Electronic Arts · 2025 · AI-first · Marketing · Gaming
Apply AI-first design approach with speed and scalability
For the global reveal of Battlefield at Tokyo Game Show, I designed an event hub page that evolved throughout the exhibition — built for 5 languages within 4 weeks using an AI-first approach.
4 weeks
Delivered with AI-first approach
5 languages
English, SC, TC, JP, KO
Tokyo Game Show
Global reveal event
Design strategy
A living event hub that excites the fans
On the booth, there's always something happening. My goal was to create a hub that reflected this dynamic in the game show — real-time, interactive, and responsive to what was going on both on-site and online.
Design details
Animation and interaction that looks the same as the game
Make the website feel like an operation brief
Create exhibition map in 3D — feels like an operation space
Scroll-based animation that highlights the event booth
Snappy blinking animation for important sections
My learning
How I run the AI-first design approach
PRD → Design system in code → Component details. I started as a design architect before becoming an interaction designer — working top-down from foundations to the smallest detail.
PRD-first
A good project foundation saves time
I wrote the technical and functional foundation in the PRD before any LLM generation. Defining high-level components with global impact up front creates a rule set the model complies with — ensuring consistency across sessions.
Website environment (React & Tailwind)
i18n and language detection
Basic color & typography tokens
Content sections — venue, agenda, influencers
Countdown and live stream logic
Design system
Respect the branding guideline for 5 languages
I imported brand guidelines and the official website, converting them into color and typography tokens for Tailwind. Localisation wasn't in the existing guideline, so I built a design system page to preview typography and components across all languages.
Reusable components
Limit design scope to reduce token usage
I defined each component's variable and value props first, then elaborated on styling and interaction. This makes each prompt more efficient in time and token usage — reducing drift across LLM sessions.
Outcome
Shipped in 4 weeks, covered by major gaming media
The site launched on time for the Tokyo Game Show reveal and was picked up by major gaming media outlets — demonstrating that an AI-first design approach can deliver production-quality results at speed without sacrificing craft.