AI is not a creative tool.
It's a consensus machine.

The sentence your brand strategy vendor will never say to you: AI produces the most agreed-upon version of everything.

Not the best version. Not the most original. The most approved-of. The statistical mean of desire, dressed in the syntax of origination.

This is not a criticism of the technology. It is a structural description of how it works and almost no one in brand strategy has followed it to its full conclusion.

Large language models are trained on the accumulated output of human expression. Everything written, made, shared, upvoted, cited. They are then refined through reinforcement learning from human feedback: human raters score outputs, and models are optimised to produce what humans rate most favourably. The result is a system trained to maximise agreement. Not truth. Not specificity. Agreement.

By late 2022, the visual consequences were already visible. Midjourney users began documenting a convergence across billions of AI-generated images: the same lighting geometry, the same face structure, the same compositional logic appearing across ostensibly distinct prompts. What looked like creative variety was, at the distribution level, a single aesthetic mode, namely the most approved-of version of 'beautiful image' averaged from the training corpus. The style became so recognisable that designers coined a term for it: the Midjourney look.

The implication for your brand is precise and uncomfortable. When every brand in a category uses the same AI tools, every brand in that category produces approximately the same outputs. Not because the tools are identical. Because the training distribution is the same. The consensus is the same. The mean they are all optimising toward is the same.

In financial services, this is already measurable. Brand analyses of major retail banks' verbal identity in 2023 and 2024 found near-identical vocabulary clusters across competing institutions: 'partner,' 'trusted,' 'growth,' 'community,' 'your future.' Different logos. Statistically identical language. The category has averaged itself into illegibility and every AI writing tool used to generate that copy drew from the same training distribution.

This is not a creativity problem. It is a structural convergence problem baked into the technology itself.

Here is what makes this difficult to solve: the problem does not look like a problem. AI-generated brand content is usually good. Competent, coherent, appropriate, polished. It passes review. It gets approved. It performs reasonably well in testing. The issue is not that it is bad. It is that it is averaged.

The strategic question has therefore changed. The question is no longer 'are we distinctive?' Distinctiveness is a surface property. The question is: what in us is genuinely non-mimetic? What could not have been generated from the consensus of what already exists?

The brands that will exit this crisis with genuine competitive position will not be the ones that use AI more cleverly. They will be the ones that understand what AI structurally cannot produce and build from that.

This is the beginning of the argument. The rest follows from this one structural fact.