AI Image Generator
AI Image Generator enables marketers to create highly engaging ads within clicks. My responsibility is to scale it responsibly by balancing user needs, business goals and long-term vision.
Role
UX lead
Areas
Strategy, Vision
Responsive web design
UI & Interaction patterns
Year
Mar 2024 - present
What’s launched
Product scale and position editing: To address the No.1 pain point, I piloted the experience enabling marketers to adjust the scale and position of their product within generated images—marking the first AI-gen editing capability at Amazon Ads.
Product scale and position editing, launched July 2024
1-click image selection: To make it easier for marketers to select AI-generated images, we leveraged pre-generation and developed a 1-click image selection experience.
1-click image selection, launched May 2024
New feature introduction patterns: As more and more new capabilities are being launched, I established new feature introduction patterns to drive product engagement.
New feature introduction, launched Oct. 2024
Side sheet migration: To provide better user experience and support more complex features, I led the transition from a modal experience to a side sheet experience.
Migration from modal to side sheet, completed Apr. 2024
Within one year of launching Image Generator, as the sole UX lead, I partnered with Product, Engineering and Science on launching a series of new features and improvements, we’ve seen
Active advertiser adoption
+1,880%
Campaign submissions
+39%
From 2.5k to 46k
Product sales
+5%
AI-GEN ads revenue
+745%
From $828k to $7M monthly
Feedback collected from advertisers, Nov. 2024
Quotes from Jay at Amazon year-end all-hands meeting, Dec 2024
Background
Amazon Ads launched the MVP AI Image Generator in October 2023, a simple AI tool for creating ad-ready lifestyle images. After its launch, I joined the team to improve and scale the product. My responsibilities include:
-
Balancing short-term business goals with the long-term vision
-
Establishing scalable patterns and mechanisms for greater impact
AI Image Generator Launch Video, Oct 2023
Long-term vision
The vision for Image Generator is to empower any advertisers to create a highly engaging creatives within minutes. To achieve this, we focus on fulfilling user needs and innovating new experiences.
Understanding the audience and discovering needs
​​
As an ongoing project, I’m partnering with UX researcher and customer success team to conduct user interviews. I also utilize the customer feedback platform and the in-app feedback selection to discover quantitative user trends.
​​​
The Image Generator serves small business owners (80% of users) seeking simplicity for major launches and big brands (65% of revenue) looking for inspiration for non-core campaigns. We prioritize ease of use while positioning the tool as a hub for creativity and inspiration.
Data breakdown by team type, by Oct 2024
Qualitative and quantitative user research, design workshops, 2024
Quotes from user interviews, 2024
Addressing the top user pain point​
​
We discovered that the biggest blocker for advertisers using images was the incorrect product size or position relative to the background—cited as the top reason for downvotes (38%). To address this, I developed the first AI-generated image editing capability. After its launch, 30% of advertisers used the feature, and downvotes related to product size or position dropped significantly to 23%.​
Advertiser adoption
30%
Negative feedback on sizing
-15%
Edited images saved or submitted
42%
Product scale and position editing, launched July 2024
Meeting user needs & igniting inspiration​
​
To support advertisers with prompt writing, I partnered with the science team to create a guided prompt feature. It provides tailored suggestions based on their products, reducing effort, improving result quality, and sparking inspiration.
Guided prompt / Ideas for you, launched Dec. 2024
Feedback collected from advertisers around UX, Nov. 2024
Live image innovation​
​
By leveraging AI text to motion capabilities, live image can remove creative barriers in rich media advertising and turn a regular display ad to eye-catching shopper experiences.
Live image WIP design
Business goals
At the end of 2023, we set an executive team goal to increase active advertiser using AI generated images from 2,523 in Q4 2023 to 50,000 in Q4 2024, an increase of 1,882% YoY. To achieve this, we focus on improving discovery and image diversity.
1-click image selection
To make it easier for marketers to select AI-generated images and improve discovery, we leveraged pre-generation and developed a 1-click image selection experience.
1-click image selection, launched May 2024
The launch of the 1-click image selection modal significantly increased the likelihood of images being selected. It enhanced the discoverability of AI-generated images, doubling their usage in campaigns (7% via the modal vs. 3% in the Image Generator).
Advertisers select an image within the modal
21%
Advertisers select an AI-GEN image within the modal
7%
Advertisers select an AI-GEN image within Image Generator
3%
Recommendation framework
​
To enhance value for advertisers and drive adoption, we automatically recommend AI-generated holiday images, enabling advertisers to refresh their creatives with a single click. We also suggest AI-generated images to advertisers who have not yet launched an image-based campaign.
Design for recommendation framework
Feature introduction patterns​
​
As we expanded the Image Generator across Amazon Ads products, I developed a standardized framework to improve feature discoverability and engagement. In collaboration with product managers, I documented component usage and set guardrails to reduce friction.
Feature introduction patterns​
Building a solid foundation
Patterns for scalability
​
As new features were added, the original modal format became a bottleneck. To enable smoother scaling, I collaborated with engineering to migrate the tool from a modal to a side sheet format. As more and more AI-GEN tools are added to Amazon Ads, I also standardized patterns for generate button, loading status so that users will have a consistent experience encountering AI.
Migration from modal to side sheet
Side sheet behavior, responsiveness and internationalization
Standardizing patterns for AI-GEN
Cross-functional collaboration improvement​
​
Noticing gaps during UAT as the team accelerated launches, I introduced a process to align UX and engineering at four key stages before UAT for early issue resolution. I also created a UX handoff checklist to standardize the process.
UX/END process improvement proposal
What’s next
​Our ultimate goal is to make Image Generator ubiquitous, seamlessly surfacing AI-generated images where advertisers need them and aligning them with brand identity.
​
It has been a remarkable year of growth (active advertisers grow from 2.5k to 46k). However, rapid growth and aggressive business goals also come with drawbacks, which led to a few feature churns. As a next step, I'll spend more time talking to users while working with Product on creating visions for 2025.