Paris—Artificial intelligence is becoming unavoidable, even in the subjective and sensorial world of cosmetics and personal care. And though discomfort over what the technology might mean for human employment is a part of that conversation, many cosmetic chemists, chemical executives, and consumer product marketers are also excited about the opportunities it offers to innovate faster and give them more time for the fun parts of their work.
For Rita Machado, a chemical engineer at the contract formulation firm Medinfar, AI is interesting as a way to widen the pool of data she can use and to eliminate some of the grunt work of combing through spreadsheets and documents. Machado, who also has had success using AI tools to screen for ingredient compatibility, for example, spoke with C&EN at the in-Cosmetics Global 2026 conference, held in Paris this year in mid-April.
Some chemical makers talk about AI tools that can actually create formulations for customers, work currently done by experts like Machado. She is watching for that possibility but not sweating in the near term. “I don’t think AI can formulate, not yet,” she said. “For now, it is more of a helpful assistant.”
Worldover and several other providers of artificial intelligence software had booths and gave talks at in-Cosmetics Global 2026, held in mid-April in Paris. Credit:
Craig Bettenhausen/C&EN
Marilou Lemaire, a chemist at the formulation firm Oriflame Cosmetics, is concerned that the guarded adoption of AI she sees in her industry today could accelerate, leading to replacement of human formulators. The sentiment among her peers about AI depends a lot on what tools they’re being asked to use, she said. Oriflame encourages its staff to use AI-enhanced search and database tools but prohibits the use of AI to build formulations, she said, an approach she’s happy with.
The talk of replacing human formulators is alarming, said Amanda Witter, a cosmetic chemist working on sun protection at the contract beauty product manufacturer Meiyume. But success for a cosmetic or personal care product hinges on analog sensations such as texture, smell, and skin feel that are hard to reduce to digital data. AI can assist with administrative tasks, help analyze ingredient lists, and play other support roles, she said. “But do I think it can replicate a sensory experience that I would want to deliver to another human? No.”
Most AI tools, at their core, do statistical modeling, suiting them to interpolate within existing data sets, said Jeffrey M. Sanders, a research lead for materials science at the computational chemistry firm Schrödinger. The problem, he said, is “almost no one has enough data to train AI models that are truly predictive and useful.” Sanders said his team was able to build an AI tool that suggests formulations built around biobased pigments, but that was only because it was able to construct an unusually rich dataset on that topic from catalogs and peer-reviewed publications by the chemical maker BASF.
Molecular simulations, Schrödinger’s bread and butter, are grounded in empirical chemical data and quantum mechanical calculations. That makes them better than AI for extrapolation tasks such as projecting how a formula will behave over time or exploring new chemical combinations, Sanders said.
Ingredient suppliers with long histories have an advantage with AI because of the data they’ve collected over the years, said Griffin Gappert, head of R&D for consumer and life sciences at the chemical maker Nouryon. Though the firm technically dates only to 2018, when it spun off from the coatings firm Akzo Nobel, its care chemical group has more than 35,000 sample formulations that it has collected over more than 100 years. That’s enough data to make the AI useful, Gappert told C&EN.
Nouryon worked with the AI developer Albert Invent to create two connected interfaces based on its database. The customer-facing version, which the firm debuted at in-Cosmetics Global, combs through Nouryon’s formulation library to find matches to plain-language searches from users. It has guardrails so it returns only existing recipes, Gappert said. An internal version used by the firm’s formulators can create new concoctions, he said, but customers won’t see them until they’ve been made, characterized, and evaluated in real life.
Some AI providers see the relatively light touch of AI in personal care as merely a transition state. Yann Chilvers, cofounder of the ingredient software firm Covalo, said during a talk at in-Cosmetics that the AI agents coming soon will be so powerful that companies will overhaul workflows around them. Up to now, he said, companies “tend to plop AI on top of their existing process and expect magic to happen. You have to redesign the system around the new tools.”
Chilvers said the result will be more time for humans to focus on high-value work. But on a less comforting note, he advised the audience to “become the human layer that makes AI valuable.”
AI will be most useful, and commercially successful, when it takes the drudgery out of personal care science and leaves the fun parts to the humans, according to Robbie Curwen of the AI software developer Worldover. Formulation scientists tend to be the day-to-day users of the firm’s AI tools, he said, and it wants to make them happy. “The core thing they enjoy is actually mixing and trying things,” Curwen said. The firm’s offerings focus instead on tasks that software can get out of the way, such as data and document management.
Humans need to be central, not just “in the loop,” even at an AI firm, Curwen said. Worldover staffers do most of the work of translating and inputting a client’s legacy data, he said, because the AI performs well only after the data are clean, fact-checked, and organized in a standard format. For similar reasons, human staff vet and enter new regulatory information into its compliance software. AI agents that scrape government websites are prone to pick up proposed rules along with finalized regulations, Curwen said.
The biobased-chemical maker Vantage Specialty Chemicals struck a cautious tone toward AI at in-Cosmetics. In a talk, Beto Pino, the firm’s vice-president for technical marketing and innovation, said he’s seen AI database tools shorten the ingredient selection process from hours or days to minutes, AI chatbots generate sales, and AI factory management suites make useful contributions to process optimization.
At the same time, Pino said, AI “photos” of cosmetic products push unrealistic and Eurocentric beauty standards. For example, in the real world, 2–3% of people are blonde. But AI will draw a blonde model 37% of the time. And although 80% of the real human population has brown eyes, he said, generative AI puts brown eyes on only 30% of its models.
In addition, because consumers can’t gauge the efficacy of a cosmetic from an artificial image, the use of AI models damages trust. “How can we talk about skin and hair when the models aren’t even real?” Unsurprising then, he said, is the trend this year toward advertising that emphasizes realness: scratched, half-empty bottles in lived-in bathrooms, photographed with grain and lens flare.
Trust and connection are what makes for repeat customers instead of one-time samplers, Andrew McDougall, director of beauty and personal care at the market research firm Mintel, said during a talk. Visible human effort and small, endearing mistakes have become critical elements in personal care marketing.
AI has helped formulators increase their success rate for new products, McDougall said. Many, in fact, would struggle to go back to a world without AI. But because the tools inherently draw on existing creations, AI is creating a sameness that turns customers off and stops brands from standing out.
“When everything looks the same, it becomes interchangeable,” McDougall said. Cost becomes the only competitive factor, and your product becomes more likely to end up in the discount aisle. AI has become a part of life in the personal care and cosmetics industry, and it can be a huge help, he said. “When we remove variance and the human element, that’s when it becomes troublesome.”
2026 American Chemical Society
