What changed: the cost of an option collapsed
For most of design history, producing one polished visual option was expensive, so clients paid for production. Image generators changed the economics of the first draft: many options now appear in minutes. When option-making gets cheap, the value moves to the surrounding decisions, which option serves the brief, why, and how it stays coherent across every surface the brand touches.
O*NET's task list for graphic designers already reflects this fuller job: meeting with clients to determine scope, developing the overall layout and production design, and presenting and defending concepts. Generators compress the middle of that list, not the ends. The ILO's task-level research groups design among occupations that are transformed rather than simply displaced, precisely because client intent, taste and accountability do not live inside the image file.
The three layers that still get paid
The first layer is direction: translating a vague business goal into a visual problem statement a generator can even be pointed at. Clients rarely know what they want; they know what they fear. Turning that into a brief remains a human conversation.
The second is systems: a logo is an afternoon, but a visual identity that survives a website, packaging, motion and a dark-mode app is architecture. Generators produce artifacts; designers produce rules that keep artifacts coherent. The third is accountable originality: knowing what you shipped is safe to ship, not too close to a competitor or a training-set artifact, licensed correctly, and defensible when a stakeholder asks why. That responsibility cannot be delegated to a model.
A realistic 90-day plan
First month: use generators deliberately for breadth on one real project, then write down which of the options you rejected and why. That rejection log is taste made visible, and it becomes portfolio material that a mood board never was.
Second month: turn one past project into a documented system: tokens, spacing, type scale, do-and-don't examples. Third month: practice the client conversation. Present one AI-assisted project to a non-designer and defend every choice without mentioning tools. Buyers hire the judgment; the tools are assumed.
