VOX is an AI assistant that turns what you say into organized actions: meetings, calls, voice notes and passing thoughts. You tap Listen and talk. The app records, transcribes, summarizes and suggests the next tasks, which you can edit before saving. It started as a rapid concept for investors and, in 2026, I redesigned it for iOS 27 with SF Pro and native Apple patterns.

Role: Product Designer (UX/UI), leading the concept and the prototype

Platform: iOS 27 (redesign of the original investor concept)

Scope: product concept, user flows, UI, quick branding and a navigable prototype

Tools: Figma and Claude

Conversations are full of decisions and to-dos that get lost once the call ends. Taking notes means not really listening, and listening means forgetting. VOX started as a rapid prototype to show investors one idea:

The design had to make a whole AI pipeline (recording, transcription, summary and task suggestions) feel like a single, simple gesture.

Sign in with Apple, Google or email, then land on a home screen built around one action: Listen. Below it sit the open tasks and the recent conversations, each with its date and length.

While recording, the user can open the live transcription, highlight a key moment with one tap or add their own notes. Speakers are identified by name, and highlights are tagged by topic.

When the conversation ends, VOX writes a short AI summary and suggests next-step tasks. Every suggestion can be edited before it goes to the list, so the user stays in control of what the AI proposes.

Each task keeps a link to the moment it was born: the quote from the conversation, the speaker and the session. It also holds the due date, reminder, list and subtasks.

A quick identity for the pitch: a soft, rounded wordmark, a dark palette with magenta accents, and a type scale based on SF Pro to follow iOS conventions.

I built a functional prototype with Claude, straight from my Figma screens. Recording, live transcript, highlights, AI summary and tasks all work, and the animations and interactions are real, not slides.

I led the concept from brainstorming and quick wireframes to high-fidelity UI and the interactive prototype. AI was part of the process: I used Claude to explore flows, generate interface variations and build the prototype in days, not weeks. AI widened the options; my judgment chose, simplified and polished what works for people, like a cleaner recording screen with live listening waves, transcripts only on demand, and clear summaries and tasks.