Revolutionizing Chemical Formulation with AI Integration at L'Oréal.
L’Oréal was developing a next-gen internal platform to support its scientists in formulating new products faster and more efficiently.
The project combined a full UX redesign of the existing tool with the integration of an AI assistant capable of turning complex data into actionable insights.
My role was to ensure the entire experience was intuitive, cohesive, and built for trust.
The core problem
As Senior Product Designer, I was brought in to help shape both the AI assistant and the overall platform experience.
Early on, I uncovered several deep-rooted issues through a research phase led by an external UX researcher:
🚫 Strong resistance to change, many users feared the AI might eventually replace their role, creating hesitation and mistrust.
🚫 Fragmented experience, the platform was cluttered, inconsistent, and lacked clear design ownership, leading to confusion and inefficiency.
🚫 Overwhelming complexity, users were flooded with data and features, with no clear hierarchy or guidance.
🚫 High cognitive load, navigating the tool required too much mental effort, slowing down workflows and impacting productivity.
To address these challenges, I redesigned the platform with a clear, consistent interface that felt familiar and seamlessly fit into scientists’ daily workflows. The new UI was both intuitive and ergonomic, reducing friction and supporting efficient decision-making.
I also introduced transparent AI features with clear, explainable predictions to reduce fear and foster trust.
Outcome
4,500+ scientists onboarded globally
+28% productivity
(measured by task completion)
High adoption and increased confidence
in AI-powered tools
Step 1: frame the problem
I Started with a series of workshops to frame the needs and understand the context.
I crossed data to have a better mapping of all existing workflows using:- Qualitative data
● The field analysis of a UX researcher 🕵🏻♂️
● A questionnaire I previously sent to a panel of chemists in the USA, France, China, Japan, India and Brazil.
- Quantitative data
● Analytics from existing process.
Key deliverable:
Because we needed to be user centric, I compiled business knowledge with 3 personas as we gained a deeper understanding of our users.
Additionally, I have some Miro boards, raw and unformatted, but crucial, such as the Vision Board and the existing user journey for creating and modifying formulas.
Given some users' resistance to change and digital transformation, as well as their fear of being replaced by artificial intelligence algorithms, these multiple sessions aimed not only to dissect and understand the user experience beyond simple business needs but also to study the impact of changes on the business.
It was necessary to involve cross-functional chapters on supporting the company's employees through change. Thus, it was gradually that this systemic journey was put in place, after several months of extremely collaborative and complex work.
Embracing AI Dynamics
The AI feature was developed by a small team over several months and had already reached an advanced stage, with predefined logic and functionality that were difficult to change.
How does it work?
The algorithm used machine learning to analyze existing formulas and automatically predict a shortlist of safer ingredient replacements when hazardous components were detected. Scientists could then validate these suggestions through lab testing.
At first, the AI made some incorrect predictions, but through continuous iteration and real-world validation, its accuracy improved rapidly, becoming a trusted assistant in the formulation process.
My challenge was to turn this rigid, technically complex system into a seamless, intuitive experience, one that balanced transparency and control, making the power of prediction and automation feel natural and easy to use for scientists in the lab.
Step 2: Design the solution
I facilitated a series of creative mapping workshops of the new user workflow including both data and engineering to the core product team.
This creative step was relatively quick since the technical solution was already well advanced. It provided an opportunity to update the technical team on the existing technological infrastructure and most important to collaborate with data analysts on the quality and format of the displayed data.
I iterated on data readyness.
Cleaning, consistency and validation to ensure useful data for users.
I supported user stories to prepare the breakdowns and deliveries for the various upcoming sprints.
After constructing the various use cases, it was necessary to go deeper into the user workflow and navigate through the constraints.
These constraints raised many questions and expectations among the teams and managers.
where to insert the algorithm data? Which data to display? How to display it? Where is the starting point of the journey into the existing navigation? What is the impact on other journey?
Key deliverable:
A Figma file containing:
Numerous iterations and variations of all screens were designed to account for all possible scenarios, informed by data insights.
Additional screens, sometimes from other tools, outside of the main workflow but impacted by it.
In the same time, I established consistent design standards by building a component library, ensuring the product’s scalability.
Final UI - Anatomy of the pages
Anatomy of the AI tool's core page: Suggestions and predictions of raw material effects for substituting ingredients in a formula
♦︎ Predictive data
Relying solely on the color proximity criterion is not enough.
It is necessary to have predictions on other key criteria that can help users make their choice, such as:
Fading: the gradual loss of intensity or vibrancy of hair color over time
Selectivity: the ability of a product to specifically target certain parts of the hair while preserving others.
♦︎ The proximity to the source
The user has an overview of the projection of a new blend.
They can even compare the impact on the product's color that they 'clean' at the time of application (T0) and an estimate of the normal color degradation 15 days later (T15).
♦︎ The other important attributes.
For any chemical formula, there is a list of attributes that are also important for making a choice, notably including a prediction of the cost of the new formula with the new ingredient and its eco SPOT score.
