The essentials
Heading into 2027, AI in perfumery has crossed from research promise to commercial fact, on a narrow but real front. The reference case is Osmo, a company spun off from Google Brain in 2022 and led by CEO Alex Wiltschko, whose machine-learning model predicts how a molecule smells from its chemical structure. Osmo has brought three novel scent molecules to market, Glossine, Fractaline, and Quasarine, and in March 2025 launched Generation, a fragrance house built around AI-assisted creation. Just as important is what the principals claim: both Wiltschko and master perfumer Christophe Laudamiel, the field's most prominent practitioner-commentator, hold that AI accelerates the perfumer's work and does not replace the nose.
So the honest status report for 2027 reads: molecule discovery, demonstrably working; AI-assisted composition, in commercial trial through Generation; fully autonomous machine perfumery, not on any credible roadmap. The sections below unpack each layer and what to watch for during the year.
Osmo: teaching machines the map of smell
Osmo's founding problem is old and hard: unlike color, smell has never had a reliable map from physical structure to perception. The company, spun out of Google Brain in 2022 under Alex Wiltschko, trained models on large datasets pairing molecular structures with human odor descriptions, producing a system that predicts a molecule's smell from its structure alone. The research lineage was validated in peer-reviewed work on the so-called principal odor map before the company's commercial turn.
The practical consequence is speed. Screening candidate molecules computationally, before synthesis, compresses a discovery process that traditionally took years of trial chemistry. It also reorients discovery toward properties the industry increasingly needs: novelty that is patentable, biodegradability, safety profiles compatible with a tightening regulatory environment. That efficiency argument, rather than any claim about creativity, is the load-bearing case for AI in perfumery as 2027 begins.
Three molecules with names: Glossine, Fractaline, Quasarine
Claims about AI perfumery became checkable when Osmo shipped materials. Three molecules have been commercialized: Glossine, Fractaline, and Quasarine. Their significance is less any single odor profile than the proof of pipeline: a computationally discovered material can now travel the whole road from model prediction through synthesis, safety evaluation, and market release. That road, historically, was the industry's slowest and most expensive.
Context matters for 2027 expectations. The conventional discovery of a major new aroma chemical has been a once-in-several-years event for the composition giants, and captive molecules are strategic assets that differentiate palettes. If AI-assisted discovery reliably shortens that cycle, the medium-term effect is a broader palette reaching perfumers faster, an infrastructural change that buyers will experience only indirectly, through new effects in finished perfumes, likely from 2027 onward.
Generation: the AI-native house experiment
In March 2025, Osmo moved downstream and launched Generation, a fragrance house whose creative process is built around its AI tooling. As a commercial experiment it tests the question the industry actually cares about: not whether a model can propose molecules, but whether AI-assisted workflows can produce finished perfumes that people choose to wear, at niche-credible quality.
For 2027, Generation functions as the field's most visible benchmark. Its releases, reception, and follow-through will shape how the rest of the industry budgets for AI tooling far more than any laboratory paper. A fair reading in mid-2026: the experiment is young, the verdict is open, and both enthusiasm and dismissal are premature. What can be said is that the test is now running in public, which was not true two years ago.
The division of labor: acceleration, not replacement
The most consistent position across the field's credible voices is a division of labor. Alex Wiltschko frames Osmo's tools as instruments that widen the perfumer's palette and shorten iteration; Christophe Laudamiel, a master perfumer who has engaged with scent technology for decades, argues the same from the composer's side: models handle search and screening, while brief interpretation, cultural judgment, and the final aesthetic decision remain human. No major house or composition firm currently claims otherwise in public.
The claim deserves scrutiny precisely because it is convenient, and 2027 is when it gets tested at scale. If AI-assisted juniors can iterate accords dramatically faster, the economics of composition teams shift even while the top of the craft stays human. The realistic near-term change is not the perfumer's disappearance but a redistribution of time: less mechanical formula iteration, more evaluation and direction. That is also, historically, what every prior tool shift in perfumery has done, from gas chromatography to headspace capture.
What to watch in 2027, and the hype to discount
Four checkpoints will mark real progress during 2027. First, adoption of Osmo-discovered materials by perfumers outside Osmo's own house. Second, Generation's release cadence and critical reception. Third, whether the composition giants, Givaudan, dsm-firmenich, IFF, Symrise, surface their own AI-discovered captives, converting internal programs into shipped materials. Fourth, regulatory fit: computationally screened molecules arriving with strong safety and biodegradability profiles would land well in a market already digesting the EU's 82-allergen labeling regime.
Discount, meanwhile, two genres of noise: "AI created this perfume" marketing where a model merely tagged along, and neuroscience-adjacent claims that borrow AI's credibility to sell mood effects that remain commercially ahead of their evidence. The sober 2027 statement stands: AI has entered perfumery as an instrument, its first commercial products exist, and the nose still decides.
Sources
- Perfumer & Flavorist, multiple articles on Osmo, its odor-prediction model, its commercialized molecules, and the launch of Generation (2024-2026). Accessed 2026-07-06.
- Contrary Research, company profile of Osmo (founding, Google Brain lineage, leadership). Accessed 2026-07-06.
- Osmo and Generation, official communications on Glossine, Fractaline, Quasarine, and the March 2025 house launch. Accessed 2026-07-06.
- Public statements by Alex Wiltschko and Christophe Laudamiel on AI as acceleration rather than replacement, as reported in the trade press. Accessed 2026-07-06.