The goal was to streamline both manual and automated test planning by introducing intelligent AI assistance. I contributed by prototyping interaction models, shaping interface behavior, and refining Copilot user flows within the Azure DevOps environment.

I explored how Copilot could be contextually introduced throughout the test creation and management workflow:

• Embedded vs. panel: Explored whether Copilot should live within the existing UI or appear as a floating panel, ultimately favoring embedded placement for contextual continuity.
• Interaction models: Designed and tested chat, inline, and modal interactions to find the right balance between guidance and user autonomy.
• Context-aware prompts: Based on historical data patterns, I created smart prompts for generating test descriptions, estimating planning time, and setting priorities.

In cross-functional syncs, I tackled several key design challenges:

• Evaluated whether Copilot should be globally accessible or embedded contextually within specific ADO workflows.
• Aligned interaction patterns with GitHub and VS Code to ensure consistency across Microsoft’s ecosystem.
• Proposed a “sandbox” space for exploratory Copilot interactions beyond structured workflows.

I also contributed to an adjacent effort: redesigning a diagnostics dashboard to help developers better interpret test failures. My focus areas included:

• Designed visualizations that told a clear story around pass/fail trends and flakiness metrics.
• Integrating Copilot-assisted insights for root cause analysis.
• Enhancing export features to support collaboration between QA and engineering.

• Established foundational UX patterns for scaling Copilot across Azure DevOps features, including test diagnostics and failure analytics.
• Accelerated decision-making on UI placement and interaction models through rapid prototyping and stakeholder testing.
• Improved cross-team collaboration by contributing to a unified vision of how AI can support engineering workflows at scale.
• Enhanced user experience by designing context-aware Copilot prompts that reduced manual input and improved test planning efficiency.

To adopt AI in a meaningful way, users must understand it, control it, and know when to trust it.

In this project, I designed Copilot as a collaborative assistant that surfaces test insights with clarity and context. For QA engineers, trust means knowing why a suggestion appears, how it was generated, and when to act on it.

My focus was on building transparent, explainable interactions—delivering assistive intelligence, not just automation for it's own sake.