100+
Prompts and scenarios tested
← All work
VOICE / AI / AUTOMOTIVE
CARA Voice Assistant
Designed an intelligent in-vehicle voice assistant that helps drivers get things done safely across mobile and Android Auto.
How can I help you?
Navigation
Media
Phone
Overview
Voice-first
Low-touch interaction model
Pilot
Internal stakeholder demo
Problem / Challenge
Drivers need to complete common tasks quickly and safely while on the road. The UX challenge was to make assistant states, permissions, errors and feature entry points clear without increasing visual or cognitive load.
Constraints / Context
- Safety-first: minimize glance time and interaction steps
- Support mobile and Android Auto constraints
- Handle fallback, model error and server unavailable states
- Design for noisy, outdoor and in-vehicle contexts
Process
01
Discover
Research users, jobs and context.
02
Define
Prioritize opportunities and success criteria.
03
Design
Prototype flows, states and screens.
04
Validate
Review through demos, UAT and feedback.
05
Deliver
Prepare specs, patterns and handoff.
Key Design Decisions
Conversational by design
Natural prompts, assistant states and recovery paths reduce friction.
Driving-optimized feedback
Auditory-first feedback with restrained visuals keeps attention on the road.
Context-aware assistance
Suggestions adapt to location, saved places and user-defined variables.
Permission clarity
Calls, location and notifications are explained before action.
Design in Action
How can I help you?
Navigation
Media
Phone
Listening state
Calm, clear indicator with suggested entry points.
Active / running state
System feedback is visible without becoming distracting.
Confirm action
Before proceeding
Before proceeding
Cancel
Confirm
Permission & confirmation
Critical actions are confirmed before execution.
Outcome
- Validated core assistant flows across multiple model and transport configurations.
- Created reusable state patterns for listening, speaking, errors and suggestions.
- Improved clarity for demos, UAT and engineering handoff.
Learnings
- Complex UX needs clear states, constraints and prioritization.
- Scannable structure matters as much as visual polish.
- Reusable patterns improve handoff and future iteration.