Introduction

Lately, I’ve had the feeling that as designers, we’re drifting into overengineering when it comes to AI.

Most conversations revolve around terminals, scripts, Claude, MCP or Code Connect. Everything seems to point toward building pipelines, automating outside the tool, or layering complexity on top of existing workflows.

And I’m not saying that has no value. It does. But more and more, I see a pattern: a small friction appears, and the immediate response is to add a technical layer.

Meanwhile, we’re overlooking something quite obvious.

The blind spot

We still treat Figma files as inherently manual and hard to maintain.

Variables, naming, modes or collections are still perceived as tedious and repetitive work.

So when friction appears, the response is predictable:

• write scripts
• build pipelines
• connect APIs
• externalize logic

The issue isn’t doing this. The issue is doing it too early.

Because Figma is already integrating AI directly into the file to handle much of this.

What you can already do inside Figma

Today you can:

• select a frame and generate variables automatically
• create modes without structuring everything manually
• apply bulk changes
• reorganize systems with file-level context

And that changes something important:

Consistency stops being an external process and becomes part of the system itself.

The file is no longer just a visual artifact. It becomes a structured system.

The real shift

When the tool itself becomes smarter, what changes is not just the workflow.

It changes your role as a designer.

Before:

• consistency depended on manual effort

Now:

• it depends on how you structure the system

And that shifts the responsibility.

Where overengineering starts

It starts when we turn design problems into technical problems too early.

Instead of asking:

“How should this system be structured?”

We ask:

“How do I automate this?”

And that’s where complexity begins.

The risk

• unnecessary technical dependency
• added complexity
• loss of control
• fragile systems

Back to craft

The value isn’t in having the most sophisticated pipeline, but in building systems that work without needing one.

Closing

AI is a powerful tool.

But it doesn’t always mean adding layers.

Sometimes it means the opposite:

simplifying, structuring better, and using what you already have.

And maybe that’s the difference between using AI… and actually designing with it.