As LLMs and MCPs enter the product design world, a systems gap surfaces: Design systems today are built for humans, not machines.
This talk explores design systems as machine-readable infrastructure that both humans and AI can understand, maintain, and build upon. I'll share findings from an 8-configuration MCP benchmark that tested the impact of metadata format on LLM reasoning, as well as the most effective metadata architecture that emerged from it.
Design system designers, design technologists, frontend engineers, and product teams working towards AI-driven design systems.
she/her
Sr. Design System Designer
Indeed
Diana Wolosin is a Senior Design Systems Designer at Indeed with 10 years of experience building systems that scale. Currently focused on building machine-readable design systems as AI infrastructure, with emphasis on evaluation frameworks and context governance. She leads Friends of Figma San Diego, and writes about AI-ready design systems on Medium. A proud Colombian, dog mom to Whiskey and Havannah, and happiest outdoors or hunting for good food.
Get access to all conference sessions, files, Q&As and networking.
Full access for one attendee
1 login per ticket
Prices in US Dollars
Live access to the conference for up to 3 team members
Prices in US Dollars
Waiting for budget approval? Get a 14-day price lock → Email us
Bonus
120 minutes total of on-demand AI sessions (multiple sessions combined) pre-recorded during our events. Available instantly to all conference ticket holders. Access immediately after purchase.
Spotify MCP
How their Design System team built their MCP
Cursor Prompting
Best practices so AI generates the design you expect
Context-based DS
How to work with helpful plugins and workflows