AI didn't change what good design leadership is. It made it legible.

There’s a tell. Every design review eventually reaches it. The moment someone asks about feasibility, or constraints, or what happens when this idea meets the team that has to build it. Watch what happens to the room.

The work that was made for the deck flinches. It doesn’t have answers. It was never designed to. Big ideas find the rocks at implementation, and deck-ware was always built to live upstream of that conversation.

This predates AI by years.

For a long time, the most important design work had a visibility problem. The things that actually mattered (finding the real problem beneath the stated one, earning the trust to say the hard thing, making something that respects the person who has to use it) didn’t show up in a portfolio. They happened in rooms. In the implementation conversation. In the quiet decision to get it right even when the deck had already been approved and everyone wanted to move on.

The surface layer was easier to see: the polished artifact, the concept that photographed well, the deck that looked like thinking.

This created a strange economy in design. Visibility and value weren’t the same thing, but they were often treated as though they were.

AI raised the floor on the visible layer. Fast.

Producing the surface (the comp, the concept, the renderable idea) no longer requires the time it used to. What took a week can happen in an afternoon. The feasibility test hasn’t changed. The implementation questions still find the rocks. AI didn’t move the rocks. It made it possible to generate more decks, more quickly, with less friction. The gap between the surface layer and the production layer got wider, not narrower.

The industry’s response has been mostly defensive. The argument you hear is about authenticity. AI-assisted work called out as less valid than work made entirely by human hands, as though the process of making is what matters rather than whether the thing works.

The irony is worth sitting with. The authenticity defense protects the production layer, the layer AI can now replicate. It’s an argument for the value of the artifact, made at the exact moment the artifact became cheaper to produce. If the work doesn’t survive the implementation conversation, it doesn’t matter who made it or how.

The defense tells you which layer people built their reputations on.

I should say: I built mine on some of the same things. Years of directing rather than making. The artifact became something other people produced while I shaped the frame around it. The career progression is a gradual handoff of the making. The deck, eventually, becomes someone else’s job. Which is fine until the deck is the only evidence that work happened.

I’ve spent the last year making internal tooling. Systems that automate production design from design tokens all the way through to documentation structured for agents. Work that doesn’t render well in a portfolio presentation. Work that exists to be used.

The feedback loop is immediate and honest. Either the system works downstream or it doesn’t. Either the documentation holds when an agent reaches for it or it breaks. No deck. No approval. Just: does it do what it’s supposed to do.

The gap between visible work and valuable work is still there. AI made it wider. The question the industry hasn’t answered yet is whether it will also make it more honest.