxford Academic and the Journal of the American Academy of Dermatology, dating back to 2022, reveal that biases within data sets stem from the lack of medical images accurately representing darker skin tones with medical conditions.
“AI is built from data, so if there is bias in the data, this same bias will be built into the AI. It is up to the team building the AI model to consider the biases that exist in the real world and take action to mitigate them,” says Renude co-founder Pippa Harman.
When building and scaling AI models, brands need to assess all skin types, which requires high-quality, varied data and skin expert assessments. It’s an investment Renude doubled down on in its AI skin analysis designed to personalise skincare routines combining AI imaging, a custom quiz and product recommendations from licensed aestheticians for consumers and retailers (the firm licensed its services to French Pharmacy-owned beauty brand Laboratoire SVR in 2023) to carve out products better suited to their skin type, tone and possible conditions. The analysis tool was developed with Harley Street consultant dermatologist Dr Justine Kluk and trained on tens of thousands of images across a spectrum of tones and skin condition severity levels like eczema or scarring. Each image was labelled by type and skin tone by licensed doctors before feeding into the AI, ensuring precision.
Hussain urges brands to go beyond colour-based analysis. “Brightness and contrast alone don’t work for darker skin tones. AI should analyse texture, shape and depth,” he says. Explainable AI (a type of AI that can explain how it makes decisions), which shows how a model reaches conclusions, can help identify and correct biases. Advanced imaging technologies like vascularity detection, thermal mapping and depth sensing further refine accuracy by reducing reliance on brightness-based cues. Whereas L’Oréal’s global director of beauty tech services Béatrice Dautzenberg, emphasises having a team that spans age and race as an important consideration for brands integrating advanced data sets into AI analysis tools. “Having a diverse team of humans coding AI’s prompts is very important as we move forward.”
Regular testing is non-negotiable. “Brands must continually evaluate AI models against diverse data sets and retrain them when discrepancies arise,” Hussain says. Tools like IBM’s AI Fairness 360 help track and mitigate bias, ensuring more equitable outcomes.
Moving marketing forward
Elsewhere, advancing models beyond data capture to help brands transition from general to personalised marketing campaigns — and at scale — will be required. Experts say this investment could advance campaigns beyond pocketed consumer segments to tailored ads. “Where I’m bullish on AI is enabling a feedback loop between brand and consumers. A conversational AI where consumers can interact with brands to act as a two-way communication channel,” marketing firm Front Row’s chief insights officer Mark Wieczorek says.
For Wieczorek, AI shouldn’t be used throughout the entire marketing funnel to facilitate hyper-personalisation in ads because the overarching goal of delivering one message at the top of the funnel will become too complex and costly. Instead, Wieczorek advises brands to focus AI efforts on mid-to-lower funnel marketing, where consumer data is at its richest. “AI is most effective when fine-tuning messaging, timing and channel placement,” he explains. Scaling AI-driven personalisation also requires infrastructure changes. “Brands need to blend data science with creative intuition and restructure teams accordingly,” says Coralie Hampson, business director at Socially Powerful, a social marketing agency. Still, broad ad campaigns won’t disappear. “They start the conversation,” Wieczorek notes, “but AI refines it.”
Yet, data privacy remains a challenge. “Consumer data should be treated as confidential, but AI tools often involve third-party developers and cross-border data storage,” warns Mona Schroedel, AI and data protection lawyer at Freeths. She urges brands to scrutinise contracts and enforce strict data access and usage policies. Despite these hurdles, costs for AI adoption are expected to drop. “Soon, an AI-powered model could cost less than a website refresh,” Wieczorek predicts, making advanced personalisation widely accessible.
Carving out customisation
Hyper-personalised beauty requires structural change. Some brands have tested using AI for tailored product formulations, such as SmartSkn’s use of AI to customise skincare, or Dr Simon Ourian’s employment of AI skin analysis with voice integration for an in-clinic experience and regimen recommendations. Function of Beauty creates bespoke haircare using a questionnaire algorithm, while dermatology-backed brands like Skin Me and Dermatica combine an AI algorithm with expert oversight for prescription-based solutions.
Still, scaling customisation is complex if brands want to move away from products dictated by traditional groupings like ‘sensitive skin’, ‘oily combination skin’, or categories such as ‘hydration’ and ‘brightening’ and calibrate systems based on skin needs. AI-based questionnaires are a starting point, but brands will need to re-think their operational flows.
“AI-powered diagnostics, flexible manufacturing and smarter ingredients selection are key,” says Hussain. Modular product systems, where ingredients are mixed based on AI assessments, could replace bulk production. Biometric data tracking could further refine AI-driven recommendations. But true personalisation demands cross-functional collaboration. “Tech, research and development, and customer experience teams must align. Personalisation can’t be an afterthought,” says Ruth De Leo, CEO of AI-powered cosmetics brand Dcypher.
Dcypher had to rebuild its entire infrastructure to scale customised formulations. “We had to build an entirely new infrastructure that could manufacture individualised formulas on demand, without generating excess waste or driving up costs. Balancing affordability, efficiency and true customisation was a challenge that traditional beauty brands don’t have to navigate,” De Leo says.
She also warns about costs from an AI investment perspective, as brands can’t afford to cut corners with advanced technology. “Developing AI-driven skin-matching technology, ensuring our database was inclusive of all skin tones and building the right manufacturing processes required resources,” she admits. But, the biggest challenge? Consumer education. “Shoppers are used to pre-made shades. We had to show why bespoke beauty is better and just as accessible.”
While hyper-personalisation won’t reconfigure beauty retail overnight, AI can enhance in-store experiences. First, “retailers can introduce AI-powered customisation on counters — think bespoke foundations or virtual try-on mirrors”, says Harman. For her, multi-brand retailers Harrods and Selfridges are already revamping beauty halls to accommodate AI-driven services from brands like Prada and Lancôme, who offer skin analysis at the beauty counter. Next, Harman recommends buyers and merchandisers leverage AI listening tools to analyse consumer spending and footprint in-store. The move will better position retailers and brands to stock what drives footfall and sales — while maximising shelf space — to advance merchandising and a consumer’s shopping experience overall.
To scale, the road ahead requires brands to rethink operations, navigate regulatory concerns and invest in AI’s long-term integration, but the cost of sitting out could be greater. “We’re on the cusp of a massive shift,” says Hussain. “AI is the future of beauty — brands can’t afford to be left behind.”
Comments, questions or feedback? Email us at feedback@voguebusiness.com.
