In the last year, consumer-facing AI has stolen all the headlines, from AI-generated product recommendations and marketing assets to enhanced skin diagnostic tools. Yet the technology is also set to upend the beauty product innovation behind the scenes.
How AI is Driving Better Beauty Innovation
Sequential learning can train AI systems on relatively tiny beauty formula data sets.
For instance, in 2023, Cambridge, Massachusetts-based Osmo made a splash with its mission of giving computers "a sense of smell" by pairing odor maps with AI to predict what a molecule smells like based solely on its structure. The goal: to create safer, more environmentally friendly aroma molecules for the flavor and fragrance market.
Elsewhere in the fragrance world, Cosmo International Fragrances created the ScentElligence AI platform to analyze data and deliver fragrance options by combining sensoriality with performance.
AI is also turning up in the wider beauty space. For example, Cosmex AI’s tool is designed to validate cosmetic product formulas against multiple regulatory guidelines, including national/regional requirements and retailer standards.
Cosmex AI reportedly screens data sets from more than 30,000 cosmetic ingredients to determine if a formula complies with a specific target. If not, the system flags the non-compliant ingredients and percentages, informs the user about the compliance issues, and recommends corrective actions to bring the formula into compliance.
No doubt, artificial intelligence has the power to help brands to create better formulations faster.
Small Data Sets? No Problem
One such AI software from Citrine Informatics reportedly allows brands—from indies to multinationals—to create better, faster.
AI platforms can shave down development timelines from 12-18 months to as little as 2.5 months, Citrine Informatics argues.Mirma at Adobe Stock
The Citrine Informatics platform was designed in response to rapid trend cycles and escalating demands for clean and sustainable beauty.
The goal, according to Citrine, is to empower brands to develop consumer-preferred products with precision.
Notably, getting started with the platform requires relatively small data sets.
"It sometimes surprises people how little data you need to start an AI project on the Citrine Platform," says Greg Mulholland, CEO and co-founder of Citrine Informatics. "A beauty product customer came to us with just 30 data points, and we were able to hit their property targets in a fifth of the standard development time with more carefully chosen experiments."
Mulholland adds, "Many of our CPG customers use our platform to predict stability of products over the long-term using short term data, speeding up their R&D process."
How AI Shortens Beauty Development Timelines
In some cases, Mulholland says, “Our customers have reformulated lines of products using naturally derived ingredients that were completely new to them in just a few months."
The average timeline for a similar project is estimated to be 12-18 months.
To cut down on a year-plus of development time, Citrine Informatics’ customers have used just a few dozen data points combined with their collective experience, classifying ingredients and their typical usage levels to help guide the AI.
They provide information on the relevant properties of the ingredients used and how they would impact the desired results.
In an industry where data can be scarce, but knowledge in the heads of researchers is rich, it is important to make it easy to train AI with a combination of test data and user knowledge.
Customers then use sequential learning via small rounds of experiments, the results of which are fed back into the Citrine Platform to train the AI and recommend the next experiments.
Each round of tests generates a small number of candidate formulations to be tested in subsequent rounds.
By testing more unusual formulations early, the AI better understands the breadth of the formulating landscape, Citrine Informatics argues, with later rounds more tightly focused on hitting targets quickly.
For brands seeking speed to market and nimble response times to trends or shifting market conditions, AI may just offer a solution.
