AI for designers
For designers, AI is most useful as a stress-test partner — not as a replacement for taste. It'll generate a hundred ideas in a minute, but you still pick the one that fits. The wins come from removing low-value work (variants, first-pass critique, microcopy) so you can spend time on the parts where craft actually matters.
What AI handles well
UX critique that finds what's broken
The problem. You designed a flow. Friends are too polite to give real feedback. You want sharp critique grounded in actual UX principles, not vibes.
What AI does. Use the UX critique prompt. Get specific friction points, drop-off predictions, and the single highest-leverage change — referenced against named principles (Hick's, Fitts's, recognition vs recall).
Image generation prompts that actually produce what you want
The problem. Generic prompts give generic images — stock-photo-feeling AI slop. You need specific outputs for hero images, illustrations, social.
What AI does. Use the image-prompt builder. It includes style anchors (specific photographers, art movements), negative prompts, and aspect ratio guidance for the chosen tool.
Wireframe in text before opening Figma
The problem. You're about to design a feature. Figma is where you go to sketch — but you should think through structure first, not pixel-push.
What AI does. Use the text wireframe prompt. It forces hierarchy decisions (primary action, secondary, what to hide on mobile) before you commit to a visual direction.
Brand voice from samples
The problem. Your brand has a voice but it's undocumented. New writers / new agency / new founder drift toward generic. You need a guide that's specific, not vibes.
What AI does. Paste your best samples. AI extracts voice signatures — vocabulary, rhythm, persona, what to avoid. Output: a voice doc that actually constrains writing instead of just describing it.
Color palette with actual reasoning
The problem. You need colors. Pinterest gives you trendy ones. Figma gives you randomly generated. Neither has reasoning tied to your audience or context.
What AI does. Use the palette prompt. It asks for audience, tone, references, accessibility constraints — then explains every color choice and confirms WCAG contrast ratios.
Empty state and microcopy that doesn't suck
The problem. You have a "no results" state. Default options: "Oops!" / mascot saying something quirky / generic text. None of them help the user.
What AI does. Use the empty state prompt. It writes copy that explains what's happening and what to do next, without fake personality.
Accessibility audit checklist
The problem. You want to ship something accessible but reading the entire WCAG spec right now isn't happening.
What AI does. Generate a checklist tailored to what you're designing (form, modal, landing page, app screen). Top-5 priority list for if you only have 30 minutes.
Tagline / hero copy variations
The problem. You're writing the hero of a landing page. Your first draft is "Empowering the future of X" — you know it's bad, but you're stuck.
What AI does. Use the tagline prompt that generates options across different angles (functional, outcome-led, pain-led, identity-led, counterintuitive) so you can compare frames, not just word-swap.
Your AI stack
Start with the foundation. Add specialized tools as the work calls for them.
Foundation LLM
Specialized add-ons
Prompts ready to use
Get started in 30 minutes
Pick an image generation tool and run 3 real prompts
10 minMidjourney or DALL-E. Use the image-prompt-builder template. Compare to what you'd get from a generic prompt — the difference is the art direction.
Generate a brand voice doc from your existing best content
10 minPaste 3-5 of your best brand-aligned pieces into the voice extraction prompt. Save the resulting doc — it's your reference for any AI-assisted writing going forward.
Run the UX critique prompt on something you shipped recently
10 minBefore defending it, see what AI flags. You'll either get useful feedback or confirm your instincts. Both are valuable.
Common mistakes
Treating AI output as the answer instead of one option among many. AI gives you 80% of the way there. The last 20% — the part that makes the design YOU — has to come from you.
AI-generated images without art direction. "A landing page hero" gives you stock-photo slop. "A black-and-white photograph of a sunlit office, shot on 35mm at f/2, mid-afternoon shadows, [reference photographer]" gives you something usable.
Skipping critique. Using AI to validate work you're already attached to defeats the point. The value is in stress-testing, not affirmation.
Outsourcing taste. AI can generate 100 color palettes. It cannot tell you which one fits YOUR brand and audience. That judgment is irreducibly yours.
Using AI image gen for production assets without checking licenses, model lineage, or whether you're comfortable with the training data origin question. Some teams ban AI imagery for legal/IP reasons.