AI and Art Direction: What Is Lost
A machine can generate an image. It cannot see.
I am a copywriter, not an art director, and I want to be transparent about the limits of my expertise before I presume to comment on someone else's discipline. But I have worked alongside art directors for thirty years. I have sat with them in studios, argued with them over layouts, watched them agonise over decisions that most people would not notice. And what I have learned from that proximity is that art direction is not the selection of images. It is the creation of meaning through visual choices. It is a form of thinking, not a form of decorating. And it is this form of thinking that AI image generation threatens to displace.
The AI image generators that have emerged over the past few years are genuinely remarkable. I do not say this lightly. I have watched them produce images that are technically accomplished, aesthetically pleasing, and occasionally breathtaking. They can generate in seconds what would take a photographer or illustrator days. They can produce variations endlessly. They can simulate virtually any style, any medium, any mood. These are genuine capabilities, and they have legitimate applications.
But they are not art direction. Art direction begins with a problem — a specific communication challenge that requires a visual solution. The art director must understand the message, the audience, the medium, the context, and the competitive landscape, and then must make a series of visual choices that serve the communication objective. Every choice is deliberate. The angle of the photograph. The colour of the background. The position of the product. The expression on the model's face. The typeface, the point size, the leading. Each of these choices carries meaning, and the art director's skill lies in orchestrating those meanings into a coherent visual argument.
The Helmut Krone Standard
Consider Helmut Krone's work on the Volkswagen campaign. The visual choices were radical and specific. A small photograph of the car, shot from a slightly elevated angle, against a plain white background. No lifestyle. No scenery. No attractive people. Just the car, small and alone, surrounded by space. Every element of this composition was a decision, and every decision communicated something specific. The small photograph said: this is a small car, and we are not ashamed of it. The white background said: we have nothing to hide. The elevated angle said: we are showing you this car honestly, without flattery. The result was a visual argument that reinforced the verbal argument — "Think Small" — with perfect precision.
An AI image generator could produce a photograph of a Volkswagen Beetle against a white background. It could even make the photograph small and surround it with white space. But it could not make the decision to do so, because it does not understand what the composition means. It does not know that a small photograph in a field of white space is a statement of confidence in a market that equated bigness with quality. It does not know that the plain background is a rebuke to the chrome-and-glamour conventions of 1950s automobile advertising. It does not know these things because meaning is not a visual property. It is a cultural property, and cultural understanding is not something that can be derived from pixel patterns.
What I Have Observed
I have watched several agencies experiment with AI image generation over the past year, and the results have followed a consistent pattern. The initial excitement is considerable. The tool produces beautiful images quickly and cheaply. The creative team is delighted. The client is delighted. The production budget is slashed. Everyone congratulates themselves on their efficiency.
Then the work goes into the world, and something curious happens. Or rather, nothing happens. The work does not fail. It does not offend. It does not embarrass. It simply does not register. It looks like everything else, because the AI generates images based on what images typically look like, and what images typically look like is what the audience has already seen. The work is technically perfect and communicatively empty. It occupies the visual landscape without altering it.
The art directors I respect most are troubled by this, and they are troubled for the right reason. They understand that their value lies not in the production of images but in the making of visual decisions — decisions that are informed by knowledge, shaped by taste, and directed toward a specific communicative end. AI tools can produce the output. They cannot make the decisions. And when the decisions are ceded to the tool — when the art director becomes a prompter rather than a director — something essential is lost.
The Photographer's Eye
I want to draw an analogy that I think is illuminating. When photography was invented in the nineteenth century, there were those who predicted the death of painting. The camera could capture reality more accurately and more quickly than any painter. Why would anyone paint when they could photograph? The prediction, as we know, was wrong. Painting did not die. It evolved. It moved toward abstraction, toward expressionism, toward forms of visual expression that the camera could not achieve. Photography did not replace painting. It clarified what painting was for.
I believe something similar may happen with AI image generation and art direction, but only if the art directors themselves understand the distinction. If they define their role as the production of images, they will indeed be replaced, because AI produces images more efficiently than they do. But if they define their role as the creation of visual meaning — as the discipline of making deliberate, informed, strategically sound visual choices — then AI is not a replacement but a tool, and a powerful one at that.
The art director who uses AI as a rapid prototyping tool — generating initial concepts quickly, then refining them with the specificity and intentionality that the tool cannot provide — will be more effective than the art director who works without AI. But the art director who lets the tool make the decisions, who accepts the first pleasing output without asking whether it serves the communication objective, who confuses beauty with effectiveness — that art director has already been replaced, not by AI but by their own abdication of judgement.
The Loss I Fear
The loss I fear is not the loss of jobs, though some jobs will be lost. The loss I fear is the loss of intention. Great art direction is intentional. Every element is chosen. Every choice has a reason. When I worked with the best art directors — and I have been fortunate to work with some of the best in London — I was always struck by their ability to articulate why they had made each decision. The typeface was chosen because its weight conveyed authority. The colour was chosen because it differentiated the brand from its competitors. The composition was chosen because it guided the eye from the headline to the product to the call to action. Nothing was arbitrary. Nothing was decorative. Everything served the idea.
AI-generated images are not intentional in this sense. They are probabilistic. They represent the most likely visual outcome given the inputs, and the most likely outcome is, by definition, the conventional outcome. Convention is the enemy of advertising, because advertising that looks conventional is advertising that is ignored. The art director's job is to subvert convention — to find the visual solution that is surprising, distinctive, memorable — and this subversion requires the kind of deliberate, purposeful decision-making that machines are not designed to perform.
I do not know what the future holds. I am a copywriter in his sixties, and I have been wrong about the future before. But I know this: the art directors I admire are not image producers. They are visual thinkers. And visual thinking — real thinking, the kind that requires understanding and intention and taste — is not something that can be automated. It can only be practised, with skill and care and the stubborn insistence that the visual choices in an advertisement are not decoration but argument.
A machine can generate an image. Only a human can decide what it should mean.