Of the six concepts presented in the finals of the L’Oréal Brandstorm global competition, the one that relied the least on artificial intelligence won the prize. Beyond this anecdote, the decision says something about the limits that the fragrance industry itself places on the automation of creativity.
The jury president had just uttered the word that, better than any argument, summed up what had unfolded over the past three days: “soul.” Cyril Chapuis, president of L’Oréal Luxe, did not say that the winning team had the best model, the best algorithm, or the best data. He said that they had delivered “a fragrance with a soul.” This statement brought to a close a competition whose assigned theme was, in fact, the exact opposite: the 34th edition of Brandstorm challenged its contestants to reinvent luxury perfume through artificial intelligence-driven technology. Six teams, selected from more than 380,000 applicants from 64 countries, took the stage at VivaTech in Paris to tackle the challenge. Only one chose to say almost nothing about it.
One brief, six responses, one withdrawal
The range of proposals illustrates the extent of the gap. The team from Australia and New Zealand built its entire concept around an image recognition model capable of translating a personal photograph into a scent composition, explicitly drawing on Modiface, the algorithmic beauty infrastructure already deployed by L’Oréal. The Korean team designed an app-controlled skin patch capable of automatically switching scent profiles based on the wearer’s schedule. The Sino-Korean team proposed a home diffuser that learns its user’s sensory preferences over time. Three out of six proposals placed a machine learning system at the heart of the experience.
The American team, for its part, presented Prada Capture: soluble sheets to be placed on the skin, organized into three combinable accords—lily of the valley, peony, and honeysuckle—packaged in a case made entirely of recycled aluminum. The technical highlight of their pitch wasn’t a neural network. It was Osmobloom, a proprietary air-flow extraction technology from L’Oréal that captures floral molecules previously considered too fragile for distillation or conventional solvents. No sensors, no predictive models, no learning curve: a promise of unprecedented ingredients, and the act of creating a fragrance returned to the user, who is called upon to decide, each morning, which of their blends will set the tone for the day.
The action that the algorithm does not sign
This decision by the jury is by no means insignificant in an industry that, over the past two years, has seen a surge in announcements of predictive tools for fragrance creation. It indicates where, within the fragrance luxury value chain, artificial intelligence is considered legitimate—and where it is not yet. During the Q&A session with the finalists, the jury members never once asked the American team about the nature of their algorithm. Instead, they asked about the fragrance’s longevity, sillage, and business model. In contrast, when faced with teams that had built their architecture around a machine learning model, every technical question inevitably—almost mechanically—boiled down to the same query: who, ultimately, signs off on the composition? The Australian-New Zealand team had to clarify, right in the middle of the Q&A session, that the composition generated by their model was systematically reviewed and adjusted by a L’Oréal perfumer before being sent back to the user for approval. Artificial intelligence made the suggestions. Human judgment made the final call.
This hierarchy—the machine that sketches, the hand that cuts—is not merely a posture of caution adopted for the occasion. It also runs through the projects that L’Oréal Luxe has chosen to showcase on the sidelines of the competition, outside the official competition.
Details. Alongside Brandstorm, L’Oréal Luxe presented a preview of “Machine Dreams Rainforests” at VivaTech—a collaboration between artist Refik Anadol and the division’s international fragrance teams, led by Véronique Ferval, in anticipation of the opening of a generative art museum in Los Angeles. The project combines images generated by artificial intelligence with twelve living fragrances, delivered through a necklace connected to biometric sensors—heart rate, skin temperature, and skin conductance. Only two technologies support the olfactory creation itself: headspace, a form of olfactory photography that captures the molecules emitted by a living object before a human nose reconstructs the scent from this data, and Osmobloom—the same airflow extraction technology used by the winning team at Brandstorm. One of the three fragrances unveiled to the Parisian public, named Petrichor Memoria, combines geosmin—the molecule responsible for the scent of rain on the earth—with a specific variety of patchouli. Another recreates the scent of a cactus flower that blooms only one night a year, deep in the Amazon rainforest, captured during a field expedition. Even at the heart of L’Oréal Luxe’s most spectacularly high-tech showcase, generative data can never replace the expedition, the human touch, or the nose.
A Statement from the Jury Chair
One question remains—one that the ceremony did not address directly, but whose conclusion suggests: What are we really teaching the young talents we invite to a competition titled “artificial intelligence” if, in the end, it is its relative absence that we are rewarding? The simplest explanation—a technophobic or nostalgic jury—doesn’t hold up: Rachel Thornton, marketing director at Adobe Enterprise, the competition’s official technology partner, was among the voters. The most uncomfortable—and most likely—explanation is different: for L’Oréal Luxe, artificial intelligence remains a brief requirement, not yet a criterion for victory. It opens the door to the competition. It does not determine who emerges as the winner.
The fragrance industry has yet to fully come to terms with the implications of this tension. Business schools that train the next generations of innovation project managers each year would undoubtedly do well to focus on this specific paradox: not to pit technology against craftsmanship, but to observe, within the same jury, exactly where the former ends and the latter takes over. In Paris, that boundary shifted on the stage of a student competition. It will shift elsewhere next time—and the question will remain the same: who, in the chain, has the final say?
Cette publication est également disponible en :
