Sephora AI Assistant Boosts Sales Conversion Over 5% in OpenAI Partnership

OpenAI

Sephora's deployment of a conversational AI shopping assistant powered by OpenAI generated an uplift in conversion rate of more than 5% by replicating in-store expert advice in digital channels.

AI-assisted summary of Réinventer l'expérience beauté grâce à l'IA — original source. Check the original for full context.

Source published: 2026-08-03

Detailed notes

## Background Cette vidéo a été publiée sur YouTube par la chaîne OpenAI le 7 juillet 2026. Selon la description, il s'agit d'une session enregistrée lors de l'événement OpenAI France en juin 2026, au cours de laquelle Géraldine Servouze, Global VP Data, AI & Insights chez Sephora, présente la stratégie IA de l'entreprise et son partenariat avec OpenAI.

## TL;DR Géraldine Servouze détaille l'intégration de l'intelligence artificielle chez Sephora à travers toute sa chaîne de valeur, de l'optimisation de la supply chain et du marketing jusqu'à la personnalisation de l'expérience client. Elle présente deux initiatives majeures conçues avec OpenAI, dont un assistant shopping IA intégrant l'application Sephora et disponible directement dans ChatGPT pour recréer l'expérience d'un conseiller en magasin. La démonstration sur application mobile illustre une recherche d'ingrédients et un achat fluide, ayant généré une hausse du taux de conversion de plus de 5 %. Enfin, la présentation met en avant trois enseignements clés : l'importance d'un parcours utilisateur guidé, la transparence explicative des recommandations et une démarche agile d'expérimentation continue.

## Atmosphere & style La présentation adopte un ton professionnel, structuré et dynamique, typique d'une intervention lors d'une conférence tech d'entreprise. L'ambiance est à la fois moderne et pragmatique, portée par une démonstration en direct qui rend l'innovation concrète tout en conservant une touche de convivialité et d'humour. L'ensemble dégage un sentiment d'agilité stratégique et d'enthousiasme face à l'impact mesurable de l'IA sur l'expérience client.

## Key takeaways - [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) L'intelligence artificielle chez Sephora est conçue comme un moteur alimentant une expérience omnicanale globale. - [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) Sephora s'appuie sur une empreinte mondiale massive de 80 millions de membres fidélisés, 2,5 milliards de visites annuelles et plus de 3 000 magasins. - [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) L'IA optimise l'ensemble de la chaîne de valeur : recommandations personnalisées, création de contenu marketing, prévisions de ventes et allocation des budgets média. - [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) L'un des objectifs principaux est de reproduire en ligne l'expérience de conseil personnalisé offerte par les conseillers beauté en magasin. - [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) Sephora a développé un Shopping Assistant IA natif et a intégré son application directement au sein de ChatGPT. - [04:24](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=264s) Une démonstration en direct montre comment l'assistant répond instantanément aux questions sur la composition ou les caractéristiques des produits avant l'ajout au panier. - [04:24](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=264s) L'assistant Shopping IA a généré un impact commercial fort, entraînant une hausse du taux de conversion de plus de 5 %. - [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) L'expérience utilisateur doit être guidée par des suggestions de requêtes proactives tout en conservant le ton de la marque Sephora. - [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) La transparence et l'explicabilité des algorithmes sont essentielles pour expliquer la raison des recommandations aux clients. - [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) L'agilité organisationnelle et une démarche continue de test-and-learn permettent d'optimiser rapidement les modèles et l'expérience client.

## Detailed walkthrough ### [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) The Global Scale of Sephora and AI Across the Value Chain

Géraldine Servouze opens her talk by greeting the audience and establishing the massive scale at which Sephora operates before diving into their technical roadmap. While Sephora is a household brand, its global footprint is truly immense, forming the bedrock for why data and technology are central to their business strategy. AI is positioned not as an isolated experiment, but as an engine powering an enhanced omnichannel experience. This entails delivering deeply personalized web touchpoints, equipping brand partners with granular business insights, and upskilling Sephora's workforce to leverage AI effectively in everyday operations.

Across the company's value chain, artificial intelligence optimizes digital, marketing, supply chain, and media operations. Digital teams rely on algorithms to tailor product recommendations, deliver next-best content, and dynamically determine optimal promotional mechanics—distinguishing between shoppers who convert via price discounts versus those motivated by complimentary gifts. Marketing units harness generative AI for scalable content creation across social channels, campaigns, and internal comms. Meanwhile, supply chain planners use AI-driven sales forecasting to streamline inventory orders, and media teams apply marketing mix modeling (MMM) and enhanced targeting to measure channel spend elasticity and allocate marketing budgets efficiently.

- **80 million** active loyalty members in the global customer database. - **2.5 billion** annual visits across web platforms and mobile apps. - Operations spanning **37 countries** with more than **3,000 physical stores**.

---

[02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) Translating the In-Store Beauty Advisor Experience to Digital

Turning specifically to customer-facing innovations, Sephora collaborated directly with OpenAI on two flagship initiatives. The core motivation stems from Sephora’s human competitive advantage: its in-store Beauty Advisors. Customers flock to brick-and-mortar stores specifically for tailored consultations on skin type, undertone matching, and hands-on application techniques. To replicate that intimate, conversational dynamic in a digital environment, Sephora created an AI-powered Shopping Assistant designed to engage in natural dialogue and provide hyper-personalized beauty advice.

