Design systems have always been about finding the right balance between rules and freedom, structure and speed.
AI is now forcing that same negotiation, just in a different layer: not “how much should teams deviate from the system,” but “how much should we hand over to a machine that can generate, document, and even design faster than we ever could.”
The honest answer, based on hundreds of practitioner responses and countless hours of deep-dive conversation is: it depends on what you’re trying to do. AI is remarkably good at the messy, expansive, diverging work, such as synthesizing legacy code, auditing inconsistencies, drafting first-pass documentation, generating dozens of UI variations in minutes. It’s far less reliable at the converging work: the precise, high-stakes decisions where every prop, every edge case, every piece of context has to be exactly right. That’s still where humans earn their keep.
None of this is settled. It’s actively being worked out in real time by the people building these systems today. Here are three conversations from Ben Callahan’s “The Question” that are worth your time if you want to see how:
Design System AI Pulse Check, with Kevin Coyle — Why practitioners are “cautiously optimistic,” and why outsourcing documentation to AI means losing the alignment that writing it was supposed to create in the first place.
Design Systems as AI Context, with TJ Pitre — A look at the messy reality of “source of truth” in the AI era, and why treating your design system as structured context — not just a component library — is the real unlock.
Where AI is Failing Design Systems, with Nathan Curtis — Nathan’s framework for where AI shines (diverging: ideation, synthesis) versus where it still falls short (converging: precise component modeling, edge cases, filling in every blank correctly).
