Policy Controls

Creating plug-and-play policies and platform-aligned UX to unblock enterprise adoption at scale.

Context

As Copilot and AI capabilities rolled out across Viva, IT admins needed flexible, secure, and centralised control over feature access. However, initial admin experiences were fragmented, non-scalable, and lacked coherence across surfaces, blocking deployments and creating privacy concerns.

Ecosystem alignment • Microsoft • Lead Designer

Understanding

Context

Designed admin controls for Copilot and AI features across Viva and MAC.

Solution

Created a scalable, plug-and-play panel aligned with platform UX and introduced a new policy grouping model.

User audience

Enterprise IT admins managing AI access, compliance, and deployment.

Result

Unblocked adoption, improved coherence across admin surfaces, and future-proofed policy management at scale.

My responsibilities

  • Drove end-to-end UX design for AI admin controls and VFAM policy management across Viva and MAC.

  • Led information architecture strategy by introducing scalable grouping patterns (e.g. org-wide vs. custom) to replace unscalable product-based tabs.

  • Initiated and pitched V2 design alignment with MAC platform standards (L2 panels), balancing delivery velocity with long-term UX coherence.

  • Collaborated cross-functionally with PMs, MAC senior designers, and platform teams to align roadmap, technical constraints, and design fidelity.

  • Facilitated multiple design critiques across internal and external teams to iterate and socialize direction.

  • Conducted unmoderated user research to validate policy categorization and IA concepts.

  • Acted as systems-level connector, applying patterns across projects to unify admin experience strategy and ensure design consistency.

  • Influenced product decisions by framing design trade-offs and advocating for strategic UX investments beyond immediate asks.

Phase 1: Establishing a foundation with VFAM

Shipped baseline controls fast, then uncovered a scalability gap in policy architecture

We launched VFAM (Viva Feature Access Management) with product-based tabs (e.g., Viva Goals, Learning) to meet urgent deployment timelines. While we knew this IA wouldn’t scale with the growing Viva suite, it allowed us to unblock adoption quickly and establish a baseline for policy management.

→ I led design execution and internal crits, while proactively identifying structural gaps we’d need to address in future iterations.

VFAM Policy creation

I designed the initial VFAM interface to support basic policy management: listing policies, and enabling create/edit/delete workflows, panel-based. Filtering was scoped but postponed due to engineering constraints tied to legacy infrastructure.

Principled deviation

A key design decision was whether to use a wizard (as recommended by MAC UX guidelines) or a single-step panel.

Based on strong customer feedback, we chose the panel approach to reduce friction, preserve context, and support high-efficiency workflows. This marked our first principled deviation from MAC’s standard patterns, one that was accepted due to the clarity of the use case and our direct validation with real admins.

This decision not only improved the initial UX but also paved the way for embedded, modular policy editing in future controls like ULC, accelerating admin workflows across surfaces.

Phase 2: Building the AI controls panel

Delivered fast with known UX debt, then leveraged platform patterns to unify admin experiences

To unblock Copilot adoption, we shipped a V1 panel in MAC with all controls on L1. While not platform-aligned, it met urgent compliance needs.

→ While designing ULC controls, I saw an opportunity to reuse its category-based logic in VFAM. I pitched this shift and led a redesign to replace the unscalable tab model with grouped policies - bringing consistency and scalability across admin surfaces.

User Data Controls

To unblock Copilot adoption, we shipped a V1 panel in MAC with all controls on L1. While not platform-aligned, it met urgent compliance needs.

Shipping with tradeoffs: a tabbed model we knew wouldn’t scale

The initial VFAM release used a tabbed layout, giving each Viva product its own tab (e.g., Goals, Learning, Insights). While we recognised early on that this structure wouldn’t scale - there were already 9 products with more incoming - it allowed us to ship a working solution quickly. This deliberate trade-off unblocked early adoption while setting the stage for a more scalable, category-based model in future iterations.

Turning a pattern into a platform: How ULC Controls informed VFAM’s evolution

While designing the ULC controls, I realised the same logic - scoping policies by category and granularity - could solve the scalability issues in VFAM. I pitched this alignment to PMs and initiated a redesign of VFAM to group policies by type (e.g., org-wide vs. custom), replacing the unscalable tab-based model and allowing for a better policy management experience.

orgw1.png
orgw2.png
1.png
2.png

Also, this was the perfect opportunity to align with the MAC guidelines and bring policies to the L2 of the panel.

Cross-surface coherence: Improved trust and discoverability by aligning admin controls with MAC-wide patterns

  1. Design for scale: Future-proofed IA to handle growing AI feature sets and diverse enterprise needs

  2. Faster unblockers: Helped PMs meet privacy/legal (CELA) requirements with modular, compliant UX

  3. Strategic ownership: Independently identified and initiated key changes that aligned roadmap goals with user-centric and platform-centric design

Partnered closely with:

  • PMs for VFAM and AI roadmap delivery

  • MAC senior designers for V2 platform alignment

  • Viva design team for internal reviews

  • Ran research independently to validate key IA decisions

FAQ

  • In VFAM, platform guidance recommended a multi-step wizard for policy creation. However, through direct feedback from IT admins, we found that a single-step panel better supported their workflow, especially when creating or editing multiple policies quickly. I pushed for a principled deviation, backed by research, and got alignment with the MAC design team. This was our first accepted departure from guidelines and set a precedent for balancing real-world usage with design systems.

  • Scalability. VFAM originally used a tab-per-product structure that quickly became unmanageable as Viva’s suite expanded. The real breakthrough came while designing ULC controls as I realised we could apply a category-based policy model across both surfaces. I pitched and drove that redesign, turning a reactive IA issue into a reusable, systems-level solution.

  • I collaborated deeply with PMs and MAC designers, and ran multiple rounds of crits internally and externally. One example: I proactively pitched the move to an L2 panel experience for MAC controls, despite the original V1 shipping on L1 for speed. I led alignment conversations, brought research and rationale, and drove consensus across teams.

  • My design decisions enabled modular reuse of controls and reduced friction for future AI features. By embedding VFAM logic directly into AI panels, we eliminated context switching and helped unblock compliance-heavy deployments. This made it easier to launch new controls without needing new infrastructure each time.

  • We had to defer advanced filtering due to legacy code constraints. With more time, I would’ve prioritised a refactor to support smarter filtering and tagging, especially as the number of policies grows. I’d also push for earlier alignment with MAC platform evolution to reduce rework between V1 and V2 designs.