PROJECT 10 / AI DESIGN WORKFLOWS
AI-assisted design across the org
Leading AI-assisted design at Unity: guidance for the design org on Figma Make and design-system skills, packaged as reusable assets for org-wide distribution.
- Role
- Senior Product Designer
- Where
- When
- 2026
CONTEXT
AI arrived, and the org needed a direction
In 2026 AI-assisted tooling landed in the design toolchain faster than the org had a position on it. Every designer was improvising their own prompts, and the result was a mess of one-off workflows that did not share anything and did not scale.
The question was not whether to use the tools. It was how to make AI-assisted design a capability of the organisation rather than a personal habit, and how to keep it consistent with the design system the org had spent years building.
MY ROLE
From offsite direction to a reusable system
I authored and presented Figma Make and System Guidance at the 2026 Unity design offsite, setting the direction for how the design org uses AI-assisted tooling, and ran the accompanying UI workshop.
From there I built two reusable skills: a Unity Design System skill for AI-assisted Figma work, and a UI Builder skill that bridges Figma designs and UI Builder through MCP. Both are packaged for org-wide distribution from a central Git repository, and were featured in the design org's Design Monthly newsletter. Alongside that I prototyped Let's Make Games, a multi-agent game-creation flow driven through the Unity CLI with n8n orchestration, and supported the UI Toolkit event binding workflow.
KEY DECISIONS
The calls that shaped the work, and what each one cost.
ENCODE THE DESIGN SYSTEM INTO SKILLS
Tension
AI tools are only as good as the guidance they are given, and a designer prompting from scratch will drift from the design system every time.
Decision
Encoded the design system into reusable skills, so the system itself becomes the guidance the AI follows, rather than each designer re-explaining it.
Tradeoff
Building and maintaining skills is real work with no single deliverable to point to. It is also the difference between AI that amplifies the system and AI that quietly erodes it.
SHARE THROUGH A CENTRAL REPO, NOT PERSONAL FOLDERS
Tension
A skill that lives in one designer's pocket helps one designer. The org needed the skills to compound.
Decision
Drove distribution through a central org-wide Git repository, so a skill built once is available to every designer and can be improved by any of them.
Tradeoff
A shared repo needs governance and coordination, which is slower than everyone keeping their own copy. That shared infrastructure is exactly what makes the effort an org capability instead of a personal one.
OUTCOME
This work moves beyond individual output to org capability, which is the point of it. The design org now has a stated direction for AI-assisted work, and a mechanism, shared skills in a shared repo, for turning that direction into something every designer uses.
I frame this honestly: it is early, and the measure of success is whether the skills get picked up and improved by other designers, not whether I built them. That adoption is the thing to watch over the next cycle.