︎
Amazon —  Senior UX Writer, Amazon Prime WW (April 2024 - Present)

Context:
To increase relevance and reach of Amazon Prime value, as well as awareness of how Prime benefits can apply to customer’s interests and orders, the Acquisition team wants to design a curated recommendation widget powered by generative AI to surface products and digital content that align with customers’ recent browsing and purchasing activity. In order to do so, the AI model will need guardrails around content creation—specifically voice, tone, and style—to ensure auto-generated messaging is tailored to the customer while remaining on-brand.


• I developed a contextual system prompt comprised of messaging rules and a content design template to train a branch of the Amazon Bedrock LLM to generate headlines and descriptive microcopy that capture Prime members’ interests and help them to achieve more with the benefits they’re already paying for
• I collaborated with Engineering to create a natural language heuristic for selecting which elements of a customer’s profile and activity to combine into relevant messaging that inspires Prime action, creating clear connections between membership and personalized value
• I trained the AI model through reinforcement learning and iterated on guidelines when edge-cases (e.g. a customer shopping on behalf of someone else) produced hallucinations

Results:
  • PriME (pronounced Pry-Mee) is a personalized GenAI CX launched on July 18, 2025 to 260 Amazon UK beta customers, and included:
  1. The generation of 260 deep, natural language customer personas based on interests identified from 24 months of activities across shopping, Prime Video, and music.
  2. A total of 818 unique, personalized, multi-modal GenAI CXs rendered through varying widgets across the Prime member experience on iOS and Android. Each featured a personalized image, contextual header and/or subheader, and 3 multi-benefit ASIN/digital title recommendations from the observed categories.
  3. Compliance with legal/privacy/marketing/UX guardrails (validated by AI and humans).
Customer response:
  • Of the 13% of survey respondents made aware of new benefits (primarily Amazon Music) thanks to PriME, 60% said they were likely to use them. This encourages the team to feature content for lesser known benefits to increase awareness.
  • 24% of CXs had a positive impact, and 6% had a negative impact on respondents’ value perception of Prime (net +1800bps positive impact), suggesting that while this method of personalization is effective, opportunity remains to improve value perception.
  • 42% felt the CX was relevant to their interests, and 28% felt the CX wasn’t. Opportunity remains to improve relevancy through both product recommendations and messaging frames for connecting interests with Prime services.

PriME system prompt 📝→🤖

Customer personas (left) & personalized AI-powered inspiration (right):





🔙