Design Systems: Then, Now, Next
Why systems built for human consistency won’t survive an AI-first workflow
Nov 23, 2025
•
Robert Myrsäter
Design systems were built for a world where humans made every pixel by hand. Their purpose was simple: keep teams consistent, reduce chaos, and make quality predictable.
That world is disappearing.
And systems designed around human behavior don’t translate to machine behavior.
Here’s the shift.
Then: Systems That Kept Humans Consistent
For years, design systems existed to counter the natural inconsistency of humans. People forget spacing rules, recreate components, and drift off-pattern. Systems brought order by defining shared patterns and reusable parts.
The source of truth lived in Figma.
Not perfect, not always aligned with production, but it gave teams something concrete to point to. Designers needed visual reference, and Figma served as the collective memory of the organization.
Now: Designers Create by Directing, Not Drawing
Design is moving from execution to orchestration. Designers now guide models, edit outputs, and shape direction more than they assemble interfaces. AI can generate UI faster than humans can document it. Once teams stop producing artifacts manually, the file no longer holds authority.
It becomes an intention surface—a place to think, sketch, and communicate.
Code becomes the accurate record of what ships. Yet even code only expresses the system; it doesn’t define it.
The source of truth now lives in the system itself—in the tokens, constraints, semantics, and rules that shape output.
Models don’t learn from screens.
They learn from logic.
Mocks illustrate intention, but only encoded constraints and behaviors guarantee consistent generation. The truth shifts from artifacts to the underlying structure that governs them.
Why Traditional Systems Break
Legacy systems rely on human judgment. Designers can interpret gaps with intent. Models can’t—they generate around them, and the results fragment quickly.
AI requires explicit structure: constraints, semantics, pattern logic, and defined ranges.
Systems built for humans assume intuition.
Systems built for machines require precision.
Next: Systems Become Generative Grammars
To produce consistent results from non-deterministic models, design systems must evolve into generative grammars: encoded behaviors, constraints, and pattern rules that both humans and models can operate within.
In practice, this looks less like documenting components and more like defining the rules that govern how they can be generated.
At that point, the source of truth isn’t a file anymore—it’s semantics the model can actually read.
The system stops being a library and becomes an engine.
The New Role of Designers
Designers shift from making artifacts to defining how artifacts are produced. They become architects of constraints, authors of visual logic, and calibrators of model behavior.
Taste moves upstream—into rules, semantics, and constraints.
Quality becomes systemic rather than manual.
The Shift
Design systems once scaled human teams.
Now they must shape machine behavior.
Teams that stay artifact-centric will struggle with AI-generated inconsistency. Teams that build systems as languages will define how future products are made.
The design system is no longer how we document design—it’s a logic.
It’s how we teach machines to create.
🌻