AI in a creative team
Who decides when agents build the tools
For a client group of 7–9 brands I came up with, designed and built a toolchain myself, working with AI agents: 12 command-line tools. A person reviews every drift between Figma and the canon and decides what to do with it. In Gromatik, my design system platform, I measured that 5 repetitions of the same prompt gave 5 byte-wise different results. I also designed an "agent in the product" route and suspended it on the same day, before implementation.
- [1] The starting point was 7–9 brands from acquisitions, ranging from a PDF brand book to full component libraries in Figma. A design system for an energy group with many brands
- [2] I came up with, designed and built the toolchain myself, working with AI agents. A design system for an energy group with many brands
- [3] The toolchain is a monorepo of 12 command-line tools with 78 test files, built in June 2026. A design system for an energy group with many brands
- [4] A person reviews every drift between Figma and the canon and decides what to do with it. A design system for an energy group with many brands
- [5] A measurement on the live stack showed that 5 repetitions of the same prompt gave 5 byte-wise different results. Gromatik: from tools for a client to a product of my own
- [6] I designed the "agent in the product" route and suspended it on the same day, before implementation. Gromatik: from tools for a client to a product of my own
- [7] Every run records its settings and a fingerprint of its input, so you can trace where a result came from. Gromatik: from tools for a client to a product of my own
Your turn · not scored
Evidence for 2.2