AC3 Social Media

Designing the future of omnichannel customer support across social and AI-powered platforms


Vision design for how Amazon CSA handle social media contacts

As part of Amazon’s Customer Service Experience initiative, I led the end-to-end UX transformation of Amazon’s social-media-based customer-service platform, which spans 704 handles across 9 global networks. The mission was to unify these fragmented experiences under a single, intelligent platform that supports scalable, consistent, and human-centered interactions.

Our long-term vision is to onboard all social media platforms into one unified customer service platform within 24 months, starting with a 3-month P0 focus on X (formerly Twitter) — which currently represents 40% of all social-media-based Amazon contacts.

My Role

As the UX lead, I was responsible for:

  • Driving research and design strategy across both customer-facing and agent-facing workflows

  • Conducting customer interviews and usability testing for early prototypes (linking Amazon accounts with X and ChatGPT)

  • Designing scalable patterns that support complex query variations across markets

  • Defining taxonomy and standardizing tagging systems to unify 9 network workflows

  • Partnering with Product, Tech, and Operations to ensure design decisions translated into measurable service efficiency

Challenge

Amazon’s existing social engagement platform (2023 data):

  • Reviewed 29.3M mentions, labeling 6.4M (21.8%) as “engageable”

  • Associates manually reviewed 13.9M (60.7%) non-engageable messages

  • Of those, 156K (1.1%) were false negatives that required manual intervention

This manual overhead significantly reduced efficiency. Our goal was to reduce 75% of these hours, targeting a 2.27% increase in Contacts Per Hour (CPH) — from 8.8 to 9.0, representing an estimated $18M annual savings based on Amazon’s $2.98 variable cost per contact (VCPC).


Journey map and initial concepts

I lead a design sprint with various cross vertical participants to understand the current challenges and to create possibilities for the future as part of this i have designed a 3 day workshop and at the end of 3rd day i came up with journey map and 2 concepts to understand customer sentiment

Prototype testing and user interviews

Wexplored how customers perceive linking their Amazon accounts with social media and AI platforms (ChatGPT) for direct, personalized support and as researcher was handling user interviews from various geographic locations

Key takeaways

Comfort & Trust

  • Customers showed higher trust linking Amazon with social platforms (familiar, conversational, brand-associated).

  • AI integrations generated curiosity but hesitance, stemming from unclear mental models and privacy concerns.

  • Trust increased when the data-sharing process felt familiar (Amazon login, clear branding, or face recognition).

“I trust Amazon in taking safety precautions — but ChatGPT feels newer; I don’t know what happens to my info.”
— Customer 1, MN (25)

Data Privacy & Transparency

Customers demanded clarity before consent:

  • Preferred seeing what data is shared before validation.

  • Wanted easy unlinking options post-resolution.

  • Appreciated transparency in the ChatGPT prototype that specified “order details, delivery info, notifications.”

“It should say before you agree to link — here’s what you’ll give access to.”
— Customer 5, NY (27)

AI vs. Human Interactions

  • Tone empathy (“Allowing us a chance to fix this”) mattered more than identity.

  • Some messages felt “scripted”; customers valued natural acknowledgment and personalization.

  • Majority cared more about efficiency and resolution, not whether support was human or AI — except for complex issues.

“If the outcome is what I want, I care less who gives me my money.”
— Customer 3, MD (56)

Communication Medium & History

  • Text messaging was overwhelmingly preferred — professional, reviewable, less error-prone.

  • Voice messages fit specific use-cases (mobile or rushed).

  • All users wanted the ability to save chat history, particularly for credits or concessions, accessible via both Amazon and social accounts.


Design Recommendations

Product & Experience

  • Introduce pre-consent transparency screens before linking.

  • Provide linking confirmation and unlink instructions post-resolution.

  • Publicly announce collaborations (e.g., Amazon × ChatGPT) to normalize new support paradigms.

  • Create segmented onboarding flows for early adopters vs. cautious users.

Design & Interaction

  • Explicitly label AI vs. Human messages.

  • Embed empathetic phrasing in both AI and agent templates.

  • Incorporate tone adaptation logic based on sentiment analysis.

  • Build modular design patterns scalable across 9 social networks.

  • Maintain secure, unified chat history accessible across all linked channels.

Proposed Designs

Customer-Facing Experience

Objective: Deliver transparent, empathetic, and efficient omnichannel support.

Design Highlights:

  • Unified Linking Flow: Clear disclosure of shared data before consent, single-tap unlink after resolution.

  • Hybrid Support Model: AI handles quick-resolution tasks (refund, reorder, delivery change) → human agents handle complex cases.

  • Trust Indicators: Verified Amazon logo, encryption badge, contextual greeting (“Hi Maya, I see your recent order for …”)

  • Chat History Management: End-of-chat option to save transcript; email copy sent automatically for concession cases.

  • Metrics Impact: Anticipated 75% reduction in manual review hours → +2.27% CPH gain → $18M YOY savings.

Tech-Support / Agent-Facing Experience

Objective: Equip associates with context, efficiency, and cross-channel visibility.

Design Highlights:

  • Unified Agent Console: Single dashboard showing linked social/AI accounts, message sentiment, and escalation history.

  • AI-Assist Layer: Suggests next-best actions or response drafts, editable by human associates for tone refinement.

  • Privacy Controls: Agents can trigger “unlink post-resolution” actions directly within the interface.

  • Conversation Timeline View: Consolidated thread of all customer interactions across channels.

  • Analytics & Feedback: Real-time KPIs (AHT, sentiment, resolution accuracy) driving continuous improvement.





Impact & Next Steps

This initiative redefined how Amazon approaches social-media-based customer engagement. By unifying disjointed networks, establishing trust-driven AI integration patterns, and grounding every interaction in transparency and empathy, we set the foundation for a scalable, secure, and efficient omnichannel experience.