Amorepacific

Objective

Reproduce online the advice that sells in stores. Amorepacific finds that the skin diagnostic converts less well on the web than in stores, where a counselor talks with the customer. The genAI app aims to close that gap by delivering personalized advice from purchase history and the diagnostic.

The deployment

Amorepacific, South Korea's leading cosmetics group (30+ brands including Sulwhasoo, Laneige, Innisfree), built a generative beauty advice app on its Azure foundation. The AI Beauty Counselor, put into service under the name AMORE CHAT on the Amore Mall online store, answers customer questions and recommends products based on their purchase history, their reviews, and the company's own expertise. It uses the GPT-4o and GPT-4o-mini models through Azure OpenAI Service, with Data Factory on Microsoft Fabric for the data and the AI Search features of Azure AI Foundry to retrieve the relevant product information. Upstream, a skin diagnostic tool questions the customer (for example rating whether the skin is oily from 1 to 5) and analyzes a photo of the face. This diagnostic has been used 2.5 million times online and in stores over four years, but it converted less well online than in stores. The genAI advice connects this diagnostic to a conversation that steers the customer toward suitable products. The group then extended the diagnostic component (Dr.AMORE, accuracy above 90% against professional devices) to several brands and touchpoints, then launched in 2026 an Amore Mall app on ChatGPT, a first in Korean beauty.

Results Proof B

The key figure (2.5 million diagnostic uses) and the stack details come from a Microsoft customer story, an official but interested source (vendor bias), which caps the case at level B. The move into production and the ongoing activity are corroborated by two official Amorepacific sources (stories.amorepacific.com releases from 2026). No quantified conversion impact is published in financial results, so no level A.

1. 1
   Collect the skin diagnostic customer, diagnostic tool

   The customer answers a questionnaire (for example rating oily skin from 1 to 5) and takes a photo of her face, online or in store.

2. 2
   Assemble the customer context data / AI team

   Consolidate purchase history, reviews, and diagnostic results via Data Factory on Microsoft Fabric.

3. 3
   Retrieve the relevant product information AI (search / RAG)

   Azure AI Foundry's AI Search features surface the in-house expertise and catalog matched to the profile.

4. 4
   Generate the advice and recommendation AI (LLM)

   GPT-4o and GPT-4o-mini via Azure OpenAI produce the conversational response and the suggested products.

5. 5
   Steer toward purchase on Amore Mall AMORE CHAT app / e-commerce

   The conversation guides the customer toward suitable products to lift the transition to online purchase.