L'Oreal Paris

Objective

Offer personalized, continuously available beauty advice to remove purchase doubt and guide toward the relevant products in the L'Oreal Paris catalog (750+ references).

The deployment

Beauty Genius is a genAI beauty assistant accessible from the smartphone (brand web, app, WhatsApp). The user asks a skin, makeup, or coloring question, gets a diagnosis, a virtual try-on, and recommendations drawn from the 750+ L'Oreal Paris products. The engine combines an LLM (GPT-4o) and a RAG system bounded to L'Oreal content to ground the answers, with visual try-on and diagnosis provided by ModiFace. Launched in the United States in 2024, with a rollout announced for other key markets in 2025.

Results Proof B

Usage figure published in L'Oreal's official annual report (primary source) and technical architecture confirmed by the Microsoft Azure OpenAI customer story. Two concordant sources, one of them primary.

1. 1
   User question client

   The user describes a skin, makeup, or coloring need, or sends a photo for diagnosis/try-on.

2. 2
   Context retrieval AI (Ada embeddings)

   The RAG system retrieves, from the L'Oreal knowledge base, the relevant products and content to ground the answer.

3. 3
   Answer generation AI (GPT-4o)

   GPT-4o formulates personalized advice and a routine recommendation in under 5 seconds.

4. 4
   Diagnosis and visual try-on AI (ModiFace)

   ModiFace performs the skin diagnosis or the virtual makeup try-on from the camera.

5. 5
   Supervision and updating brand / data team

   The brand team updates the catalog, checks the content guardrails, and tracks aggregate conversations.