BEAUTY BAE - GEN AI CHATBOT
INTRO
Every day, thousands of customers visit our brick-and-mortar stores, seeking personalised beauty services from our expert Beauty Advisors. With Beauty Bae, we're bringing that same level of tailored, expert beauty guidance right to our users' fingertips through our app.
​
Customers can now ask beauty questions naturally and conversationally, receiving personalised product recommendations in an engaging and secure environment.
PHASE 1: PROOF OF CONCEPT
We developed a proof of concept to evaluate the viability of Beauty Bae. The chatbot was introduced during the product discovery stage of the purchase funnel, with its entry point placed as an icon on the bottom-right corner of the listing page. To help draw users’ attention, we incorporated a subtle animation into the button.
INITIAL RESULTS AND FINDINGS
The initial pilot delivered promising results, achieving a conversion rate of 12.59%, significantly higher than the app’s typical 2–3%. However, overall traffic and usage of Beauty Bae remained low, with only a 2.4% click-through rate, indicating that users were not actively engaging with the feature.
OPTIMISATION: PRODUCT LISTING PAGE
To improve Beauty Bae’s usage rate, we increased the number of entry points across the user journey.
​
Beauty Bae now also appears after users exceed a certain scroll depth on the product listing page, allowing it to assist those who are actively browsing and exploring options.
​
Additionally, the suggested prompts within the chat are dynamically tailored to the specific listing category the user is viewing, making the experience more relevant and useful.
OPTIMISATION: PRODUCT DETAIL PAGE
We also integrated Beauty Bae into the product description and reviews section of the Product Detail Page.
With a single tap, users can send a pre-generated prompt to ask about product suitability or receive a summary of reviews. This enables them to make more informed and confident purchase decisions.
PHASE 2: SOLVING FOR AI ADOPTION
Following the initial launch of Beauty Bae, we gathered usage data and conducted additional user testing to further refine and enhance our AI implementation within the app.
The results revealed that the main barrier to adoption was the need for users to actively initiate the feature. From a user-journey perspective—one that had already been optimised—introducing the chatbot felt like a detour and introduced unnecessary friction.​
DESIGNING FOR AI
​Reconsidering the value of AI, particularly its ability to automate repetitive tasks and support decision-making, we returned to the drawing board with a stronger emphasis on the AI’s user experience.
​​​​
We divided the user journey into two key phases and tailored our AI implementation differently for each phase to better support the user's needs:
-
Product Discovery phase
-
Purchase Intent phase
PRODUCT DISCOVERY PHASE
Beauty Bae was integrated into the search function to better support users’ needs and enhance their discovery experience. Traditionally, users had to recall specific brand or product names to find what they were looking for, limiting exploration and potentially missing relevant options.
With AI-powered search, users can now interact naturally—using queries in plain language—and receive more accurate, relevant, and personalised results. This not only improves product discovery but also encourages exploration, increases engagement, and helps users make more informed purchasing decisions.
PURCHASE INTENT PHASE
At its core, Sephora Beauty Bae is a natural language processing model. This makes information-dense areas such as product descriptions and customer reviews ideal opportunities for the AI to summarise content and offer tailored recommendations.
​
Rather than requiring users to manually trigger a prompt, integrating Beauty Bae directly into the product detail page—with its output generated by default—creates a far smoother and more intuitive user experience.