A field guide to productive AI workflows and ethical practice for designers, artists, and engineers who build things.
Before you touch a tool, build the right mental model. AI in design is not magic, not a shortcut, and not a replacement for your judgment. It is a translation engine: it translates your intent into form. The quality of what comes out is directly determined by the quality of what you put in.
AI is a translation engine, not a creativity engine. It translates your intent into form. Vague intent produces vague output — usable, pleasant, and forgettable. A specific emotional argument, a precise context, a clear constraint: these produce output that means something because you gave it something to mean.
AI knows nothing about your project until you tell it. It doesn't know your design argument, your user, your tone, your constraints, or your influences. Every time you start a session or a new prompt chain, you are starting from zero unless you actively feed in context. Your job is to build that context — and keep feeding it in.
AI can generate a hundred options in the time it used to take you to sketch one. None of them are good or bad until you say so. Choosing what aligns with your design argument — and being able to say why — is where your authorship lives. Accept nothing you can't explain. Discard freely. The curation is the design work.
These workflows are organized around what you're trying to build — not around specific tools. The tools change; the design practice doesn't. In every workflow, the principle is the same: you bring the intent, AI brings the execution speed. The judgment is yours.
AI doesn't know your project. You have to tell it — and you have to tell it every session, because AI has no memory between conversations. The single highest-leverage practice in this entire guide is writing a short Design Intent Document and pasting it at the start of every AI conversation. It takes five minutes. It saves you from spending the whole session wrestling AI back toward your argument.
These aren't rules imposed from outside — they're the practices of a designer who takes their work seriously. Transparency is a posture of confidence, not confession. Knowing where your authorship lives is a professional skill, not an admission. Each of these commitments makes you a better designer, not just a more ethical one.
Your authorship is in the choices you made: what context you built, what emotional argument you brought, what you kept and what you discarded, and why. You can't claim authorship of work you accepted without evaluation. You can absolutely claim authorship of a rigorous, AI-assisted design process — if you can account for the decisions.
Transparency about AI use is a professional norm that's still being established. Be part of establishing it well. Name specifically what AI contributed to your work and what you contributed. The strongest version of this isn't "I used AI" — it's a precise account of the division of labor that reveals the quality of your design thinking.
When you prompt "in the style of [artist]," AI reproduces patterns extracted from that artist's work — work that was in the training data, often without the artist's consent. That artist's style, developed over a career, is being imitated at scale. Use style prompts with full awareness of what you're doing, and consider whether the original artist would want their visual language used this way.
AI image and text models overrepresent certain bodies, aesthetics, cultures, and worldviews because that's what the training data contained. If you don't actively push back on defaults, you're inheriting and reproducing those biases in your work. Notice who appears in your generated images. Notice whose design aesthetic reads as "neutral." It isn't.
Generating images, running large language models, and training AI systems consume significant amounts of energy and water for cooling — far more than most users realize. This cost is invisible in the interface, which makes it easy to use AI habitually rather than intentionally. Use AI deliberately: iterate with purpose, not volume.
Prompts, images, and text you enter into AI tools may be stored and used to improve future models, depending on the platform and your account settings. Treat AI conversations as semi-public spaces. Don't paste proprietary work, personal data, research data about human subjects, or anything that would be harmful if made public.