Personalised Gift Cards Using GenAI
Personalised Gift Cards Using GenAI
Reimagining Digital Gifting for Amazon Pay Users
Reimagining Digital Gifting for Amazon Pay Users
Client
Client
Amazon
Amazon
My Role
My Role
UX Designer
UX Designer
Project Type
Project Type
Innovation Sprint
Innovation Sprint
Project Duration
Project Duration
Jan 2023- Sept 2023
Jan 2023- Sept 2023
Projected Impact
Projected Impact
Could generate over 180 crores in additional annual revenue
Will reduce manual effort by ~30 minutes for each new gift card design
Expected to improve page performance by minimizing reliance on static assets
Likely to boost engagement through personalized and culturally relevant gifting
Could generate over 180 crores in additional annual revenue.
Will reduce manual effort by ~30 minutes for each new gift card design.
Expected to improve page performance by minimizing reliance on static assets.
Likely to boost engagement through personalized and culturally relevant gifting.
Background
Background
Amazon Pay’s gift card experience was functional but limited. Users could only choose from a fixed set of static designs, which often didn’t reflect specific occasions or cultural moments.
This project aimed to use Generative AI (via Amazon Bedrock) to make gifting more personal, emotional, and scalable. Designed as part of an innovation sprint, it was scoped as a proof of concept to test feasibility before a full rollout.
Amazon Pay’s gift card experience was functional but limited. Users could only choose from a fixed set of static designs, which often didn’t reflect specific occasions or cultural moments.
This project aimed to use Generative AI (via Amazon Bedrock) to make gifting more personal, emotional, and scalable. Designed as part of an innovation sprint, it was scoped as a proof of concept to test feasibility before a full rollout.
The Aim
The Aim
Enable personalized gifting at scale
Enable personalized gifting at scale
Allow users to generate custom card images and personalised messages using GenAI
Allow users to generate custom card images and personalised messages using GenAI
Reduce operational overhead
Reduce operational overhead
Reduce the operational load on internal teams maintaining hundreds of static designs
Reduce the operational load on internal teams maintaining hundreds of static designs
Expand cultural relevance
Expand cultural relevance
Expand the reach and resonance of gift cards by offering culturally relevant and diverse content
Expand the reach and resonance of gift cards by offering culturally relevant and diverse content
The Problem Space
The Problem Space
Limited Personalization Options
Limited Personalization Options
Users were limited to static designs, resulting in generic and emotionally flat experiences.
Users were limited to static designs, resulting in generic and emotionally flat experiences.
Message Creation Fatigue
Message Creation Fatigue
Users found it hard to craft meaningful messages quickly, often settling for generic text.
Users found it hard to craft meaningful messages quickly, often settling for generic text.
700+
700+
static gift card designs are manually created and maintained by the design and ops teams.
static gift card designs are manually created and maintained by the design and ops teams.
30%
30%
drop-off rate during customization due to limited design/message options (internal hypothesis)
drop-off rate during customization due to limited design/message options (internal hypothesis)
Manual Design Creation Slows Teams
Manual Design Creation Slows Teams
Each design takes 2–5 days and cross-team effort, causing delays and bottlenecks.
Each design takes 2–5 days and cross-team effort, causing delays and bottlenecks.
Cultural Gaps
Cultural Gaps
Many regional festivals or languages (e.g. Bihu, Odiya) have no design representation.
Many regional festivals or languages (e.g. Bihu, Odiya) have no design representation.
Research and Strategy
Research and Strategy
I worked closely with the product and research teams, using the insights they gathered about user problems. My role was to turn those insights into easy-to-use design flows and safe ways to bring GenAI into the experience.
I worked closely with the product and research teams, using the insights they gathered about user problems. My role was to turn those insights into easy-to-use design flows and safe ways to bring GenAI into the experience.
