Medium · Workflow
Study build → moderate → report
AI helps build and moderate studies; humans still own the research question and the call.
Purpose
Maze’s AI hub (study builder, moderator, reporting, MCP) scales research ops; NN/g’s testing-AI method reminds us evaluation design is still human. Medium practice: AI drafts the study and drafts themes; the researcher locks the hypothesis, sample, and go/no-go.
Steps
1. **Lock the research question and success metric** — before opening Study Builder.
2. **Draft with AI** — generate tasks/screener; edit every item for bias and clarity.
3. **Pilot one session** — human or AI-moderate; check the quality metrics / transcript sanity.
4. **Run the n** — stop rules agreed up front; do not “keep collecting because AI is cheap.”
5. **Report with themes + clips** — accept AI themes only after a human pass; publish a one-page decision for design.
When to use / Skip
Use for unmoderated or lightly moderated product tests at medium scale. Skip for generative discovery that needs deep moderated interviews only — AI moderate is not a substitute for expert interviewing. Not for legal/medical claims without compliance review.