Artificial Intelligence has learned to make perfume. The first AI fragrance from the collaboration between IBM and Symrise hit the Brazilian market in 2019. In 2023, all perfumers at the major scent companies started routinly using AI when working on formulas. By the end of 2024 you can launch your perfume brand with AI in 2 weeks. What does it mean for consumers and professionals? Analysis of a trend that will shape the future of the perfume industry.
Philyra: when big data & big fragrance come together
Today, all major international fragrance companies employ AI technologies. Germany-based Symrise was one of the first to explore the opportunities offered by artificial intelligence. Its AI is named โPhilyraโโa term rooted in ancient Greek meaning the โcompanion of creationโ. This algorithm is a sophisticated iteration of IBMโs Watson, an enterprise-focused AI platform capable of serving multiple industries. Watson first catapulted into public consciousness in 2011 when it won on the game show Jeopardy. It later morphed into Chef Watson to create culinary delights.
These public-facing ventures, however, primarily functioned as PR stunts. Watsonโs main utility is less glamorous but more impactful, invisibly powering B2B applications require processing and analyzing immense volumes of data. One such data-intensive application is perfumery, where IBM has formed a strategic alliance with Symrise, one of the Big Five of the fragrance industry.
This partnership signifies what Symrise regards as the most groundbreaking transformation within the perfume industry since the 1980s, when synthetic molecules were first introduced into formulas (another interesting recent innovation is the development of a new generation of water-based perfume).
Philyraโs prowess lies in its ability to seed up the typically long process of scent creation. What used to take years, can now be done in mere days. The algorithm ceaselessly iterates and optimizes new versions and can act as a limitless source of innovation.
Symrise is a hotbed of innovation even if you donโt take AI into account. The firm recently investment $6 million into Phlur, a startup thatโs redefining the perfume market by tailoring its eco-friendly and health-conscious fragrances to individual consumer tastes, and exclusively offering them online.
This strategic investment paid off when Phlurโs Missing Person fragrance ignited a TikTok craze, propelling the brand into the global spotlight among millennials. Itโs a pity the brand is only selling in the USA.
Coming back to AI, Symrise is hardly alone in its quest to harness its power for olfactory innovation. All major scent producers across the globe, from Switzerlandโs Givaudan with its Carto program to Geneva-based Firmenichโs Scentmate, and American IFFโs Codex, are investing in artificial intelligence. They regard AI as an indispensable foundation for shaping a profitable, future-proof trajectory in the fragrance industry.
How does creating perfume with AI work?
IBM Research and Symrise are teaming up to create perfume with artificial intelligence. Theyโre using Philyra AI tool to sift through Symriseโs massive database that has been built over 200 years of working in fragrance and flavors industry. The AI digs into 3.5 million chemical formulas, as well as information about how well these formulas sell in different countries and among different demographic groups. In addition, the algorithm remembers what scent family the formula belongs to (woody, floral, etc.), what gender itโs most commonly ascribed to, and even the degree to which itโs perceived as pleasant.
The AIโs analysis uncovers what are known as โwhite spacesโ or โempty spacesโโthese are unexplored combinations of chemical compounds that havenโt been used in existing perfumes. Philyra takes advantage of these gaps to generate innovative suggestions that have a high likelihood of resonating with consumers.
For instance, when tasked with updating a 12-year-old fragrance, the algorithm strategically removed elements that were considered dated and increased the concentration of currently trending sandalwood. In an unexpected move that surprised experts, Philyra autonomously decided to incorporate another popular wood noteโcedarwoodโinto the mix.
Itโs worth noting that professional perfumers typically have around 3,000 scent ingredients at their disposal. A single perfume composition might range from as few as seven to as many as hundreds of distinct aroma molecules. The scope for creativity here is virtually boundless, given the immense number of possible ingredient combinations.
Philyra goes beyond simply generating new combinations. It incorporates a nuanced understanding of how different components can either complement or clash with each other and which can be replaced with other alternative raw materials. The program aims not just for innovation but for crafting scents that will be genuinely appealing to people.
One of its key features is a โcreativity switch,โ which allows perfumers to select the level of inventiveness they want. A setting of 0 prompts Philyra to sift through historical data and consumer preferences to suggest formulas with the highest likelihood of mainstream acceptance. On the other end of the spectrum, a setting of 10 unleashes the softwareโs capability to explore uncharted territories, generating highly unconventional and rare combinations. This is automated creativity.
Philyra 2.0: better, faster, more sustainable
In Summer 2022, Symrise unveiled Philyra 2.0, an updated version of its AI system, to industry experts. Notably, this iteration of the computational model has been fine-tuned to prioritize sustainability when suggesting fragrance compositions. It is designed to favor ingredients that are both environmentally friendly and renewable from Symriseโs expansive ingredient catalog.