I collaborated with stakeholders to understand:
I collaborated with stakeholders to understand:
Customer frustrations with the current experience
Customer frustrations with the current experience
Time-consuming process of creating and launching new gift card designs
Time-consuming process of creating and launching new gift card designs
Technical capabilities and limitations of Amazon’s GenAI APIs
Technical capabilities and limitations of Amazon’s GenAI APIs
Legal and ethical concerns around user-facing AI outputs
Legal and ethical concerns around user-facing AI outputs
Research Highlights
Research Highlights
High User Demand for Personalization
High User Demand for Personalization
78% of users wanted more control and personalization options
78% of users wanted more control and personalization options
Manual Design Creation Slowed Ops
Manual Design Creation Slowed Ops
Ops teams spend up to 30 mins creating each unique gift card design manually
Ops teams spend up to 30 mins creating each unique gift card design manually
AI Filters Not Foolproof
AI Filters Not Foolproof
Amazon Bedrock’s filters reduced but didn’t eliminate risks of offensive or biased content
Amazon Bedrock’s filters reduced but didn’t eliminate risks of offensive or biased content
Legal Red Flags on Content Safety
Legal Red Flags on Content Safety
Legal flagged offensive content, and AI bias as major go/no-go issues
Legal flagged offensive content, and AI bias as major go/no-go issues
Early Ideation & Concept Exploration
I explored how users approach gift personalization, which led to two early concept directions based on their needs:
Prompt-Driven Customization
This allows users to describe exactly what they want, offering complete flexibility but carrying legal and content risks.
Occasion-Based AI Suggestions
A safer and structured approach that lets users choose from curated AI-generated options based on the occasion and recipient.
After reviewing technical feasibility and user desirability with stakeholders, I evolved these into the Northstar and MVP flows.
Early Ideation & Concept Exploration
I explored how users approach gift personalization, which led to two early concept directions based on their needs:
Prompt-Driven Customization
This allows users to describe exactly what they want, offering complete flexibility but carrying legal and content risks.
Occasion-Based AI Suggestions
A safer and structured approach that lets users choose from curated AI-generated options based on the occasion and recipient.
After reviewing technical feasibility and user desirability with stakeholders, I evolved these into the Northstar and MVP flows.
Designing for AI : Two Versions
Designing for AI : Two Versions
The end goal was to give users full creative autonomy. To get there, I developed two approaches:
The end goal was to give users full creative autonomy. To get there, I developed two approaches:
The Northstar flow, which allows users to describe exactly what they want, offering complete flexibility but carrying legal and content risks
The MVP flow, a safer, structured approach that lets users choose from curated AI-generated options based on the occasion and recipient.
While the Northstar flow represents the long-term vision, the MVP was prioritized for initial development due to its feasibility and lower risk.
The Northstar flow, which allows users to describe exactly what they want, offering complete flexibility but carrying legal and content risks
The MVP flow, a safer, structured approach that lets users choose from curated AI-generated options based on the occasion and recipient.
While the Northstar flow represents the long-term vision, the MVP was prioritized for initial development due to its feasibility and lower risk.
The Northstar
The Northstar
Let users express any design idea through natural language prompts.
Let users express any design idea through natural language prompts.
Maximum creativity
Maximum creativity
Emotional resonance
Emotional resonance
The MVP
The MVP
Deliver personalisation safely using structured inputs.
Deliver personalisation safely using structured inputs.
Safer content generation
Safer content generation
Faster load times
Faster load times
Designing for AI : Balancing UX and Risk
Designing for AI : Balancing UX and Risk
I approached GenAI not as a feature drop, but as a system that demands trust, clarity, and thoughtful design.
I approached GenAI not as a feature drop, but as a system that demands trust, clarity, and thoughtful design.
I designed not just for what could go right, but for what could go wrong. My goal was to make collaboration with AI feel safe and empowering.
I designed not just for what could go right, but for what could go wrong. My goal was to make collaboration with AI feel safe and empowering.
Designing for Edge Cases & Risk
Designing for Edge Cases & Risk
I worked with PMs and legal to anticipate worst case scenarios not just to avoid harm, but to intentionally design for safety and trust.
