AI-Integrated Product Design Workflow
Cushly: Tax App for Freelancers — Designed & Shipped in 10 Days with Claude
6
Phases
15
Prompts
10 days
Research to Production + Shipped

Role
Designer & Builder
Project
Cushly - Personal
Platform
iOS - SwiftUI
Date
May 2026
OVERVIEW
A full product lifecycle, compressed into days.
Traditional product design is slow — weeks of research, stakeholder rounds, handoffs, and delayed engineering.
This project flowed seamlessly: research informed IA, IA shaped the design system, and the design system became production code. No handoffs. No delays.
The result: a complete 6-phase workflow where I directed and Claude accelerated — delivering a full production iOS tax app, built by one person.
THE CHALLENGE
No tool built for how Australian freelancers actually manage tax
Freelancers juggling GST, BAS, and irregular income need something mobile-first, accurate, and reassuring — not another accounting dashboard. Existing tools like Rounded, Xero, and Wave miss the gap: a financial wellbeing product built around Australian tax reality and calm, proactive design.
WHAT I DID
What I owned
The full product lifecycle as designer and builder — research, IA, design system, prototyping, user testing, SwiftUI production code, QA, and App Store submission. Claude accelerated execution; I directed scope, validated against ATO requirements, and reviewed every output.
What I delivered
Research, JTBD pain points, and competitive analysis (Rounded, Xero, Wave)
5-tab IA and 3 end-to-end user flows (expense, BAS, onboarding)
Interactive prototype, moderated user testing, and iteration
Production SwiftUI app — MVVM, SwiftData, ATO tax engine
QA accessibility audit and shipped to the App Store (Australia)
IMPACT
What one person shipped in ~10 days
10
Days · research to App Store
6
Phases · zero handoff loss
0
Third-party dependencies
KEY INSIGHT
The opportunity wasn’t a better accounting tool — it was the first financial wellbeing tool built around the psychological and practical reality of Australian freelance life. AI accelerated production; every distinctive product decision still came from the designer.

HOW
Process in brief
Research & discovery — pain points grounded in Australian tax law, competitor gaps identified
IA & flows — 5-tab model and critical flows mapped before any screen design
Design system & prototype — tokens to Figma MCP, then Claude Design for clickable testing
Build, QA & ship — Claude Code in Xcode, accessibility audit, App Store launch
TAKEAWAY
Prompting is a design skill — review is not optional
Output quality tracked directly with spec quality. Domain knowledge and critical review mattered as much as speed — the QA audit surfaced real accessibility and performance issues in AI-generated code. The designer’s role shifted from producing documents to directing, deciding, and validating.

