Design governance in the age of generative AI

Generative AI tools are making designers faster. They are also making design systems decay faster. A tool that can generate a plausible-looking component in seconds doesn't know that your team chose 12px border radius for a reason, or that the secondary button uses a glass treatment because someone tested it against the terracotta variant and measured the click-through rate.

The output looks right. It just isn't right. And the gap between "looks right" and "is right" is where design systems die.

The decay pattern

Design systems don't fail dramatically. They fail incrementally. A new designer joins the team and copies a component from the wrong screen. A developer builds a variant because the documented one doesn't match what's in Figma. Someone uses an AI tool to generate a layout and the generated tokens don't match the system's tokens. Each deviation is small. Over six months, the system is unrecognizable.

The traditional fix is more documentation. More guidelines. More Figma page rules. More Confluence pages nobody reads. This doesn't work because the problem isn't that people don't know the rules. The problem is that the rules aren't present at the moment the decision is being made.

What governance actually needs

Design governance needs three capabilities that documentation can't provide.

First, detection: the ability to scan a design file and identify where it deviates from the system. Not just "this color is wrong" but "this color was changed from the canonical value on March 3rd by someone who probably didn't know the original was chosen for WCAG compliance."

Second, context: when a deviation is detected, the system should surface the original decision, why it was made, who made it, and what the correct version looks like. This turns a violation into a learning moment instead of a friction point.

Third, memory: every design decision should be captured with its rationale so that future designers inherit context, not just components. When someone asks "why is this 12px and not 8px?" the system should have an answer that doesn't depend on the person who made the decision still being on the team.

Why we are building Tenet

We've spent 18 years building and maintaining design systems for enterprise clients. We know firsthand how quickly they decay. The systems we built for Deem lasted because we stayed for 7 years and carried the context in our heads. That doesn't scale. It doesn't survive team turnover. And it certainly doesn't survive generative AI producing design variations faster than any human can review them.

Tenet is the tool we wished we had on every engagement. It scans, fixes, and remembers. It starts inside Figma, where the decisions happen. And it's built specifically for the world where generative tools create faster than humans can govern.

If you're managing a design system and noticing the drift accelerating, that's the signal. The system isn't failing because your team doesn't care. It's failing because the velocity of creation has outpaced the velocity of governance. That gap is what Tenet closes.