In addition to developing this native assistant, Sephora expanded its reach by integrating the Sephora application directly into ChatGPT. This strategic integration meets customers natively inside an AI ecosystem they already use daily, bridging conversational AI discovery with commerce.

---

[04:24](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=264s) Live Demo: AI Beauty Assistant in Action and Business Impact

Walking the audience through a live demo on the mobile app interface, Géraldine illustrates how a typical customer journey unfolds. A user searches for "mascara" and lands on a specific product page, where they open the **AI Beauty** chat interface docked at the bottom of the screen. The shopper queries the ingredient profile—asking *"What ingredients are in this mascara?"*—and the assistant instantly parses and returns the full formulation breakdown. The customer then inquires if the formula is waterproof before seamlessly adding the item to their basket and checking out without breaking conversational context.

The commercial impact of this conversational assistant significantly exceeded internal baseline expectations. While standard retail personalization projects typically generate a modest conversion lift of around 1%, Sephora's AI Beauty Assistant drove a substantial surge in sales conversion, demonstrating the tangible economic value of conversational commerce at enterprise scale.

- The AI Shopping Assistant generated an uplift in the conversion rate of **more than 5%**.

---

[05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) Key Lessons Learned: Guidance, Transparency, and Continuous Testing

Reflecting on the development and rollout of the AI Beauty Assistant, Géraldine highlights three core operational lessons:

1. **Guided User Experiences Drive Higher Engagement:** Rather than leaving users with a blank text prompt, providing proactive suggestions—such as *"Gifts under €50"* or *"Summer beauty products"*—dramatically increases interaction rates. Acknowledging with a touch of humor that *"people are naturally a bit lazy"* and do not always want to type, structured guidance reduces friction. Crucially, the UX must preserve Sephora's signature brand voice so shoppers feel they are talking to a genuine Sephora Beauty Advisor rather than a generic bot. 2. **Explainability and Algorithmic Transparency:** Trust is paramount in responsible AI deployment. When the assistant suggests a product, it explicitly clarifies the rationale—connecting recommendations back to prior purchase history, known brand affinities, or specific user constraints. 3. **Rigorous "Test-and-Learn" Mindset:** Ongoing experimentation via extensive A/B testing allows engineering and product teams to analyze behavioral patterns, identify gaps, and deploy rapid iterative updates.

Géraldine concludes by emphasizing that while frontier models from OpenAI provide the foundational power, the ultimate differentiator is Sephora's organizational agility to test, adapt, and build durable, value-additive experiences alongside its customers.

## Notable quotes > [04:24](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=264s) What ingredients are in this mascara? — Customer

> [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) Gifts under €50 — Géraldine Servouze

> [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) Summer beauty products — Géraldine Servouze

> [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) people are naturally a bit lazy — Géraldine Servouze

## Mentioned Sephora, OpenAI, Géraldine Servouze, OpenAI France, ChatGPT, AI Beauty Assistant, Shopping Assistant

[Source](https://www.youtube.com/watch?v=sLYXRA5Ay9g)

中文简报

丝芙兰携手OpenAI推出AI助手，销售转化率提升超5% 丝芙兰通过部署由OpenAI支持的对话式AI购物助手，在数字渠道复刻门店内专业顾问的建议，使转化率提升了5%以上。 Géraldine Servouze 详细介绍了 Sephora 在其整个价值链中对人工智能的整合，涵盖从供应链与营销优化到客户体验个性化的各个环节。她展示了与 OpenAI 合作打造的两项重大举措，其中包括一款整合了 Sephora 应用程序且可直接在 ChatGPT 中使用的 AI 购物助手，旨在重现门店内顾问的服务体验。移动端应用的演示展现了成分查询与流畅的购买流程，并已推动转化率提升超 5%。最后，演讲强调了三项核心经验：引导式用户路径的重要性、推荐理由的解释透明度，以及持续实验的敏捷方法。 Sephora 依托全球 8000 万活跃会员及 3000 多家门店的庞大规模，将 AI 深入应用于全价值链，涵盖个性化推荐、营销内容生成、销量预测与媒体预算分配。 (0:05) 为了在数字化渠道中复现线下美妆顾问的专业咨询体验，Sephora 与 OpenAI 合作打造了原生 AI 导购助手，并将应用直接集成至 ChatGPT 中。 (2:38) 现场演示展示了 AI 助手即时解答产品成分等疑问并支持无缝加购的能力，该导购助手已推动销售转化率提升超过 5%。 (4:24) 项目核心实践经验表明：主动提供预设提示语能显著降低交互门槛，同时明确推荐原因并依托持续的 A/B 测试迭代是成功的关键。 (5:33)

