Case study
Information Architecture
Creating a shared structure to improve decision quality at scale
Focus Systems decision quality, scalability
Work type IA framework, principles, alignment
Output the shared model for feature placement and surface behavior
Context
As the mobile keyboard evolved, it accumulated features faster than it developed a shared structure. Grammar suggestions, AI rewrites, voice input, and growth surfaces were added incrementally, often solving local problems without a unifying system.
This created recurring questions across Design, Product, and Engineering: where new capabilities should live, how users should discover AI features without cluttering the interface, and how to scale across mobile and desktop without fragmenting behavior. The issue wasn’t UI polish. It was missing information architecture.
The core problem
The keyboard lacked a clear mental model for both users and teams. Features existed, but relationships between them were unclear. Decisions were repeatedly debated from scratch, and small additions risked long-term complexity. As AI capabilities expanded, the cost of not having a system increased.
The key decision
Design the structure before designing the interface. Instead of starting with screens or flows, define what the keyboard fundamentally is, how capabilities relate, and where responsibility begins and ends across surfaces.
This meant temporarily slowing UI execution to create a scalable framework.
My role
I led the information architecture work end-to-end: framing the problem as a system issue, defining scope and principles, aligning mobile and desktop behavior, and translating strategy into a usable framework.
I partnered with Product, Growth, AI, and Platform teams to ensure the system supported user needs and business goals while protecting the product’s core value as a writing-first companion.
The leverage move
I introduced a four-layer IA model that gave teams a shared language for discussing scope, behavior, and trade-offs across platforms.
-
Entry Surface — a context-aware starting point that responds to typing state and system signals.
-
Feature Hub — a predictable, user-initiated space for discovering and accessing tools.
-
AI Assistant Surface — the primary writing workspace, with compact and extended states depending on task depth.
-
Support Surface — feedback, confirmation, and learning moments tied to outcomes.
This structure became the shared language for discussing new ideas, resolving trade-offs, and evaluating scope.
Trade-offs and constraints
We deliberately chose not to optimize for maximum discoverability upfront, expose every capability at all times, or treat the keyboard as a general-purpose assistant. Instead, we prioritized clarity over density, progressive discovery over immediate exposure, and consistency across platforms over local optimization.
What changed
The impact was primarily organizational and strategic. Teams shared a common understanding of how the keyboard works; new features could be introduced without reiterating fundamentals; and mobile and desktop behavior aligned more naturally. Conversations shifted from “where does this go?” to “does this fit the system?”
Validation approach
We validated the IA through shared understanding, not solely on usability metrics: could teams clearly explain what each surface was for, could every feature map cleanly to a single surface, and did transitions between states feel logical and predictable. When answers converged, the architecture held.
Reflection
This project reinforced that design leadership is often about creating clarity that others can build on, not producing visible artifacts. By defining structure early, we enabled faster, more confident execution later, and protected the product from incremental complexity as AI capabilities continue to evolve.
In one sentence
I led a system-level redesign of the keyboard’s information architecture, creating a scalable framework that aligned teams, improved decision quality, and enabled AI capabilities to grow without fragmenting the user experience.