Challenge:
Creatively using AI to build a custom reference tool for the design team to improve project timelines and assist with scale
Design + content systems, content governance, information architecture
Business Need
For the games content design team, the biggest recurring drain was tracking down historical decisions and rationale scattered across Figma comments and Google Docs — with no central search tool that reliably indexed them. When Netflix's design org gave directives to discover new ways we could use AI to improve our workflows.
Proposed Outcome
Use AI tools to find the best way to build a scalable reference tool that made historical decisions and content rationale findable without manual searching — grounded entirely in work the product design team had already done.
Context
- Content and product designers could lose anywhere from 2 to 5 hours per week tracking down historical decisions and rationale
- Netflix had no central search tool that reliably indexed documentation or Figma files nor were memos and product specs in a reliable file location, making institutional knowledge effectively invisible over time
- The games content design space was complex and specialized — a general AI tool wouldn't know Netflix's specific patterns, taxonomy, or design decisions
- The initiative had no established playbook; guardrails, rules, and use cases had to be defined from scratch
Solution
Designed and built a custom AI skill using Claude and Cursor — a games-specific content design reference tool that surfaced existing product design work, supported brainstorming, and helped generate copy grounded in established patterns.
Built it by:
- Identifying the core problem: not that content didn't exist, but that it wasn't findable or usable at the speed modern content design required
- Feeding the skill existing documentation and Figma files — [X documents and Figma files — to be added] — so it could reference real games content design decisions rather than inventing new ones
- Setting strict guardrails: the skill would only surface content created by the product design team, would never invent copy or content unprompted, would always ask for clarification if a request was unclear, and would never suggest copy strings without being explicitly prompted
- Using the skill to consult on active projects — including brainstorming error messaging that fit within the existing message library, building matrices for upcoming features (covering key milestones based on similar features from widely known apps), generating additional engineering asks for fallback copy when dynamic elements couldn't load on new features, and locating source documents and original rationale without manual searching
- Making the skill self-serve and available to the broader team.
Outcome
- Once implemented, it was downloaded by content designers, product designers, and made available to product marketing to help frame branding and content alignment decisions
- Recovered an estimated 2-5 hours per week per content designer and product designer previously lost to manual research
Used actively on multiple shipped projects before and after launch, including achievement systems, error messaging, and game invite copy - Skill was fully built, in active use, and being refined at time of handoff — [X documents and Figma files fed into the skill — to be added]
What I'd Revisit
The skill was built and adopted quickly, but leaving before the full rollout meant I wasn't able to measure its longer-term impact on team velocity or content consistency. If revisiting, I'd establish baseline metrics upfront — time spent on research per project, number of escalations for historical context, copy iteration cycles — so the tool's impact could be tracked and used to make the case for expanding it beyond games content design to other product areas.
