Case Study
Amazon Q for Developers
Utilizing the latest AI technology to make developing software easier and dare I even say… delightful?
Role
Lead designer
Team
AWS AI/ML
Time
2021-2024
Areas
INTERACTION
AI/ML
SYSTEMS
Delivering a Minimum Lovable Product
I joined this project as the first designer to transform an early AI/ML prototype into a user-friendly product that could suggest code to developers as they work. The main challenge was finding a way to present these suggestions without interrupting developers' focus and workflow. Through user interviews and research, I discovered that developers were hesitant about AI-generated code and wanted full control over any changes to their work. This insight led me to design a new interaction pattern that gave users more control and transparency.
Early wireframe, showing an inline suggestion as "ghost text"
Early prototype showing suggestions, used for usability testing
Final UI is simple, but making automatic suggestions feel snappy and unobtrusive took many iterations.
The solution proved successful – after launching to a private beta in early 2022, the product quickly grew to over 3,000 monthly active beta users who could now write code more efficiently while maintaining their preferred workflow.
As of 2024, this feature has the highest acceptance rate on the market.
2. Building a Suite of Features
Following our successful private beta, we faced the exciting challenge of preparing for our public preview while balancing new opportunities and user feedback. I focused on two key areas: improving the core experience and exploring new applications of our technology. Through user research and testing, I identified critical improvements needed in performance and keyboard interactions, while also discovering high-value opportunities like security risk detection and team-specific code suggestions.
User Journey map, summarizing research findings and feedback from Private Beta
This research shaped our product roadmap, leading to a successful staged rollout—starting with a closed preview in June 2022, expanding to larger organizations, and culminating in a full public launch in early 2023.
Example of additional functionality we added: scan for security issues in code.
New user onboarding, based on common pitfalls identified from testing and feedback.
3. Transition to AI Agents
We recognized that simply suggesting code wasn't enough—developers needed a way to communicate their intentions and make complex, multi-file changes. Working with a small group of developers, we reimagined our product as a conversational AI assistant that could understand and help with development tasks.
Diving into modern conversational design with variable output from an LLM
Through extensive user testing, we learned that users expected quick responses, needed clear examples, and preferred to start fresh conversations for each task. The result was Q Developer Agent, which launched to public preview in late 2023 and quickly became the #1 Verified autonomous code generation agent (source: SWEbench.com).
Agent welcomes users to set expectations on latency and required input.
Agent provides high-level updates to provide transparency and communicate progress.
Once code generation begins, agent provides updates so users can follow logic.
Once suggestions are ready, user can open a "diff" view to review changes to their code.
Building on this success, I worked with PMs to identify opportunities to expanded our vision by creating specialized AI agents for unit testing, documentation, and code review. It's around this time that more designers joined my team, so I led and worked alongside a small group of designers to bring these to life as a cohesive set of powerful agents within the same product. We launched these agents in November 2024!
CEO Matt Garman at re:Invent 2024. Watch the full announcement here!
Demo: 'Explore' AI Agents and Quick Start from Chat