Reading summary

Sephora deploys artificial intelligence across its global operations to translate its in-store beauty advisory model into digital commerce, according to Géraldine Servouze. Enterprise Scale and Cross-Functional AI Integration (0:05) Sephora's data and technology strategy operates against a global footprint comprising 80 million active loyalty members, 2.5 billion annual visits across web and mobile platforms, and more than 3,000 physical stores across 37 countries. Rather than treating artificial intelligence as an isolated experiment, the company embeds algorithms across its commercial value chain to optimize an omnichannel retail model. Operationally, digital teams employ algorithms to customize product recommendations and dynamically adjust promotional mechanics, distinguishing customers motivated by discounts from those incentivized by complimentary gifts. Marketing teams use generative AI for scalable content production across channels, supply chain planners utilize AI-driven sales forecasting for inventory management, and media teams apply marketing mix modeling to assess channel spend elasticity. Conversational Advisory and Commercial Impact (2:38) To replicate the high-touch consultations delivered by in-store Beauty Advisors regarding skin tones, undertones, and application techniques, Sephora collaborated with OpenAI on two primary initiatives: an embedded AI Shopping Assistant and a direct integration within ChatGPT. In a demonstration on Sephora's mobile app, the assistant parsed complex ingredient lists for a mascara and confirmed waterproof specifications within the chat interface, enabling the customer to add items to their cart and check out without exiting the conversation. According to Servouze, while baseline retail personalization initiatives typically produce a conversion lift of roughly 1%, Sephora's conversational assistant generated a conversion rate increase of more than 5%, evidencing direct commercial returns from natural-dialogue shopping interfaces. UX Guidance, Explainability, and Iterative Testing (5:33) Deployment of the conversational assistant surfaced three foundational operational practices. First, providing structured, proactive prompts—such as curated categories like gifts under 50 euros or seasonal selections—substantially improved interaction rates over an open, blank text box while preserving Sephora's brand voice. Second, recommendation transparency proved essential for user trust; the system explicitly explains why a product is suggested, tying outputs to previous purchase history, stated constraints, or brand affinities. Finally, Sephora adopted a test-and-learn operational framework. Engineering and product teams use extensive A/B testing to analyze user behaviors and gaps, rolling out iterative software updates rather than relying on static system deployments. Servouze maintains that while foundational models provide technical capabilities, an enterprise's ability to maintain organizational agility, conduct continuous testing, and design guided customer experiences remains the primary differentiator.

精炼阅读摘要

丝芙兰（Sephora）将人工智能确立为驱动全渠道零售体验的核心引擎，依托与 OpenAI 的深度合作，将其线下美妆顾问的专业咨询能力复刻至数字化界面与生成式生态中，成功推动销售转化率提升超过5%。全球规模底座与全价值链智能化协同 (0:05) 丝芙兰的技术路线图建立在其庞大的全球零售网络之上。根据演讲者 Géraldine Servouze 披露的数据，丝芙兰在全球拥有超过 8000 万活跃会员，网站与移动应用的年访问量达 25 亿次，业务覆盖 37 个国家和逾 3000 家实体门店。该庞大规模促使数据与技术成为公司整体商业战略的基石，AI 并非孤立的边缘实验，而是深度嵌入各业务环节的核心动力，旨在为消费者提供高度个性化的触点，为品牌合作伙伴提供颗粒度细致的商业洞察，并赋能内部员工的日常协作。在具体业务链条中，人工智能已被全面应用于数字化体验、市场营销、供应链和媒体投放。数字化团队利用算法提供个性化商品推荐与内容分发，并能动态识别促销机制偏好——例如精准区分更倾向于价格折扣的顾客与受赠品激励转化的顾客；市场部门采用生成式 AI 工具进行社交媒体、营销战役与内部沟通内容的规模化创作；供应链团队借助 AI 销售预测算法优化库存采购计划；媒体运营团队则运用营销组合模型（MMM）与精细化定向技术评估渠道支出弹性，从而更具成本效益地分配营销预算。数字化重塑美妆顾问体验与转化增长 (2:38) 丝芙兰直面消费者的技术创新核心，在于将实体门店中“美妆顾问”（Beauty Advisors）的人性化核心优势迁移至线上。针对消费者在肤质评估、底妆色号匹配以及专业上妆手法等维度的个性化咨询诉求，丝芙兰与 OpenAI 展开深度合作，一方面在原生移动应用中内置了具备自然对话能力的 AI 购物助手，另一方面直接将丝芙兰应用集成至 ChatGPT 生态中，在用户日常高频使用的外部 AI 环境中实现产品发现与商业闭环。在针对移动端界面的实际功能演示中，用户搜索睫毛膏并进入详情页后，可直接在屏幕底部的 AI 对话窗口发起询问；系统能迅速解析并呈现完整成分配方，并在解答该产品是否防水后，支持用户在不中断对话交互的前提下直接将商品加入购物车并完成结算。在商业表现方面，Servouze 指出行业内常规的零售个性化项目通常仅能带来约 1% 的转化率提升，而丝芙兰的 AI 购物助手带来了超过 5% 的转化率增长，验证了对话式商务在大规模零售场景下的实质经济回报。落地运营原则：引导交互、透明度与测试 (5:33) 在 AI 助手的开发与推广过程中，丝芙兰总结出三条关键落地准则。首先，结构化的引导式体验比空白输入框更能激发用户互动。由于部分消费者并不愿意主动输入长文本，界面通过提供预设提示选项（如“50欧元以下礼物清单”或“夏季美妆好物”）显著降低了交互摩擦，同时交互文案严格保持了丝芙兰特有的品牌语调，避免流于机械化应答。其次是算法的可解释性与透明度，这是建立负责任 AI 信任机制的关键；当系统推荐特定产品时，会向顾客主动说明推荐依据，将其与用户过往的购买记录、明确偏好的品牌或特定使用限制关联起来。最后，工程与产品团队贯彻严格的“测试与学习”（Test-and-Learn）机制，通过大规模 A/B 测试持续监测用户行为模式与潜在短板，实现功能与体验的快速迭代演进。 Géraldine Servouze 总结认为，尽管 OpenAI 研发的前沿大模型提供了底层算力与技术支柱，但零售企业真正的长期壁垒，在于组织自身能否保持高度敏捷，通过持续试验与快速适配，在真实客户场景中构建持久且具备增量价值的交互体验。

