Prompt Engineering for Designers: From Brief to Pixel
Prompts are the new design brief. Here is the structure that consistently turns a vague ask into usable output.
Most bad AI output is a bad brief in disguise. Designers already know how to write briefs — we just have to apply that muscle to prompts.
The four-part prompt
- Role — who the model is pretending to be
- Context — the constraints, audience, brand
- Task — the one specific thing you want back
- Format — the shape of the answer (paragraph, table, JSON, Figma to production with AI description)
Treat the first reply as a draft
If you accept the first answer, you are using the model at 30% of its capability. Follow up with "now make it shorter," "now argue the opposite," "now write it for a skeptical CFO."
Save the winners
The prompts that work become reusable templates — your personal design system for thinking. Keep them in a Notion page or a learning AI design with Claude Project, not in your scroll history.
Related AI Design Posts
Learning AI Design with Claude: A Designer's Starter Kit
How to get from zero to shipping AI-assisted design work using Claude as your daily thinking partner.
ChatGPT vs Claude vs Gemini for UX Workflows
I ran the same week of design work through all three. Here is where each one actually pulled its weight.
From Figma to Production with AI: A Modern Designer's Stack
The handoff is dead. Here is the AI-assisted stack I use to ship designs straight into a live React app.