On the scenic Manhattan waterfront, amidst a men’s homeless shelter and a medical examiner’s office, work is underway at a luxurious lab where a summer plum is being transformed into fragrance code. This cutting-edge effort is led by Osmo, a fragrance technology startup aiming to revolutionize the scent industry with artificial olfactory intelligence. The company boasts quick fragrance compounding that can produce a custom perfume sample in just 48 hours, allowing clients to receive their orders in the time it takes to get an Amazon Prime shipment.
Historically, developing a fragrance is a drawn-out process. After receiving a client brief—often a mood, a memory, or a specific concept—perfumers typically engage in formulation trials that can span weeks or even months. They work through numerous modifications, or “mods,” meticulously refining each iteration to evaluate its balance, projection, and drydown. The raw materials themselves might require years of cultivation, with additional time needed for bottling, regulatory checks, packaging, and comprehensive testing. Consequently, a fragrance can take anywhere from six months to over a year to reach store shelves, particularly in the luxury sector, while variability in climate affects the aroma of fragrant materials year by year.
Osmo aims to disrupt this traditional approach by digitizing elements like the plum and accelerating key processes surrounding odor molecule analysis and transport. During a recent scent conference, I encountered the fragrance derived from this digitized plum, shared with me by an independent perfumer. As we gathered around the scent blotter, it sparked a debate among attendees regarding its character. Critics noted it felt too medicinal and too sanitized, questioning the absence of nuances like bruising or decay.
“Where’s the craftsmanship? Where’s the perfumer?”
Having judged numerous fragrances for international awards and worked on machine learning initiatives at various startups, I understand the appeal of well-crafted scents and the temptations offered by technological advancements. To me, the digitized plum revealed a striking aroma, albeit larger than life and possibly engineered. Its presence was unmistakable. However, a more profound question loomed: while AI permeates the world of fragrance, are we enhancing the accessibility of beauty, or are we stripping away its essence?
The integration of AI in the fragrance sector is already established, influencing the vast majority of the scents encountered by consumers. The leading four fragrance conglomerates—DSM-Firmenich, Givaudan, IFF, and Symrise—utilize AI in their production processes. Givaudan’s Carto system assists perfumers in polishing their formulas, while DSM-Firmenich’s EmotiON is designed to create scents aimed at enhancing well-being. These technologies are incorporated not only at technical hubs but also in fragrance education globally. The shampoos, conditioners, luxury fragrances, and household products found on shelves have all been influenced by these industry giants and the AI systems they deploy. Notably, Frank Voelkl, a principal perfumer at DSM-Firmenich and creator of popular aromas, applies AI in his daily creative tasks, illustrating its beneficial role in managing technical challenges, thus allowing him to concentrate on the more imaginative and instinctive facets of his craft.
Heather, an emerging perfumer in France, notes that the use of AI has become common among her classmates. “Most of my peers have employed AI for every project or inquiry. Gen Z interacts with it as a fundamental tool—unlike older generations who see it more as an auxiliary resource.” The distinction lies in Gen Z’s integration of AI into their creative workflow, employing it for material selection and refining fragrance profiles, contrasting with older generations who utilize it primarily for inspiration or research purposes. For newer creators, AI has evolved from mere assistance to a critical component of the perfumery process.
Professor Pierre Vouard from FIT acknowledges the mixed blessings of this technological shift. “The art of manually compounding fragrances, knowing the precise measurements of each component, could soon fade. But how essential is that to the craft?” He observes AI’s integration in his classroom, speculating that this technological development might democratize fragrance creation by significantly lowering production costs, yet it raises inquiries about artistry and the role of traditional perfumers.
“There’ve only been about 100,000 fragrances ever made. I want there to be millions.”
Perfumer Michael Nordstrand expresses similar concerns regarding AI-driven fragrance companies. He contends that these businesses bypass professionals and target consumers lacking the expertise to evaluate scents critically. Furthermore, he notes that Osmo has not disclosed crucial details regarding the datasets or formulas used in its fragrance models, despite repeated requests for clarity. When I inquired, Osmo did not provide specific metrics, only stating that it is still in the developmental phase of its system. Despite promoting its lead perfumer, Christophe Laudamiel, the company has not identified any other perfumers within its structure.
Historically, in 1995, fewer than 400 new fragrances were launched worldwide, while in 2023, that figure surpassed 3,000. Osmo’s founder, Alex Wiltschko, envisions this figure growing substantially. “There’ve only been about 100,000 fragrances ever made. I want there to be millions,” he asserts, emphasizing the importance of new tools in enhancing global beauty.
However, expansion comes with potential environmental repercussions. When discussing energy consumption, Wiltschko claims that Osmo’s graph neural networks require considerably less power in comparison to models such as ChatGPT. He states that the energy requirements are minimal and emphasizes the efficiency of their model, which can be trained in under an hour, unlike much larger models that take months. Nevertheless, he acknowledges that Osmo does not currently track its energy use and declined to provide lifecycle assessment data for evaluation against traditional fragrance houses.
So, is the energy consumption negligible or too ambiguous to measure?
Most consumers remain uninformed about the extent of AI already integrated into their beauty products, as well as its environmental impact. The lack of transparency is increasing. Some independent brands, like House of Bo, have begun utilizing deepfake AI technology in promotional videos, failing to disclose this to customers. “I feel condescended to,” comments fragrance consultant LC James, as he highlights the hidden labor and environmental costs involved. Moreover, some online retailers have exploited the likeness of perfumer Teddy Haugen in unauthorized promotions, creating videos where his appearance and voice are manipulated to promote fragrances he had no association with.
Fragrances have deep-rooted origins. For instance, orris root requires years to cure before it can be used, while sandalwood also demands extensive cultivation time. Natural materials need careful harvesting, aging, and blending. AI-driven labs like Osmo are capable of delivering customized samples within a mere 48 hours. This remarkable speed, while appealing, raises concerns about a growing detachment from the raw, organic beauty from which fragrances are traditionally derived.
Matt Belanger, co-owner of the New York City fragrance retailer Stéle, highlights this tension in the industry. “Many brands mislead by presenting themselves as perfumer-led while utilizing generators that replicate existing fragrances. The essence of fragrance lies in the time, courage, and significant effort involved in its creation, which starkly contrasts with simply pressing a button for quick results.” His partner Jake Levy adds, “Countless companies operate merely as an automated service. If transparency regarding AI usage were established, it would enhance our respect for the brands.” The team at Stéle diligently reviews the backgrounds of every brand they carry, emphasizing the need for more dialogue about AI’s role in perfumery. Nordstrand cautions, “If we don’t take charge and discuss AI’s place in this field, we risk letting it spiral out of control… it’s reminiscent of Jurassic Park—we were focused on whether we could, without considering whether we should.”