Detailed notes — zh

## Background 该视频由 OpenAI 于 2026 年 8 月 3 日发布在 YouTube 平台上。内容记录了 2026 年 6 月 OpenAI 法国活动中，丝芙兰（Sephora）全球数据、AI 与洞察副总裁 Géraldine Servouze 的主题演讲，主要探讨了丝芙兰如何依托 OpenAI 模型推进 AI 战略、打造智能导购助手并将其集成至 ChatGPT，以实现客户体验的规模化创新。 ## TL;DR 演讲首先展现了丝芙兰庞大的全球会员与零售体量，阐述了 AI 如何深度赋能个性化推荐、营销创意、供应链预测及媒介组合模型等全价值链环节。为在线上复刻实体店美妆顾问（Beauty Advisors）的专业服务，丝芙兰联手 OpenAI 推出了智能导购助手并深度接入 ChatGPT，实现了对话内一键加购与结算，推动线上转化率实现了超过 5% 的突破性增长。分享最后总结了落地生成式 AI 的三大法则：以引导式提示词降低用户输入门槛、提供透明的推荐理由以落实负责任 AI，以及依托密集 A/B 测试持续敏捷迭代。 ## Atmosphere & style 整场演讲充满前沿商业峰会的专业风范与活力，演讲者娓娓道来、幽默从容，开篇即以宏大的全球运营体量数据牢牢吸引全场目光。配合生动的手机投屏实机交互演示，复杂的底层算法被转化为极其直观丝滑的购物闭环，带来了很强的现场感染力。从展示超预期商业回报时的自豪与振奋，到复盘实战经验时的务实严谨，整体氛围既具前瞻创新的启发性，又兼备扎实落地的工业级说服力。 ## Bulletin - [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) 丝芙兰依托全球 8000 万活跃会员和年均 25 亿次数字端访问的庞大体量，将 AI 全面融入供应链预测、媒介投资模型（MMM）及差异化个性化营销的全链路中。 - [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) 通过与 OpenAI 合作打造智能导购助手并入驻 ChatGPT 原生生态，丝芙兰实现了从成分咨询到一键加购结算的闭环，推动线上转化率净增超 5%，远高于业内常规个性化约 1% 的提振幅度。 - [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 在生成式交互实战中，丝芙兰验证了预设提示词的“引导式体验”远优于空白输入框，且对话必须严格保持专业美妆顾问（Beauty Advisor）的温暖品牌语调。 - [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 丝芙兰强调透明可解释的推荐逻辑是负责任 AI 的底线，并指出组织团队与算法协同进化的敏捷学习机制才是零售企业在 AI 时代最核心的长期竞争壁垒。 ## Key takeaways - [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) 丝芙兰在全球拥有 8000 万活跃会员、37 个国家超过 3000 家实体门店，每年承载逾 25 亿次线上访问，庞大的全球体量推动 AI 成为全渠道业务的核心引擎。 - [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) AI 深度贯穿数字端个性化运营，不仅实现产品与内容千人千面，还能根据用户心理偏好差异化分发打折促销或附赠礼品策略。 - [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) 在运营支撑侧，生成式 AI 用于规模化创作营销素材，供应链借助预测模型优化采购决策，媒介团队则通过市场营销组合模型（MMM）最大化投放回报。 - [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) 线上 AI 导购的核心目标是将线下实体店专业美妆顾问（Beauty Advisors）因人而异、有温度的咨询体验精准复制至数字渠道。 - [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) 丝芙兰采取了 App 内置智能助手与直接接入 ChatGPT 原生生态的双重布局，直接触达高频使用大模型生态的终端用户。 - [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) 移动端 AI Beauty 助手实现了从解答产品成分、防水属性等细分诉求，到会话内直接一键加购并完成结账的端到端无缝转化链路。 - [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) AI 导购助手的上线使全站转化率净增超过 5%，大幅超越了传统零售个性化优化通常仅 1% 左右的收益水平。 - [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 经验法则一：引导式体验的转化效果显著优于空白对话框，主动提供预设提示词（如预算选礼）能有效契合用户天性并大幅降低交互门槛。 - [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) AI 助手的交互必须严格契合丝芙兰的品牌语调（Tone of Voice），确保用户感受到的是专业亲切的美妆顾问，而非生硬冰冷的机器人。 - [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 经验法则二：坚持推荐逻辑的透明度，主动向用户解释推荐理由（如历史购买偏好或特定肤质），以此践行负责任人工智能（Responsible AI）。 - [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 经验法则三：秉持敏捷的“测试与学习”（Test-and-Learn）心态，通过高密度的大规模 A/B 测试持续从用户真实交互数据中驱动算法迭代。 - [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 接入先进大模型只是起点，零售企业真正的护城河在于团队与算法协同进化、敏锐洞察并快速落地的组织学习能力。 ## Detailed walkthrough ### [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) 丝芙兰的全球体量与贯穿全价值链的 AI 战略布局在开场致辞中，演讲者首先与台下观众互动，抛出了一个引人深思的问题：大家都熟悉丝芙兰（Sephora），但大家是否真正了解丝芙兰在全球范围内的庞大体量？为了让大家对这家美妆巨头的覆盖规模有更具象的认知，她列举了一组震撼的数据：丝芙兰在全球客户数据库中拥有多达 8000 万名活跃忠诚度会员，其官方网站与移动端应用程序每年承载的访问量高达 25 亿次；业务版图横跨 37 个国家，线下运营超过 3000 家门店。