NN/G: the design process is compressed, not dead

Published on March 13, 2026, this article by Sarah Gibbons and Huei-Hsin Wang from Nielsen Norman Group addresses a narrative that has grown louder as AI tools accelerate design work: that formal design processes have become unnecessary.

The compression argument

The article’s main claim is that experienced designers who appear to skip process steps are not actually skipping them — they have internalized those steps through years of practice. The authors call this “compressed process.” What looks from the outside like intuition is structured thinking that has been done so many times it no longer requires explicit scaffolding.

AI tools change the timeline but not the underlying structure. A designer who once needed months to internalize how a certain navigation pattern breaks under edge cases can now encounter and test it in hours. This narrows the distance between novice and expert in the execution layer, but it also makes the framing step more visible. When AI handles generation, what remains to the designer is problem definition — the part that was always the harder work.

When shortcutting is safe and when it is not

The article draws a practical line between contexts. In mature problem spaces with established patterns, solution-first approaches can work: the pattern is known, the edge cases are understood, and an experienced designer can generate a reasonable output quickly. In novel, high-stakes, or regulated contexts — medical software, new market categories, legally complex products — skipping discovery is a liability. The appropriate process depends on the kind of problem, not on the designer’s speed preference.

The article names this skill “process literacy”: the ability to match the right level of rigor to the actual problem. It is the judgment call that AI cannot make on behalf of a designer, because it requires understanding the context of use rather than the form of the artifact.

Implications for prompting

The article implicitly applies to how designers brief AI tools. A prompt that goes straight to “design a settings screen for X” skips the problem framing that would constrain the output usefully. The designers who get more reliable AI outputs are those who do more definition work before opening the prompt box — the same habit that distinguished good designers before AI was part of the workflow.

Who should read this

Designers who are being told to skip process in the name of speed, and managers who need a vocabulary for explaining why some design work takes time despite access to faster tools. Also useful for anyone thinking about how AI changes the novice-to-expert curve in design: the gap is narrowing in execution, but it is not disappearing in judgment.