Faster releases without a sloppier product

Getting started

Wing Bank’s mobile app is one of Cambodia’s most-used financial products—transfers, wallets, QR pay, and onboarding for a growing customer base. Over four years as UX/UI Lead, my job was to keep the experience coherent while the roadmap accelerated. This case study covers how we refreshed high-traffic flows and changed how design worked with product and engineering.

The Challenge

Wing Bank’s mobile app was where Cambodia banked on the phone: transfers, wallets, QR pay, onboarding. Feature demand outpaced design capacity. Every release added screens; patterns drifted; product managers arrived with folders of feedback—app store reviews, call-center notes, chat logs—not a single prioritized problem.

The business needed velocity without a sloppy UI. My challenge as UX/UI lead was not to draw more screens. It was to compress the loop from “we heard 500 complaints” to “here is one testable flow” while keeping engineering unblocked.

[!NOTE] Insert 16:9 image: collage of Wing app screens showing pattern drift (mixed button styles, inconsistent success states) before the refresh initiative.

The Breakthrough

We did not bolt AI onto the customer product first. We used AI inside the design operation—where the bottleneck actually was.

1. Feedback synthesis at scale

I built a lightweight pipeline (export → structured prompt → human edit) to cluster thousands of qualitative comments into ranked themes: onboarding OTP failures, transfer limit confusion, QR receipt anxiety. What used to take two days of workshop sticky notes became a same-day brief product could react to.

[!NOTE] Insert 4:3 image: anonymized screenshot of a synthesis output—theme clusters with volume counts and example quotes (blur PII).

2. GenAI for wireframe iteration—not final UI

For high-churn flows (wallet top-up, transfer confirmation), we used generative tools to explore layout density and step count in hours, not days. I treated outputs as disposable scaffolding: wrong brand, wrong components—but useful for “What if we collapsed these three steps?”

My rule: AI proposes structure; designers own system, accessibility, and brand. Every promoted concept was rebuilt in Figma with Wing’s tokens and reviewed with engineering for edge cases.

[!NOTE] Insert 16:9 image: three low-fi wireframe variants for transfer confirmation—annotated which variant won usability testing and why.

3. Agile UI that survived release cadence

The breakthrough release was a transfer + wallet refresh: one navigation model, one form pattern, one success/error language. We shipped behind a feature flag, measured drop-off per step, and iterated weekly—design in the same sprint rhythm as engineering.

The moment that proved the model

A product lead asked for “just add a button” on the home screen. The synthesis brief showed 68% of related complaints were about finding transaction history after QR pay—not home-screen discovery. We redesigned receipt + history as one path. QR repeat usage in the test cohort rose 23% in four weeks; home-screen clutter did not.

[!NOTE] Insert 3:2 image: final transfer receipt screen with prominent “View history” and share actions—highlight the path users actually needed.

The Outcome

Over four years on the product, the workflow change mattered as much as any single screen:

Area	Before → After
Priority flow concept → dev-ready handoff	~3 weeks → ~10 days
Cross-journey UI consistency (internal audit)	baseline → +40% on top 5 flows
Usability issues found post-release (tracked)	−30% on refreshed transfer/wallet paths
Designers on shared pattern library	2 → 5 contributors

Customer-facing AI features came later (smart search suggestions, smarter transaction labels). Because we had already proven a fast, evidence-led pipeline, those features shipped on patterns the team trusted—not as one-off experiments.

[!NOTE] Insert 16:9 image: Wing design system in Figma—form fields, buttons, and QR success states used across squads.

What this case study shows: growth-stage consumer mobile, agile UI at release speed, and using AI in the workflow to synthesize feedback and accelerate exploration—while keeping craft and handoff quality with engineering.