在这样的全球体量下，人工智能与前沿技术并非边缘探索，而是早已被置于丝芙兰整体战略的核心引擎位置。公司致力于通过技术构建更卓越的全渠道体验（omnichannel experience），一方面在数字渠道打造极致个性化的交互，另一方面向合作品牌提供更深度的业务数据洞察，同时确保内部员工能够掌握前沿的 AI 工具与技能，将其充分融入日常工作。 AI 已经深度渗透进丝芙兰的整个业务价值链。对于数字化团队而言，AI 是推动网站精细化运营的关键工具，它不仅能提供精准的产品推荐与“千人千面”的个性化内容分发，还能因人而异地推送最契合的促销机制——因为数据表明，有些顾客对直接的折扣让利更敏感，而有些顾客则更偏好获得赠品或礼物，系统会据此量身定制触达策略。在市场营销层面，生成式 AI 大幅赋能创意生产，被广泛用于社交媒体营销、广告活动物料以及内部沟通内容的规模化生成；供应链团队则借助 AI 算法大幅提升销售预测的精准度，帮助企业科学计算向合作品牌订购库存的最优量级；而在媒介投放领域，媒体团队利用市场营销组合模型（Marketing Mix Models, MMM）精准评估各渠道预算增减的边际效应，从而实现营销投资回报的最大化与全媒体渠道的精准获客。 - 全球会员与规模：拥有 8000 万活跃会员，年均网站及 App 访问量达 25 亿次，遍布 37 个国家、3000 多家实体门店。 - 数字化与个性化：基于用户行为偏好自动匹配最优内容，差异化匹配打折优惠与赠品机制。 - 运营全链路赋能：涵盖生成式 AI 营销内容创作、供应链销售预测与采购决策、媒介混合模型（MMM）及精准广告投放。 ### [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) 联手 OpenAI：AI 导购助手与无缝购物体验演示谈及直面消费者的前端体验，丝芙兰与 OpenAI 展开了深度的战略合作，并重点推进了两项核心举措。第一项举措是将线下实体店的王牌体验复刻到线上。在线下门店中，丝芙兰最引以为傲的核心优势之一便是专业的美妆顾问（Beauty Advisors）。顾客之所以热衷于走进实体店，正是为了获取美妆顾问的面对面指导——他们会根据顾客的肤质、肤色推荐最适配的产品，并悉心传授上妆手法与使用技巧。为了让线上用户也能享受同等水准的专业关怀，丝芙兰构建了一款智能导购助手（Shopping Assistant），通过自然的拟人化对话，为顾客提供高度个性化的护肤美妆建议与精准选品。第二项战略举措则是将丝芙兰官方应用程序直接整合进 ChatGPT 平台，让品牌能够直接入驻消费者高频使用的原生 AI 生态，实现零距离触达。演讲者随后在现场通过手机投屏演示了移动端“AI Beauty”助手的交互过程。画面中，用户在搜索栏键入“睫毛膏”（mascara）进入单品落地页后，随即唤起了屏幕底部的 AI 对话浮层，直接提问：“这款睫毛膏包含哪些成分？”系统瞬间抓取数据库并完整列出了该产品的全成分表。用户紧接着追问该产品是否具备防水功能，AI 亦如实回复该款并非防水型。演讲者风趣地补充道，产品是不是防水并非重点，真正的亮点在于整个交互与交易链路的无缝衔接——用户无需跳出对话，即可在当前界面直接将商品一键加入购物车并顺畅结算。这一创新举措带来了极其亮眼的商业回报：上线后为网站转化率带来了超过 5% 的净增幅。演讲者特别强调，对于丝芙兰这样庞大基数的零售巨头而言，5% 的转化率提升是一个极其惊人的数字；业内绝大多数常规个性化优化带来的转化提升通常仅在 1% 左右，因此这代表了重大且实打实的商业价值突破。 ```mermaid flowchart LR accTitle: 丝芙兰线上 AI 导购转化链路 accDescr: 展示用户在丝芙兰 App 内从搜索商品到唤起 AI 问答、最终实现快速加购与结账的闭环流程。 subgraph S1["商品浏览"] A["搜索关键词<br/>(例如：睫毛膏)"] --> B["进入商品详情页"] end subgraph S2["AI 深度交互"] B --> C["唤起 AI Beauty 对话"] C --> D["成分分析与特性问答<br/>(如防水性能确认)"] end subgraph S3["转化变现"] D --> E["对话内直接一键加购"] E --> F["直接前往结账<br/>(转化率提升 >5%)"] end ``` - 核心诉求：线上还原实体店美妆顾问（Beauty Advisors）因人而异的专业肤质诊断与使用建议。 - 渠道延伸：推出官方 Shopping Assistant，并将 App 深度集成至 ChatGPT 原生环境。 - 业务转化成果：相较于常规个性化项目仅约 1% 的转化提振，AI 导购助手带来了超过 5% 的转化率爆发式增长。 ### [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 打造生成式导购的三大核心法则与未来展望在回顾研发 AI 美妆助手的实战历程时，演讲者总结了团队沉淀出的三条至关重要的实操经验。第一条法则是“引导式体验（guided experiences）的转化效果永远优于空白输入”。所谓引导，就是在交互界面主动向用户呈现推荐提示词（prompts），例如“50 欧元以下的礼物推荐”或“夏日美妆必备单品”，而不是把白纸一张的对话框丢给用户去绞尽脑汁输入。演讲者直言，人性本质上“天生带有一点懒惰”，普通用户并不总是愿意在手机上敲打冗长的文字，因此系统提供的引导越贴心，用户的互动频次就越高。与此同时，交互体验必须保持极致丝滑，更重要的是严格遵循丝芙兰标志性的品牌语调（tone of voice）——用户在交流时应该感觉对面坐着的是一位体贴温暖的丝芙兰专业美妆顾问，而非冷冰冰的通用 AI 机器人。第二条法则在于“透明度与推荐原因解释”。当 AI 助手向消费者推荐某款护肤品或彩妆时，系统有时会明确向其说明推荐理由，例如这是基于顾客过往的购买记录、偏好的特定品牌，或是系统此前记录的用户肤质特征。丝芙兰认为，坦诚地告知顾客“为什么推荐这款产品而非另一款”至关重要，这也是落实负责任人工智能（Responsible AI）的核心底线。第三条法则则是始终秉持敏捷的“测试与学习”（test-and-learn）心态。团队通过开展密集的大规模 A/B 测试，日以继夜地从海量用户与助手的真实交互数据中捕捉洞见，进而以极高的迭代节奏快速上线系统优化，让体验日臻完善。在分享的尾声，演讲者展望了 AI 技术带来的广阔空间。她表示，尽管丝芙兰在特定领域已经走在前列，但当下展示的场景依然处于飞速演进的初级阶段。团队依然保持着每天学习、持续精进的创业状态。她感触颇深地总结道，除了接入像 OpenAI 这样强大的技术基座和尖端算法外，零售企业真正拉开差距的核心资产，在于团队与算法共同学习、协同进化的敏捷能力；唯有将敏锐的业务学习能力与前沿技术融为一体，才能打造出经久不衰、真正为消费者沉淀核心价值的长青体验。 - 经验一：提供预设提示词（如“Gifts under €50”）进行启发式引导，降低输入门槛，严格维持美妆顾问的人性化品牌语调。 - 经验二：坚守推荐透明度，向用户解释推荐背后的历史偏好或数据依据，践行负责任 AI 理念。 - 经验三：保持高频 A/B 测试与数据驱动的快速迭代机制。 - 战略总结：强大的基础模型是催化剂，但团队敏捷进化与持续迭代的机制才是创造持久用户价值的核心壁垒。 ## Notable quotes > [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) 丝芙兰在全球客户数据库中拥有多达 8000 万名活跃忠诚度会员，其官方网站与移动端应用程序每年承载的访问量高达 25 亿次；业务版图横跨 37 个国家，线下运营超过 3000 家门店。 — Géraldine Servouze > [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) 公司致力于通过技术构建更卓越的全渠道体验（omnichannel experience），一方面在数字渠道打造极致个性化的交互，另一方面向合作品牌提供更深度的业务数据洞察，同时确保内部员工能够掌握前沿的 AI 工具与技能。 — Géraldine Servouze > [00:05](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=5s) 有些顾客对直接的折扣让利更敏感，而有些顾客则更偏好获得赠品或礼物，系统会据此量身定制触达策略。 — Géraldine Servouze > [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) 在线下门店中，丝芙兰最引以为傲的核心优势之一便是专业的美妆顾问（Beauty Advisors）。 — Géraldine Servouze > [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) 为了让线上用户也能享受同等水准的专业关怀，丝芙兰构建了一款智能导购助手（Shopping Assistant），通过自然的拟人化对话，为顾客提供高度个性化的护肤美妆建议与精准选品。 — Géraldine Servouze > [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) 将丝芙兰官方应用程序直接整合进 ChatGPT 平台，让品牌能够直接入驻消费者高频使用的原生 AI 生态，实现零距离触达。 — Géraldine Servouze > [02:38](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=158s) 上线后为网站转化率带来了超过 5% 的净增幅。演讲者特别强调，对于丝芙兰这样庞大基数的零售巨头而言，5% 的转化率提升是一个极其惊人的数字；业内绝大多数常规个性化优化带来的转化提升通常仅在 1% 左右。 — Géraldine Servouze > [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 引导式体验（guided experiences）的转化效果永远优于空白输入。 — Géraldine Servouze > [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 人性本质上‘天生带有一点懒惰’，普通用户并不总是愿意在手机上敲打冗长的文字，因此系统提供的引导越贴心，用户的互动频次就越高。 — Géraldine Servouze > [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 用户在交流时应该感觉对面坐着的是一位体贴温暖的丝芙兰专业美妆顾问，而非冷冰冰的通用 AI 机器人。 — Géraldine Servouze > [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 坦诚地告知顾客‘为什么推荐这款产品而非另一款’至关重要，这也是落实负责任人工智能（Responsible AI）的核心底线。 — Géraldine Servouze > [05:33](https://www.youtube.com/watch?v=sLYXRA5Ay9g&t=333s) 零售企业真正拉开差距的核心资产，在于团队与算法共同学习、协同进化的敏捷能力；唯有将敏锐的业务学习能力与前沿技术融为一体，才能打造出经久不衰、真正为消费者沉淀核心价值的长青体验。 — Géraldine Servouze ## Mentioned Géraldine Servouze, 丝芙兰 (Sephora), OpenAI, ChatGPT, Shopping Assistant, AI Beauty, 市场营销组合模型 (Marketing Mix Models, MMM), 负责任人工智能 (Responsible AI) ## Video description Découvrez comment Sephora réinvente l'expérience beauté grâce à l'intelligence artificielle. Lors de cette session enregistrée à l'occasion de l'événement OpenAI France en juin 2026, Géraldine Servouze, Global VP Data, AI & Insights chez Sephora, présente la stratégie IA de l'entreprise et la manière dont elle s'appuie sur les modèles d'OpenAI pour proposer des expériences toujours plus personnalisées à ses clients, tout en accompagnant ses équipes et ses partenaires. Géraldine revient notamment sur le développement du Shopping Assistant de Sephora et sur l'intégration de l'application dans ChatGPT. Elle partage les enseignements tirés de ces projets, de l'importance d'une expérience utilisateur guidée, à la transparence des recommandations et à une approche continue de test and learn pour améliorer l'expérience client à grande échelle. [Source](https://www.youtube.com/watch?v=sLYXRA5Ay9g)

