AI-generated imagery for marketing creative, advertising and social feeds produces exactly two kinds of work: the kind nobody notices, and the kind everybody hates.
The hate is loud and largely earned. Antipathy toward AI slop is at an all-time high, and most of what triggers it deserves the contempt it gets. What gets lost are the trained creatives that are running AI pipelines successfully every day, producing beautiful, impactful work for brands you already trust.
Both come from the same tools, and the difference isn’t the software.
Generative AI tools were trained on everything that already exists, which means every output pulls toward the statistical center of what came before. You cannot optimize for originality. Midjourney, Nano-Banana, and Dall-E will only ever produce an image that could plausibly have appeared anywhere, made by anyone, for any brand. Bland acceptability is easy to produce at scale. If you want distinctive, that requires someone who knows the difference, and that someone will never be software.
The invisible standard:
That’s the standard: not “does it pass?” but “does it disappear into the work?”. Getting the right output requires two things that consumer tools can’t provide on their own: a model trained on your brand’s specific visual language, and a creative who knows what they’re doing.
At rbb, we build brand-trained models. We teach those systems how your brand composes shots, what your lighting is, which aesthetic choices make you recognizable before anyone sees a logo. The model learns your visual vocabulary instead of averaging the internet’s. That’s one part.
The other piece is the person running it. Our art directors understand image-making at a technical level, which is why the outputs don’t look like everyone else’s. They direct the AI the way they’d direct a photographer or illustrator: with intention and a specific result in mind, with the judgment to know when to keep going and when to edit. They also know when not to use it. When a studio shoot serves the work better. When stock photography is the right call. When an AI element should be composited in traditional photography rather than standing alone. The decision about which tool to use is as consequential as how well you use it.
The conversation worth having:
The AI-in-creative discourse has settled into a binary: you’re either embracing the future or resisting it. Both camps tend to skip the harder questions.
Using generative AI in professional creative work is an ethical decision, regardless of whether anyone treats it that way. The industry’s obligation to the people who do this work for a living didn’t expire when the technology got interesting. Transparency with clients and consumers about what’s AI-generated and what isn’t is the bare minimum, not a virtue. The goal shouldn’t be to obscure the process. It should be to make the process irrelevant to the quality of the result.
The replacement narrative deserves more skepticism than it usually gets. Engaging with these tools doesn’t require accepting the premise that they’re substituting for human creative intelligence, because they’re not. A well-deployed AI production pipeline changes what art directors spend their time on. It doesn’t change what art directors are for. The conceptual thinking, the creative judgment, the brand knowledge, the relationship between a maker and their work, none of that is in the tool. The tool handles volume and execution.
The current generation of AI tools won’t produce breakthrough creative work on their own. The architecture won’t allow it. They converge toward the mediocre, and mediocre is a sophisticated average of what already exists. Nobody cares about it. You can fight that tendency with training, taste, and creative direction, but the intelligence driving that fight is still human. Whether a future iteration of the technology changes this is an interesting but irrelevant question. Right now, the tool is powerful within a specific range and knowing where that range ends is my job.
The Bottom Line:
For brands, a well-deployed AI production pipeline changes the math on specific problems.
Campaign concepts that previously required full reshoots every time a channel or format changed can be scaled across placements without starting over. Social content that’s been defaulting to generic stock because bespoke assets cost too much to produce at volume can get imagery that’s actually on-brand. Seasonal variations can be built within guidelines rather than drifting from them.
Building and maintaining a brand-trained model with proper creative oversight is an investment. But it lets that investment work harder, and it lets smaller brands hold a visual presence next to organizations spending significantly more on production.
Breakthrough work is still the goal, whatever the budget, whatever the tool. That hasn’t changed.