Capstone Project
I collaborated with Microsoft’s CX team to design an AI-powered access management dashboard from zero to one, streamlining permission requests and surfacing flagged activities to improve team productivity, operational efficiency, and data security at scale.
View the full prototype here.
At enterprise scale, employees interact with countless tools and datasets daily. Managing who has access to what, especially during onboarding, offboarding, or team shifts, quickly becomes complex and error-prone.
Problem Definition
Two users. One system. Very different needs.
We translated our findings into two personas and mapped their journeys using service blueprints, which exposed breakdowns in communication, missed notifications, and overly manual tasks.
Ideation
Zeroing In on Broken Access Flows and Overwhelmed Managers
I co-led an ideation workshop with Microsoft stakeholders to align cross-functional teams and identify high-impact solutions. Using research themes and HMW statements, we prioritized key pain points and focused on two core challenges: streamlining new hire access and supporting managers with security audits.
User Flow
Integrating AI features to surface access issues and simplify processes
I mapped the future state end-to-end user flow—from audits to approvals—to uncover where automation and AI could streamline the process. This aimed to solve key friction points and inefficiencies, especially around justifications, reviews, and notifications.
I led the design of the manager dashboard, focusing on surfacing high-priority actions and simplifying access audits.
The Solution
01. Access Dashboard: Visibility without Guesswork
“What does a manager need to feel in control of their team’s access?"
→ The Access Dashboard surfaces key security insights like compliance status, access gaps, and flagged activities, helping managers instantly assess team health and act on risks.
02. Intelligent Audit Planning: From Reactive to Proactive
“How can we reduce friction in audit operations?”
→ Managers can schedule audits based on team behavior, calendar events, and key deadlines. Copilot automates follow-ups and surfaces 'Priority Actions Today', streamlining both planning and execution.
03. Copilot-Powered Auditing: Smarter, Faster Decisions
“How can we guide decision-making without overwhelming users with data?”
→ In the Audit Dashboard, managers review flagged activities with AI-generated risk tagging and recommended actions, enabling fast, informed, and stress-free auditing. To further reduce cognitive load and accelerate communication, Copilot auto-generates a ready-to-send message in Microsoft Teams.
Usability Testing & Stakeholder Feedback
01. Replacing Inline Justification with a Copilot Side Panel
Original Issue: Inline AI explanations were distracting and disrupted users’ ability to scan and act on flagged activities.
Changes: I moved Copilot’s justifications into a dedicated right-hand panel, allowing users to reference detailed insights without losing visibility of the full audit table and interrupting the workflow.
02. Adding Calendar Integration to Schedule Audits
Original Issue: Managers wanted more control over when audits were conducted, especially when managing competing priorities.
Changes: I added a calendar scheduling feature to the dashboard where managers could choose a date, automatically sync it with their calendar, and receive reminders.
03. Adding AI Confidence Tooltips to Build Trust
Original Issue: Users were hesitant to trust AI-generated risk levels without context.
Changes: I introduced hover-based tooltips on risk levels showing AI confidence scores to increase transparency and trust in AI recommendations.
Final Prototype
Meet Nexus: Smarter Access, Powered by AI
Reflection
1. Designing with AI, intentionally
This project pushed me to go beyond surface-level UX and to consider how AI could not only automate tasks but actively support decision-making. I had to think critically about when and how to surface AI suggestions so they felt helpful, not overwhelming. Designing for trust, transparency, and explainability became just as important as designing for usability.
2. Guiding with Trust and Control
I learned how microinteractions like tooltips and confidence scores build trust without overwhelming users. One of my favorite parts of the project was crafting flows that gently nudged users to take action without making the system feel overly automated, supporting confident decisions rather than replacing them.
3. Iterating Through Feedback
Collaborating with teammates and stakeholders helped sharpen the design at every stage. Iteration was key to refining Copilot’s tone, behavior, and overall value, and it reminded me that good AI design isn’t static, but shaped through thoughtful collaboration.