This shift resonates with the current zeitgeist where consumers are increasingly scrutinizing the environmental impact of the products they use, and where sustainability claims are becoming a standard selling point for many fragrance brands, whether in packaging or in the molecular make-up of the scent itself.
The revamped algorithm possesses the capability to not only create new sustainable fragrances but also to optimize existing formulas in terms of environmental impact. For example, it offers suggestions for replacing previously used chemical compounds with renewable or biodegradable alternatives, all without compromising the olfactory experience.
Human perfumers: do we still need them?
While Philyra 2.0 represents a technological advancement, it hasnโt completely eliminated the need for human expertise. Even with this update, a human perfumer still has the final say in refining the formula to achieve the perfect olfactory balance. The nuance and intuition of a seasoned perfumer are elements that AI hasnโt yet been able to replicate.
However, starting from 2022, all of Symriseโs fine fragrance creators have integrated Philyra 2.0 into their workflow. Although human experts still make the final decisions, itโs safe to say a splash of AI innovation is now embedded in each perfume we encounter on the shelves of beauty stores.
This unleashes a lot of creative potential, as human perfumers are now freed from many time-consuming technical and manual tasks, but also means the profession is going to change dramatically. Time and cost savings always mean fewer people are needed to do the job.
Exceptionally talented scent visionaries and established star noses donโt need to worry about employment. However, if I were finishing school and dreaming of entering the industry as a perfumer today, Iโd be careful to think twice. Big fragrance companies are quick to deny it, but a lot of technical jobs and also creative jobs are going away in the next decade.
Weโll take a closer look at the AI implications for the perfumer profession below in the outlook section.
First AI perfumes: where can I smell them?
Philyraโs first assignment came from O Boticario, a Brazilian cosmetics giant, which commissioned two millennial-focused fragrances: Egeo On You and Egeo On Me. The AI analyzed what had been resonating with this demographic in recent years and made a series of scent formula suggestions. Symriseโs experts the refined these options into final perfumes.
One these fragrances combines exotic spices with a milky-creamy undertone. This blend was so original that a Symrise staffer admitted it would have been unlikely for a human to come up with it. The elements it employed were not part of the usual range of raw materials perfumers work with. The second fragrance, aimed at young girls, is a mix of floral notes, toffee, lychee, and patchouli.
Both scents were tested with focus groups where they proved very successful. They even outperformed the competition when pitted against current Brazilian bestsellers. O Boticario launched both fragrances in 2019.
In another test case in 2019, Symrise deployed Philyra to co-create the experimental fragrance Berlin #3.0 with the help of 4,300 visitors of Bikini Berlin mall. AIโs first task was to formulate scents for various Berlin districts based on their descriptions. The algorithm was assisted by Senior Perfumer Marc vom Ende.
The action then moved to Bikini Berlin mall, where visitors were invited to test the fragrances. They were asked to match them to their impression of Berlin in order to decide, which one best represents the city. The AI adapted the formula dynamically using the resulting feedback until #Berlin 3.0 perfume was finalized.
Symrise heralded this experiment as a milestone in customer participation. And itโs true โ you canโt imagine a more involving way to create a perfume for general public. We have to keep in mind though, that it was still up to the human expert to fine-tune and balance the resulting composition.
More commercial projects followed. In 2021, the South American Belcorp corporation launched Esika Kalos Tech menโs fragrance that was entirely developed by Philyra. And in 2022, Thailand witnessed the launch of AI Awakening Cologne by Tros (belongs to Neo Corporation).
As you can see, itโs the budget-friendly mass market brands that are pioneering the AI integration. High-end niche houses that claim artistry and individuality as their raison dโetre, had always had the figure of the human perfumer as a visionary central to their identity and value proposition. Itโs understandable they are finding it challenging to justify AIโs computational efficiency without tarnishing their artistic ethos.
In 2022, Symrise has formally announced that ALL its fine fragrance perfumers are utilizing Philyra 2.0 in their work process. Itโs safe to assume other major fragrance companies (like IFF, Firmenich and Givaudan) to do the same. The advantages โ ranging from speed to the extensive array of raw materials that the neural network can incorporate into its work process โ are too big to pass by.
So we can be certain,ย that AI has also permeated the realm of high-end luxury and niche scents. The degree of AI involvement โ did the perfumers consult if for inspirational references before getting to work, or did they only touch-up the formula the algorithm created โ is rarely disclosed.
