AI Dashboard for Microsoft Entra

The Problem

Enterprise administrators were overwhelmed by hundreds of security signals. Existing dashboards surfaced data, but didn’t help admins understand what mattered most or what to do next.

Design Goals

Design an AI-first dashboard that could:

• Prioritize security issues
• Explain why something mattered
• Recommend the next action
• Build trust in AI recommendations

My Role

• Product Design
• AI UX Strategy
• Interaction Design
• Executive Prototyping
• Design Systems
• Cross-functional collaboration

Research / Journey

We began by synthesizing existing telemetry, customer conversations, and product feedback to understand where administrators were spending the most time and where decision making became difficult.

Rather than redesigning the existing dashboard, we explored what an AI-first administration experience could look like. Through rapid concepting, we evaluated multiple approaches for surfacing recommendations, explaining AI-generated insights, and guiding administrators toward their next best action.

Using Figma, AI-assisted prototyping, and close collaboration with engineering, we quickly evolved ideas into interactive prototypes that tested recommendation frameworks, dashboard layouts, and reusable AI interaction patterns.

This rapid design-engineering loop allowed us to validate concepts early while shaping a scalable foundation for future AI-powered administration experiences across Microsoft Entra.

Building a Decision Intelligence Framework

One of the biggest design challenges wasn’t designing another dashboard - it was creating a reusable framework for presenting AI-generated recommendations consistently.

Rather than displaying isolated alerts, every recommendation followed the same interaction model so administrators could quickly scan, understand, and act.

Each recommendation card included:

• AI-generated summary
• Severity indicator
• Intelligence source
• Recommended next action
• Supporting context

This consistency reduced cognitive load while making AI recommendations predictable and trustworthy.

AI Recommendation Cards

The dashboard unified multiple sources of intelligence into a single experience while preserving transparency around where recommendations originated.

Copilot Generated Insights

Generative AI synthesized security signals into concise, human-readable recommendations that explained what changed, why it mattered, and suggested the appropriate next step.

Agent Generated Recommendations

Autonomous security agents continuously monitored the environment and surfaced completed investigations, delegated tasks, or recommended actions requiring administrator review.

Machine Learning Signals

Backend machine learning models surfaced deterministic detections and risk signals while maintaining the same visual interaction pattern used across AI-generated experiences.

A Unified Card Pattern

Every recommendation shared a consistent structure regardless of where the intelligence originated.

Each card included:

• Source indicator (Copilot, Agent, or Machine Learning)
• Severity badge
• AI-generated two-line summary
• Primary action: Review
• Secondary action: Investigate with Copilot

Keeping interactions consistent helped administrators quickly recognize recommendation types while building confidence in AI-assisted decision making.

Designing for Trust

Introducing AI into enterprise administration required more than generating recommendations.

The experience needed to communicate:

• Why a recommendation appeared
• What evidence supported it
• Where the intelligence originated
• What action was recommended
• When human review was required

Every interaction reinforced administrator control while reducing unnecessary cognitive effort.

Impact

This exploration established foundational interaction patterns for AI-first administration experiences across Microsoft Entra.

• AI-first administration concepts
• Decision intelligence dashboards
• Conversational administration workflows
• Agent-powered recommendations
• Reusable AI interaction patterns
• Human-in-the-loop review experiences

The work helped shape a broader vision for how enterprise administrators move beyond monitoring dashboards toward AI-assisted decision making.

End to End Demo