AI Won’t Replace Design Systems. It Will Replace the Tedious Parts.
For years, building a design system meant spending a surprising amount of time doing things that weren’t particularly creative.
Create a button.
Duplicate it.
Create another size.
Add hover, pressed and disabled states.
Set up properties.
Rename layers.
Check spacing.
Repeat.
None of this is meaningless. Good systems depend on this structure.
But a lot of it is mechanical work.
And that is exactly where AI is starting to become interesting.
Recently, I’ve been exploring how AI-assisted tools like Figma Agent can fit into real design workflows.
What surprised me wasn’t that AI could “design.”
It was how quickly it could remove the repetitive setup between an idea and a usable system.
From hours of setup to a starting point
Imagine I need a complete button system.
Traditionally, I might create the base component, define sizes, build variants, configure states, organize properties and make sure everything follows a consistent structure.
Multiply that across inputs, cards, navigation, modals and other components, and a significant amount of time disappears into production.
With a detailed enough instruction, AI can now generate much of that initial structure remarkably quickly.
That changes something important.
Not the design thinking.
The setup.
Instead of beginning with an empty canvas and manually constructing every variation, I can begin with something structured and spend more time evaluating it.
That distinction matters.
Generation is not the same as designing a system
AI can give me components.
But components alone are not a design system.
I still need to decide:
Which tokens should exist? How should states behave? Is the component accessible? What should be flexible and what should remain locked? Will this structure survive when the product grows? Are naming and properties understandable to another designer? Does the system create consistency without limiting useful variation?
These decisions require context.
A component might look perfectly correct and still be badly designed as part of a larger system.
That is why I don’t see this as:
Designer vs. AI.
I see it as a shift in where designers spend their time.
The mechanical layer is becoming automated
Design tools have automated parts of our workflow for years.
Auto Layout removed repetitive positioning.
Components reduced duplication.
Variables made tokens more manageable.
AI feels like another step in that progression—just significantly larger.
It can potentially take over more of the repetitive production layer:
Generating initial components.
Creating variants.
Exploring states.
Producing repetitive patterns.
Creating starting structures.
Helping prototype interactions.
The interesting question isn’t whether AI can perform these tasks.
It increasingly can.
The better question is:
What should designers do with the time this gives back?
The value moves toward decisions
If I spend less time manually producing twenty component variants, I can spend more time thinking about whether we actually need twenty variants.
That is where I think the real opportunity is.
Less time on mechanical production.
More time on:
Architecture. How does the system fit together?
Consistency. Are we solving the same problem in five different ways?
Accessibility. Does the system work for more than the ideal user and ideal state?
Scalability. Will these decisions still make sense when one product becomes five?
User experience. Are we creating a better experience—or simply a more organized Figma file?
AI can accelerate execution.
But acceleration makes good judgment more important, not less.
Faster production can also create more mess
There is another side to this.
When creating components becomes almost effortless, it becomes equally effortless to create too many.
More variants.
More patterns.
More options.
More inconsistency disguised as productivity.
AI can generate ten solutions before we have properly understood whether we need one.
So restraint becomes part of the designer’s job.
The goal should not be:
How much can AI generate for me?
It should be:
What is actually worth generating?
A good design system is not valuable because it contains hundreds of components.
It is valuable because it gives a team just enough structure to make good decisions consistently.
The designer becomes more of a system architect
I think this is part of a broader shift happening across product design.
The repetitive layer is becoming cheaper.
The value is moving upward.
From creating components to defining systems.
From drawing states to defining behavior.
From maintaining files to maintaining logic.
From producing variations to deciding which variations should exist.
That makes the designer’s role less about operating the tool and more about directing the system.
AI becomes the production partner.
The designer remains responsible for judgment.
Where I think this is heading
I don’t think we are moving toward a future where designers simply type:
“Create my design system.”
and receive a perfect production-ready system.
Real products are too contextual for that.
But I can imagine a workflow where designers define the principles, constraints, tokens and behavior—and AI handles much more of the repetitive construction underneath.
That would be a very different use of a designer’s time.
And potentially a much better one.
The part I’m most interested in exploring now is where tools like Figma Agent eventually sit in that workflow.
Are they primarily generators?
Prototyping assistants?
System maintainers?
Or eventually something closer to a production partner that continuously helps designers build, audit and evolve a system?
We’re still figuring that out.
But one thing already feels clear:
AI doesn’t need to replace the designer to fundamentally change design work.
It only needs to remove enough of the tedious work that we can spend more time on the decisions that actually matter.