CASE STUDY

AI Assistant Widget

An AI assistant designed to move from a small entry point into a deeper workspace, while keeping the same conversation and context across web, mobile, and kiosk environments.

Role UI & UX Designer
Tools Figma
AI Assistant Laptop Mockup

Project Context

Available without taking over

The assistant needed to stay available without taking over the product around it. I designed it as a flexible interaction layer: lightweight for quick access, but able to expand when the conversation or task needs more room.

Challenge

Change presence without changing the product

The challenge was not simply fitting the same chat interface into different screen sizes. The assistant needed to change how present it was without feeling like a different product.

  • Keep quick access lightweight and unobtrusive.
  • Preserve context as users move into deeper interaction.
  • Give each state a clear purpose.
  • Adapt the experience across web, mobile, and kiosk/totem use.

My Role

Designing Bubble, Overlay, and Fullscreen

I designed how the assistant moves from a small entry point into overlay and fullscreen work. The job was to keep those states connected as space grows, and to adapt the same presence logic across desktop, mobile, and kiosk.

Interaction Model

Interaction States

Bubble state preview

Bubble

Quick access without interrupting the current task.

  • Always available
  • Minimal screen space
  • Clear entry into conversation
Overlay state preview

Overlay

More room for conversation while preserving the underlying context.

  • Opens over the existing product
  • Supports richer interaction
  • Easy transition between work and chat
Fullscreen state preview

Fullscreen

A dedicated workspace when the assistant becomes the main task.

  • Supports longer conversations
  • More room for history and tools
  • Removes competition with the underlying interface

Interaction Flow

Users can enter through the Bubble, expand into an Overlay, and move into Fullscreen when the task needs more focus—without restarting the conversation.

AI assistant interaction flow from bubble to overlay to fullscreen

System Design

Platform Adaptation

The interaction model stays consistent across platforms, while the interface adapts to the available space, interaction method, and viewing distance.

Platform adaptation across web, mobile, and totem environments

Web

  • Lightweight Bubble entry
  • Overlay preserves the surrounding product
  • Fullscreen supports deeper work

Mobile

  • Fullscreen becomes more natural
  • Touch-first controls and simplified hierarchy
  • Conversation readability is prioritized

Kiosk / Totem

  • Clear separation between avatar and conversation
  • Larger touch targets and readable content
  • Designed for public use and greater viewing distance

Process

One model, three states

I first mapped what the assistant needed to support at each level of engagement, then designed the states around those needs. This created one interaction model for Bubble, Overlay, and Fullscreen rather than three separate interfaces.

Key UX Decisions

Designing for progressive presence

Progressive disclosure

Start with the smallest useful state and reveal more interface only when the task needs it.

Preserve context

Use the Overlay as a bridge between quick assistance and focused work instead of forcing an immediate context switch.

Give each state a job

Bubble, Overlay, and Fullscreen support different depths of interaction rather than acting as three sizes of the same window.

Adapt the system, not just the screen

Keep the core behavior recognizable while adapting layout and controls for desktop, mobile, and kiosk/totem use.

Final Design

One assistant, different levels of presence

Bubble, Overlay, and Fullscreen stay one assistant. Conversation and controls continue as presence grows from a small entry point into fuller work.

Design Outcome

Presence that scales with the task

Bubble → Overlay → Fullscreen lets the assistant change how much space it takes without becoming a different product. Quick access and deeper work stay related across desktop, mobile, and kiosk.

See something you’d challenge?

I like talking through design decisions. If you would have approached something differently, I’d like to hear it.

Next project

Explore another project

See how I approached a different product problem.