Design systems with AI: what changes and what doesn't
AI speeds up interface generation, but a design system still needs human judgement. How I integrate AI tools into my workflow without losing coherence.
The role of AI in design systems
AI has changed how I generate interfaces. I can iterate faster, explore more variants and reduce the time between idea and working prototype. But one thing hasn’t changed: a design system needs judgement.
What AI does well
- Generate component variants from a description
- Suggest colour palettes coherent with existing tokens
- Speed up repetitive code (grids, layouts, forms)
- Spot inconsistencies in spacing or typography
What still needs a person
A design system is not a collection of components — it’s a set of decisions. Decisions about hierarchy, about what gets visual priority, about how it scales when the product grows.
AI doesn’t know why you chose 8px spacing instead of 12px. It doesn’t understand that secondary button exists because the primary competes with another CTA in the same view. Those decisions come from business context, from the user and from the conversation with the team.
My current workflow
- Define tokens and rules — I do this manually. Tokens are the foundation of everything.
- Generate base components with AI — I use tools like v0 or Claude to get a first draft.
- Review and refine — This is where judgement comes in: consistency, accessibility, performance.
- Document decisions — Not just the what, but the why.
AI is a velocity multiplier, not a substitute for judgement. Use it to move faster, not to think less.
Conclusion
If you’re building a product and want to use AI in your design workflow, start by having clear principles. AI amplifies what you already have — if you have good judgement, you’ll move faster. If you don’t, you’ll move faster in the wrong direction.