I worked with PMs and legal to anticipate worst case scenarios not just to avoid harm, but to intentionally design for safety and trust.
We identified high-risk prompt categories and created proactive UX guardrails to mitigate them.
We identified high-risk prompt categories and created proactive UX guardrails to mitigate them.
Legal Risk
Legal Risk
Copyright Infringement, Misinformation or Fake Claims
Copyright Infringement, Misinformation or Fake Claims
User Experience Risks
User Experience Risks
Poor handling of failed generations
Poor handling of failed generations
Cultural and Ethical Sensitivities
Cultural and Ethical Sensitivities
Religious or cultural insensitivity, Public figure misuse
Religious or cultural insensitivity, Public figure misuse
Content Risks
Content Risks
Offensive or obscene content, Disturbing imagery, Brand misuse
Offensive or obscene content, Disturbing imagery, Brand misuse
UX Guardrails I Designed
UX Guardrails I Designed
My goal was to empower users without overwhelming them, while balancing safety with creativity. To address risks without limiting expression, I designed the following guardrails:
My goal was to empower users without overwhelming them, while balancing safety with creativity. To address risks without limiting expression, I designed the following guardrails:
Guided Task Flow
Guided Task Flow
Guided users through one task at a time using progressive disclosure and clear layout.
Guided users through one task at a time using progressive disclosure and clear layout.
Safety & Sensitivity
Safety & Sensitivity
Added tooltips and fallback messages for sensitive prompts and safer outputs.
Added tooltips and fallback messages for sensitive prompts and safer outputs.
Affordance Clarity
Affordance Clarity
Clearly marked editable and regenerable content to keep users in control.
Clearly marked editable and regenerable content to keep users in control.
MVP Prompt Limits
MVP Prompt Limits
Replaced open prompts with dropdowns in the MVP to minimize risky inputs.
Replaced open prompts with dropdowns in the MVP to minimize risky inputs.
Content Preview
Content Preview
Introduced a “Review & Confirm” screen with content guidelines and editing options.
Introduced a “Review & Confirm” screen with content guidelines and editing options.
Error Recovery
Error Recovery
Included regenerate and revert actions to help users recover from poor AI results.
Included regenerate and revert actions to help users recover from poor AI results.
Next Steps
Validate the MVP with Real Users
Test the MVP to see if users understand the flow and find the AI suggestions useful.
Evolve the Northstar Vision
Keep building toward open-ended customization as AI tools and legal policies mature.
Broaden Visual Style Options
Add more design styles to better reflect user intent and diversity.
Key Takeaways
Safety and creativity can work together
I learned how to balance giving users freedom while keeping the experience safe and responsible.
Real-world constraints made me think differently
Designing around legal and ethical risks pushed me to go beyond usual UX patterns.
This project changed how I approach design
This project made me more systems-minded and helped me design better for scale and emerging tech like GenAI.
Next Steps
Validate the MVP with Real Users
Test the MVP to see if users understand the flow and find the AI suggestions useful.
Evolve the Northstar Vision
Keep building toward open-ended customization as AI tools and legal policies mature.
Broaden Visual Style Options
Add more design styles to better reflect user intent and diversity.
Key Takeaways
Safety and creativity can work together
I learned how to balance giving users freedom while keeping the experience safe and responsible.
Real-world constraints made me think differently
Designing around legal and ethical risks pushed me to go beyond usual UX patterns.
This project changed how I approach design
This project made me more systems-minded and helped me design better for scale and emerging tech like GenAI.
Final Notes
Final Notes
This project taught me to think beyond the screen, to design for risks, edge cases, and everything that could go wrong (and how to make it right). The Northstar flow might not launch today, but it’s changed how I think about designing for scale and what’s next.
This project taught me to think beyond the screen, to design for risks, edge cases, and everything that could go wrong (and how to make it right). The Northstar flow might not launch today, but it’s changed how I think about designing for scale and what’s next.