For example, IFFโs AI Codex was used to create two perfumes for the niche brand Vyrao (The Sixth and Sun Rae). Yet descriptions of these fragrances on brandโs website do not mention AI participation. Instead, they emphasize that every scent contains โa supercharged Herkimer diamond crystalโ energized by the brandโs healer, thus featuring the (belief-based) esoteric aspects and omitting the (science and fact-based) cutting edge technology.
Regardless of whether a perfume is 100% AI-generated or only contains a 5% AI contribution, algorithm involvement doesnโt inherently make the scent โlesserโ in quality or value compared to an entirely human-crafted one. Itโs more about the emotions and sentiment we associate with manual versus technological craftsmanship. With niche perfumery being all about emotions, however, this can be a crucial factor tipping the scales at the cash desk.
AI and perfume future: threats and opportunities
AI and Jobs: scaling down?
Symrise emphasizes that the use of their Philyra AI does not pose a threat to jobs within the fragrance industry. The algorithm serves as a tool to assist rather than replace human experts, primarily handling data analysis and mundane tasks to identify white spaces and forecast potentially successful fragrance options. Formulas with the highest likelihood of success still require human evaluation and fine-tuning.
Claire Viola, VP Digital Strategy Fragrance at Symrise, describes the technology as an apprentice, responsible for performing tedious routine tasks and enabling master perfumers to devote their attention to the production of innovative fragrances without the burden of repetitive testing or prolonged searches for new possibilities.
Despite the optimistic assurances, I think this story talks down the risks AI poses to the perfumery profession. Once AI begins to supplant apprentices and interns, the pathway for individuals to accumulate the vital experience needed to ascend to mastery in the field will be compromised, potentially leading to a reduction in training positions within major companies.
Seasoned perfumers agree with these concerns about the future of their industry. While talking to some of the top noses at the TFWA 2023 trade fair in Cannes, one message kept popping up: while established professionals have little to fear, the trajectory for nascent perfumersโparticularly those in the midst of academic trainingโstands to become markedly more challenging.
Big fragrance companies like Symrise and Givaudan offer apprenticeships. Each year, a few lucky newcomers get the chance to work side-by-side with a master perfumer. The idea is simple: these masters pass down their wealth of experience, which often includes trade secrets you canโt find in any textbook.
Historically, apprentices have been tasked with significant preparatory and post-production duties. For example, they would carry formula variations chosen for testing to the compounding department. There they would help the staff mix the ingredients drop-by-drop, following exact measurements to make test samples.
This wasnโt a quick job and involved a lot of waiting around. But that downtime was actually valuable. It allowed the apprentices to chat with material experts and soak up tons of useful information. Theyโd learn about ingredient details and innovations in a way that no formal lecture could ever provide.
This sort of hands-on, informal learning is priceless and only possible when people spend time together. Apprentices didnโt just learn while mixing formulas; they interacted with nearly all the companyโs departments, from marketing to production. That broad experience is under threat if artificial intelligence starts doing the jobs traditionally reserved for human trainees.
With the rise of digitalization and AI, many tasks have become simpler and faster. Gone are the days when trainees would walk through corridors with formulas in hand. Now, those formulas are sent digitally to compounding machines that prepare the required samples within minutes, all without human involvement.
Sure, itโs efficient and saves time and money. But for the apprentice, it means losing a valuable source of learning. They no longer need to leave their desk and spend time with colleagues from other departments.
In the past, training gave you a broad overview of all internal processes, almost by necessity. Todayโs apprentices are more like narrow specialists with fewer opportunities to look beyond their immediate tasks. This specialization has its benefits but can limit their ability to come up with creative solutions that require a broader perspective.
Moreover, apprentices used to be essential for aiding the work of experienced perfumers. They didnโt just prepare lab samples; they also made suggestions. Now, AI handles those tasks. A perfumer in a large company can manage without a host of assistants, so the number of apprenticeships will be cut down to the minimum needed to maintain staff. Well-paid training positions will be harder to find.
Breaking into the profession is getting noticeably harder. The barriers are already high and will become even higher for young people who donโt come from a line of perfumers where knowledge is passed down generations and canโt afford to spend years gaining experience without a steady income.
Itโs understandable that big international beauty companies are keen on using computer technologies for product development. The market is fast-paced, and staying ahead means constantly releasing new products. The risk of launching a sales flop is high. So, itโs tempting to remove the slower โhuman factor,โ prone to errors in judgment, and rely instead on lightning-fast software and massive databases. However, this might turn out to be a shot-sighted perspective if it means loss of human talent over time.
AI and sustainability: not the main problem
Introduction of sustainably thinking AI models like Philyra 2.0 is a welcome development. However, I believe their global impact will be negligible.