Source quotes

[0:05] Géraldine Servouze: 丝芙兰在全球客户数据库中拥有多达 8000 万名活跃忠诚度会员，其官方网站与移动端应用程序每年承载的访问量高达 25 亿次；业务版图横跨 37 个国家，线下运营超过 3000 家门店。 [0:05] Géraldine Servouze: 公司致力于通过技术构建更卓越的全渠道体验（omnichannel experience），一方面在数字渠道打造极致个性化的交互，另一方面向合作品牌提供更深度的业务数据洞察，同时确保内部员工能够掌握前沿的 AI 工具与技能。 [0:05] Géraldine Servouze: 有些顾客对直接的折扣让利更敏感，而有些顾客则更偏好获得赠品或礼物，系统会据此量身定制触达策略。 [2:38] Géraldine Servouze: 在线下门店中，丝芙兰最引以为傲的核心优势之一便是专业的美妆顾问（Beauty Advisors）。 [2:38] Géraldine Servouze: 为了让线上用户也能享受同等水准的专业关怀，丝芙兰构建了一款智能导购助手（Shopping Assistant），通过自然的拟人化对话，为顾客提供高度个性化的护肤美妆建议与精准选品。 [2:38] Géraldine Servouze: 将丝芙兰官方应用程序直接整合进 ChatGPT 平台，让品牌能够直接入驻消费者高频使用的原生 AI 生态，实现零距离触达。 [2:38] Géraldine Servouze: 上线后为网站转化率带来了超过 5% 的净增幅。演讲者特别强调，对于丝芙兰这样庞大基数的零售巨头而言，5% 的转化率提升是一个极其惊人的数字；业内绝大多数常规个性化优化带来的转化提升通常仅在 1% 左右。 [5:33] Géraldine Servouze: 引导式体验（guided experiences）的转化效果永远优于空白输入。 [5:33] Géraldine Servouze: 人性本质上‘天生带有一点懒惰’，普通用户并不总是愿意在手机上敲打冗长的文字，因此系统提供的引导越贴心，用户的互动频次就越高。 [5:33] Géraldine Servouze: 用户在交流时应该感觉对面坐着的是一位体贴温暖的丝芙兰专业美妆顾问，而非冷冰冰的通用 AI 机器人。 [5:33] Géraldine Servouze: 坦诚地告知顾客‘为什么推荐这款产品而非另一款’至关重要，这也是落实负责任人工智能（Responsible AI）的核心底线。 [5:33] Géraldine Servouze: 零售企业真正拉开差距的核心资产，在于团队与算法共同学习、协同进化的敏捷能力；唯有将敏锐的业务学习能力与前沿技术融为一体，才能打造出经久不衰、真正为消费者沉淀核心价值的长青体验。

