Upstream Influence Isn’t Unlocked by Tool Access

In 2024 I held a belief that I now think was wrong, and I want to be specific about how it was wrong because the error is common.

The belief: as AI tools improve, designers will naturally move upstream. Better tools mean more leverage. More leverage means more influence over product direction. The capability growth would expand the design function automatically.

What I observed over the following year: engineering adopted AI tools faster, more deeply, and with more organizational recognition than design did. Not because engineers are smarter or more adaptive. Because the tools mapped directly onto their existing workflow abstractions. GitHub Copilot sits inside the IDE, produces code, gets evaluated by whether it runs. The feedback loop is tight and the output is legible to the organization.

Design tools produce artifacts whose value is harder to attribute and harder to evaluate. When a designer uses AI to explore more directions faster, the organization often doesn’t see the exploration — it sees one final output, the same as before, slightly faster. The leverage is invisible.

More importantly: I watched leadership at multiple organizations interpret AI adoption in design as productivity tooling. Faster comps. More iterations in less time. The same work, cheaper. Not strategic leverage. Not upstream influence. Faster execution of the same downstream role.

This wasn’t tool failure. It was a framing failure, and it was a failure I contributed to by building tools and frameworks before building organizational understanding of what the tools were for.

The revised belief: upstream influence for design in an AI context isn’t unlocked by tool access. It’s unlocked by shaping the organizational understanding of what AI capability means and where it requires design judgment to be usable. That’s a different job. It requires working at the level of mental models and decision-making processes, not workflows and artifacts.

That realization is what shifted my focus toward capability surface mapping, AI literacy infrastructure, and calibration as an organizational discipline. Not because those are interesting intellectual frames, but because they’re the actual leverage points that tool adoption wasn’t providing.