As far as raw materials are concerned, the perfume industry, particularly at the premium level, is already exemplary. Major fragrance corporations are investing heavily in sustainable cultivation of plant-based raw materials, driven largely by self-interest. After all, if the flowers and trees essential for perfume-making canโt grow due to climate change, the industry is doomed.
The major environmental impact of perfumes doesnโt come from the use of โenvironmentally unfriendlyโ scent molecules the AI is programmed to favor, but rather from packaging waste.
Although there have been strides made in packaging sustainability in recent years, the industry as a whole is still far from solving this issue. Unfortunately, an AI focused on creating the most eco-friendly fragrance formulas canโt contribute to resolving this particular challenge.
As the market is flooded with an exponentially increasing number of new releases, it doesnโt matter how sustainable the formulas inside the bottles are. The amount of packaging used to transport, store and display all this olfactory abundance will thwart any attempts at making perfumery as a whole truly sustainable.
AI and dupes: copying made easy
People like to be smart shoppers, and with prices for popular niche perfumes like Baccarat Rouge, Erba Pura, Layton or Delina exceeding several hundred dollars, no wonder fragrance dupes are all the rage online.
YouTube videos and TikTok clips about dupes for famous fragrances are often getting more views than the reviews of the originals. Young consumers want the same scents, but much cheaper.
One potential application for artificial intelligence could be in the development of such dupes, or strikingly similar copies of popular perfumes. The software could help avoid brand copyright infringement by analyzing and altering the formulas. This would ensure that, to human perception, the scent remains the same, yet technically, it wouldnโt be an exact copy.
Itโs worth questioning whether such a scenario should be welcomed.
Classes and masses: intensification of market segmentation through AI
Artificial intelligence programs will speed up and intensify market segmentation between budget-friendly mass market and high-end perfumery.
The use of AI software will save a lot of time, personnel, and costs, but perfume prices are unlikely to decrease for the consumers. Instead, the biggest financial beneficiaries of the new technology will be industry giants like LโOreal, Estee Lauder, Puig and Coty.
Smaller players and niche perfume houses are also dipping their toes into the AI waters. Some AI tools and platforms are already very affordable and target small-batch projects with low budgets (for example, Scentmate by dsm-firmenich). Some high-end niche brands wonโt use (or admit to using) AI for ideological reasons. But as soon as it becomes the most cost-efficient option the floodgates will be open.
As a result, the perfume world will likely become even more polarized, with a massive AI-driven mainstream market on one end and an even more exclusive and expensive artistic niche on the other.
The situation mirrors that of food consumptionโsome people find fast-food meals acceptable, valuing their affordability, speed, and convenience. Others would never drink a Starbucks coffee or eat a frozen pizza, opting instead to invest more money and time to ensure the quality and origin of their food and to enjoy the ambiance of fine dining establishments.
At the same time, currently there is a big trend towards conversion between niche and mass market fragrance profiles, driven both by market forces and AI innovations.
Itโs intriguing that IBM named the program after the ancient Greek mythological figure Philyra, who was supposedly the goddess of perfume. However, the Wikipedia entries focus more on the fact that Philyra gave birth to a centaur and was so horrified by its appearance that she preferred to be transformed into a linden tree rather than look at her child. It remains to be seen whether software-driven โPhilyrasโ will also produce terrifying fragrance mutations.
Perfume expert and founder of Niche Fragrance Discovery Agency Perfumestry, Yana, is an industry expert with years of experience in discovering and analyzing trends in the high-end fragrance market segment. She supports new brands in expanding their market presence and PR work. Yana shares her passion for olfactory culture at the weekly meetings of the Berlin Perfume Club, which she co-founded.
ALSO INTERESTING
I’m extremely inspired with your writing and the topics you cover. If I may ask, do you have some background in AI? Just asking, as this is so comprehensively covered here in regards to perfume. Anyway keep up the excellent quality writing, it is uncommon to see this quality of information in a free magazine today…
Thank you so much! My background with AI is not technical (not an engineer or coder), but the topic is of great interest to me because it is transforming the economy in general and perfume business in particular, and I’m always interested in how business models behind fragrances work (used to work as a business analyst before switching to the world of scent)
You’ve done a great job researching the topic! It’s the first time I found actual analysis and quality information about the role of AI in perfumery on the internet. May I quote your piece in my paper?
Dear Martin, thank you for your feedback! Very glad you found my piece interesting. Of course, do quote it in your paper! What is the name of it? Would love to read it!
Aw, this was an incredibly insightful post. Thanks for taking the time and making an actual effort to create a really good articleโฆ I’m just wondering… since chatgpt and midjourney became available I procrastinate even more than usual and don’t seem to get anything done. Do perfumers working with AI feel the same?
Haha… good point! I hope not! Otherwise we’ll never get anything new to smell!