Skip to content
Article Nielsen Norman Group Sep 2026

Nielsen Norman Group: AI in the editorial workflow

Nielsen Norman Group published a piece in September 2026 documenting how their team actually uses AI in the article production process—not as a policy statement, but as a description of what editors do on each draft. The author, Raluca Budiu, has edited NN/G articles since 2013 and walks through four specific applications that have become part of their standard workflow.

The context matters because NN/G is a firm where editorial standards are part of the product. Every article goes through at least two editors who review for logical coherence, UX accuracy, and copy quality. That rigorous multi-revision process is not something AI replaces—it’s the frame within which AI is given specific, bounded tasks.

The four applications Budiu describes are: clarity (using tools like Copilot and Grammarly to condense text without losing meaning, with human verification of each change), format compliance (AI generates summaries that fit exact character limits or platform specifications), content adaptation (reformatting existing article content for courses, reports, videos, and podcasts), and argument critique (flagging potential fact errors and logical inconsistencies for human review and resolution). In each case, the scope is narrow and the output is verified.

The central claim of the article is that editorial responsibility cannot be delegated. Humans decide which ideas are worth publishing, and humans ultimately stand behind what goes out under the NN/G name. AI can accelerate or assist any step in that process, but it cannot replace the editorial judgment that gives the final work its credibility.

This is a useful model for design practitioners who write—whether that means documenting patterns, publishing research reports, contributing to publications, or maintaining a team design blog. The instinct to use AI for full draft generation and then ship it is common, but the NN/G framework shows a different posture: assign AI to discrete tasks inside a human-managed workflow rather than treating it as a replacement for the process itself.

The article is short enough to read in full and specific enough to adapt directly. For teams that produce a lot of written output—design rationale, case studies, research summaries—it offers a principled starting point for deciding which parts of your writing workflow can absorb AI assistance and which ones still require a person.