Building one experience layer for every agent, surface, and workflow, on a GRC platform trusted by 1 Million users.

At Diligent, I lead experience design for Platform AI and Core Services, helping shape the interaction foundation for an emerging agentic platform within a global GRC ecosystem serving 75% of the Fortune 500. My work spans foundational platform experiences, agent interaction patterns, human-in-the-loop workflows, and cross-surface experiences designed to scale as the platform expands with new agents and capabilities.

Building the proactive layer.

The platform could answer questions. It couldn’t yet tell users what needed their attention.

Platform AI began as a reactive Q&A experience. The next step was to make it proactive: surface the work that matters across Diligent products and help each user understand what requires attention, why, and what to do next.

The product brief established the opportunity, but much of the experience model was still undefined. I led the design of Attention Pane V1, translating that direction into the interaction architecture, rules, states, and build-ready experience needed to make it real.

Designing a focused, actionable layer between intelligence and work. I wasn't just designing responsive versions of the same UI. I was thinking about the underlying interaction contract.

Turning a product concept into an interaction system.

I translated the Attention Pane concept into a coherent interaction system: defining how actionable items are curated and structured, how users move from awareness to action, how items behave across their lifecycle, and where human control is required in agent-driven workflows. The result was a persistent work surface designed around context, action, state, and user control.

One interaction contract.

Multiple surfaces. I extended the V1 thinking beyond the homepage to explore how the same agent decision could travel across surfaces. Rather than designing separate interaction models for every host, I defined a shared decision payload — context, required input, preview, validation, and human confirmation — that could be rendered natively by Platform AI, collaboration tools, or external AI environments.

The interface changes. The interaction contract stays consistent.

I drove the Attention Pane from a high-level product concept to a build-ready interaction system, aligning UX and technical architecture throughout discovery and into implementation.

The first version of the Attention Pane established a shared view of what needs a user’s attention across the GRC ecosystem. The next challenge was personalization: different roles, responsibilities, and risk priorities require different signals, levels of detail, and delivery patterns.

01/ Designing the personalization layer (combine with intro)

As the Attention Pane evolved from a shared cross-domain view into a more personalized experience, I began designing the Briefing Agent: a configurable layer that learns what matters to each user and shapes what information is surfaced, how much context they receive, and how the briefing reaches them.

From requirements to experience model.

The product requirements defined the core configuration capabilities, but left critical interaction questions unresolved. I decomposed the model, mapped system states and edge cases, and translated the requirements into a coherent configuration experience.

One configuration surface, multiple dimensions of personalization.

Unlike traditional settings, these controls configure future agent behavior. That required designing not only the happy path, but also how users understand delayed effects, changing permissions, unavailable signals, paused states, and recovery.

Building human control and reusable interaction patterns into Platform AI.

As agents move from answering questions to taking action, the design problem shifts from conversation to control.

I worked across Platform AI, Product, Engineering, and Design to define how human judgment enters agentic workflows, how those interactions scale beyond individual features, and how AI-specific behaviors can become reusable platform patterns rather than one-off solutions.

Defining the first human intervention patterns.

I partnered with the Agentic Fabric engineering team to define the first reusable human-intervention patterns for agent workflows. I translated the interaction model into states and rules, then validated the contract with engineering so UX aligned with how workflows actually pause, resume, and proceed.

Scaling human oversight with agent autonomy.

The larger question wasn’t how to design an approval button. It was when a human should be in the loop at all.

I expanded the implementation work into a broader HITL discovery to help Product frame how human oversight should evolve as agent autonomy increases. The direction explored calibrated autonomy: human intervention changes based on consequence, reversibility, uncertainty, and task context rather than applying the same approval model everywhere.

Human-in-the-Loop is a system, not a modal.

My research reframed HITL as a platform capability that spans more than interaction design. Human oversight only works when governance, orchestration, interface behavior, evidence, and learning operate together.

Making agent interactions reusable.

Once agent behaviors repeat across products, the next challenge is making them reusable without creating a parallel AI design system.

I helped frame A2UI as a behavior and workflow layer built on top of Diligent’s existing design system. The goal was to standardize AI-specific states and interactions while preserving shared foundations, accessibility, and visual consistency across products.

AI-specific behavior can be new. AI-specific styling should be new only when necessary.

Building the system across teams.

I worked across implementation, discovery, and platform strategy at the same time, connecting immediate engineering needs with a longer-term interaction model that could scale across Diligent.

From concept to shared component behavior.

The first A2UI work focused on data-visualization patterns, with HITL interactions forming a subsequent layer. I collaborated with Product, Engineering, Design Systems, and designers across business units to understand downstream needs and feed those priorities into the shared component direction.

The work moved from individual controls to a reusable model for how people supervise, redirect, and collaborate with agents.

Across HITL and A2UI, my role was to connect interaction design, agent behavior, engineering constraints, and platform governance into a system that could support both immediate delivery and longer-term scale.