Skip to content
08 · Field Guide

Craft by discipline

Everyone can prompt. The value is bringing different forms of expertise into one build.

Product design

AI tends to create screens before it understands a journey. Product designers should slow that impulse down.

Provide:

  • the job the user is trying to accomplish;
  • entry and exit points;
  • decision points and dependencies;
  • priority of actions;
  • state transitions;
  • what should be remembered;
  • failure and recovery behaviour;
  • acceptance criteria.

Ask the agent to describe the flow in plain language before building it. Review whether each interaction has a clear system response and whether the simplest path remains visible.

Visual design

AI tends to default to familiar visual clichés. Visual designers should give it a system, not a mood-board adjective cloud.

Provide:

  • typographic roles and scale;
  • layout grid and max widths;
  • spacing rhythm;
  • semantic colour roles and contrast expectations;
  • component geometry;
  • icon style;
  • imagery or illustration principles;
  • motion purpose and tempo;
  • explicit anti-references.

Convert repeated visual values into design tokens. Review the rendered page, not only the code. Check multiple widths, long content, focus styles, zoom, and real imagery.

Content design

AI can produce fluent but generic copy. Content designers should define what each message must achieve.

Provide:

  • user intent and emotional context;
  • message hierarchy;
  • approved terminology;
  • voice characteristics with examples;
  • reading-level expectations;
  • patterns for labels, instructions, errors, and confirmations;
  • words or claims to avoid;
  • localisation considerations.

Review content in the live interface. A sentence that works in a document may fail inside a narrow card or at the moment of error. Never let placeholder copy become accidental final copy.

User research

AI can accelerate preparation and synthesis, but it can also flatten nuance or invent certainty.

Use it to help:

  • turn product assumptions into research questions;
  • draft neutral discussion guides;
  • generate edge cases for a study plan;
  • structure notes;
  • cluster de-identified observations;
  • identify contradictions and missing evidence;
  • draft a prototype specifically for a research question.

Researchers should keep an evidence trail: observation → interpretation → insight → opportunity. Ask the AI to cite which observations support each synthesis statement, then review the original material yourself.

Working as one team

The disciplines should not hand prompts to one another in a linear waterfall. Work around shared artefacts:

  • a one-page product context;
  • a journey and state map;
  • an evidence and assumptions table;
  • design and content principles;
  • the running prototype;
  • an issue list ordered by user impact;
  • a shared definition of done.

The prototype becomes a conversation object. Researchers sharpen what must be learned, product designers shape behaviour, content designers shape understanding, and visual designers shape perception and hierarchy.