Source quotes

[4:24] Customer: What ingredients are in this mascara? [5:33] Géraldine Servouze: Gifts under €50 [5:33] Géraldine Servouze: Summer beauty products [5:33] Géraldine Servouze: people are naturally a bit lazy

Original transcript

[0:05] Hello everyone. I'm delighted to be here with you this evening. Does everyone know Sephora? Great. But does everyone know just how large Sephora is globally? Perhaps not. [0:19] Before I introduce some of our AI initiatives, let me share a few figures. We have 80 million active loyalty members in our customer database and 2.5 billion visits each year across our websites and mobile apps. [0:44] We're present in 37 countries and operate more than 3,000 stores. For us, AI and technology are at the heart of our strategy. They enable us to deliver a better omnichannel experience by creating more personalized experiences on our websites, providing our brand partners with richer data and insights to better understand their business at Sephora, and ensuring our employees have the right tools and AI skills to make the most of AI in their day-to-day work. [1:26] At Sephora, AI supports every part of our value chain. It helps our digital teams optimize our websites through greater personalization, offering product recommendations, the next best content for each customer, and the most relevant promotions. [1:48] Some customers respond better to discounts, while others prefer receiving gifts. We tailor the experience accordingly. AI also supports our marketing teams. Generative AI now enables us to create more content for social media, advertising campaigns, and internal communications. [2:12] Our supply chain teams use AI to improve sales forecasting, helping us determine how much inventory to order from our brand partners. Our media teams also use AI to optimize marketing investments through marketing mix models that help us understand where to invest and what impact increasing or decreasing spend on a channel will have. [2:38] AI also helps us target customers more effectively across media channels. Focusing specifically on the customer experience, we've worked with OpenAI on two key initiatives. The first was recreating the in-store Sephora experience on our website. [3:01] One of Sephora's greatest strengths is our Beauty Advisors. They work in our stores, offering personalized advice, and that's one of the main reasons customers come to Sephora. They recommend the best products based on skin type, skin tone, and explain how to apply different products. [3:25] We wanted to recreate that experience online through a Shopping Assistant that enables natural conversations with customers, offering personalized recommendations, tailored product suggestions, and beauty advice specific to each individual. I'll dive a little deeper into that in a moment. Our second initiative is that the Sephora app is now integrated directly into ChatGPT. [4:02] This allows us to engage with our customers through a platform they already use every day—ChatGPT. Now I'd like to focus on the first use case, our Shopping Assistant. I'll show you a short demo and talk you through it. [4:24] Here we're looking at the Sephora app. It's slightly cropped—sorry, you can't quite see the top of the screen. [4:30] A customer types "mascara" into the search bar and lands on the mascara product page. [4:37] They then open AI Beauty, which you can see at the bottom of the screen, and ask, "What ingredients are in this mascara?" Our Beauty Chat immediately provides the full ingredient list. They then continue the conversation by asking whether the mascara is waterproof. This one isn't, but that's not really the point. [5:00] The customer can then add the product directly to their basket and proceed straight to checkout, continuing their shopping journey. We've found that this has become a real driver of conversion on our website. [5:17] We've increased our conversion rate by more than 5%. At Sephora's scale, a 5% uplift is huge. Most personalization initiatives typically deliver around a 1% increase in conversion, so this has had a significant business impact. [5:33] I'd like to share a few lessons we've learned from building the AI Beauty Assistant. The first is that guided experiences always perform better. [5:53] By "guided," I mean giving customers suggested prompts instead of expecting them to type everything themselves—for example, "Gifts under €50" or "Summer beauty products." [6:11] We've found that the more we guide the experience, the more people interact with the chatbot. I think people are naturally a bit lazy—they don't always want to type. So it's important to guide the experience wherever possible. [6:26] The user experience is equally important to ensure adoption. Everything has to feel seamless, and the assistant has to reflect Sephora's tone of voice. Customers should feel like they're receiving advice from a Sephora Beauty Advisor, not simply interacting with AI. Maintaining that tone of voice is incredibly important. [6:53] The second lesson is to explain our recommendations and be transparent. When the chatbot recommends a product, we sometimes explain why—for example, based on the customer's purchase history, brand preferences, or other information we know about them. [7:15] Being transparent about why we're recommending one product over another is important to us. We also want to ensure we do this responsibly from an AI perspective. Finally, our third lesson is to always adopt a test-and-learn mindset. [7:32] We run a large number of A/B tests and constantly learn from how customers use the Beauty Assistant. That enables us to release improvements quickly and continuously enhance the experience. [7:52] To conclude, I'd say that AI represents an extraordinary opportunity for us. Even though we're already quite advanced in certain areas, the two use cases I've shared today are still evolving rapidly. We're learning every day. [8:12] That's why this test-and-learn approach is so important—we're continuously improving. Beyond having access to powerful technologies like OpenAI and advanced algorithms, it's really our ability to learn and evolve together that allows us to create experiences that are both sustainable and genuinely valuable for our customers. [8:41] Thank you.