Case study / 05 — Mobile App · AI

AI SPEAK voice chat bot.

Intuitive interfaces for an AI-powered speaking and language app built around voice recognition and conversational practice.

Role
UI/UX Designer
Platforms
Mobile
Category
Mobile App · AI
Type
AI Language App
0
Platforms
0
Process phases
0
Key decisions
0
Research insights

OVERVIEWThe idea

An AI speaking companion that helps users practice a language by talking. The design challenge: make voice-first interaction feel natural for non-tech-savvy users.

THE PROBLEMWhat had to be solved

Voice UIs fail when users don’t know when to speak, what to say, or whether the app heard them. Onboarding and feedback states carry the entire experience.

MY ROLEWhat I owned

Designed onboarding, conversation and feedback screens; contributed to AI prompt structuring and voice-feedback UX patterns.

DESIGN PROCESSFrom idea to interface

01 Voice-interaction research
02 Onboarding flow
03 Conversation UI patterns
04 Feedback & correction states
05 High-fidelity UI
06 Prototype

METHODThe workflow, end to end

01
Step 01
Voice-interaction research
02
Step 02
Onboarding flow
03
Step 03
Conversation UI patterns
04
Step 04
Feedback & correction states
05
Step 05
High-fidelity UI
06
Step 06
Prototype

KEY DECISIONSWhat made it work

🎙️

Clear speaking states

Listening, thinking and responding are visually unmistakable.

🧸

Friendly onboarding

Three screens to first conversation — no forms, no lectures.

Gentle corrections

Mistakes surfaced as encouragement, not red error text.

📈

Visible progress

Streaks and fluency scores that motivate daily practice.

RESEARCH & INSIGHTSWhat shaped the direction

Mobile-first discovery: where is this app used, one-handed or two, in what moments of the day? Session mapping and competitor teardowns showed that onboarding and first-minute experience decide retention long before feature depth matters.

The first minute decides

If users don’t reach value in the first session, there is no second session. Onboarding was designed as the product, not a tour of it.

👍

Thumbs set the layout

Primary actions live in the natural thumb arc; anything above the fold’s reach is for reading, not tapping.

🔔

Attention is borrowed

Every notification and interruption spends trust. The design asks for attention only when it returns value.

WHO IT'S FORDesigning for real people

PRIMARYThe curious newcomer

Downloaded on a recommendation, gives the app ninety seconds to prove itself. Needs instant value, zero forms and a reason to return tomorrow.

SECONDARYThe habitual user

Opens the app daily and wants friction gone: remembered context, one-tap resumption and progress made visible.

DESIGN LANGUAGEThe visual system

A friendly, high-contrast mobile language: large touch targets, one primary action per screen, motion used to explain rather than decorate. The palette stays calm so content and progress cues stand out, and dark mode is a first-class citizen rather than an inverted afterthought.

One primary action per screen
Thumb-arc layout
Explanatory motion
First-class dark mode
Progress made visible

TESTING & ITERATIONPressure-testing the design

Prototype tests focused on the first-run experience: where do new users hesitate, misread or bail? Each round shortened onboarding and moved value earlier, while returning-user flows were tuned for one-handed speed.

PROJECT SCREENSInside the design

AI Speak — home & lessons AI Speak — voice conversation AI Speak — lesson prompt

Auto-scrolling — hover any screen to pause & lift

OUTCOMEThe result

A voice experience that feels like talking to a patient tutor — full flows on Behance.

View full project on Behance

KEY LEARNINGSWhat I'd carry forward

01 Onboarding is the product’s first feature, not its manual.
02 Design the returning-user path as carefully as the first-run path.
03 On mobile, restraint reads as quality.
← Previous
SaaS Subscription Module