I built an AI workspace for our design team
I built the shared repository, moved our existing work into it, and tested the setup.
AI work needed
a shared workspace
We needed to ship faster with AI. But separate setups made it hard to track tasks, review work, and build together
I planned for a system
that could grow
I chose one repository for web, iOS, and Android so agents could adapt features across platforms. Shared skills and tasks would help it grow with the team and better models.
One workspace, shared
context
Our design system, templates, tasks,
and agent instructions now live together. Everyone builds from the same context.
We started building
the workflow
I set up the folders and agent instructions.
A colleague and I started moving and testing templates for all three platforms.
Next: turn meeting transcripts into tasks for agents to build. Our goal is a design factory.
More product decisions, less manual work
I helped the team spend less time drawing screens.
We describe ideas by voice, make product decisions,
and review what agents build.
The design factory is still in progress. My lesson: build
for better models without starting over.
Local AI tool for YouTube content research
View case study
