When an entire industry outsources its thinking to the same handful of models, the differences between companies flatten out and the skills employees were hired for quietly rot.
At nine on a Monday morning, 28-year-old copywriter Lukas Weber starts his week by checking on a small army of AI agents. A newsletter draft, several client proposals and an analysis of a recent campaign were all produced by them; he only reviews. His team shrank this year, so his workload grew, and the agents are how he keeps up. 'The expectation from our executives is I shouldn't write at all,' he says.
He is not unusual. His whole company works this way, and so do its competitors.
Everybody hired the same employee
Everyone is building on the same small ecosystem of models. Your marketing plan and your rival's marketing plan come out of the same underlying brain.
The result is a convergence in style. Em dashes suddenly appear everywhere because the models love them; newsletters, ads and websites read as though one company made them all. In a 2026 study, a Duke researcher and a Technion cognitive scientist tested 22 commercial models against more than 100 people on standard creativity tasks. Individual models sometimes beat individual humans, but the models converged on strikingly similar answers while the humans did not.
What happens when you straighten the river
In 1962 the US Army Corps of Engineers decided central Florida's Kissimmee River had too many bends and flooded too often, so they straightened and dredged it. The flooding stopped, but oxygen levels in the water fell, sediment washed into Lake Okeechobee, nearly 35,000 acres of floodplain habitat were destroyed and roughly 90% of the wintering waterfowl disappeared. From 1999 they spent more than a billion dollars putting 44 miles of bends back.
Organisations have bends too: an idea gets discussed, challenged by colleagues, refined. AI removes that, and employees scurry off to produce things alone. Output rises, but so does the odds that nothing your company makes could only have come from your company.
The cheap end and the expensive end
Researchers predict a barbell-shaped market: near-zero-cost AI output at one end, verifiably human work selling at a premium at the other, and a hollowed-out middle. The 'human in the loop' review jobs move to cheaper labour markets, because if the output is all the same, why pay a premium to oversee it? The AI companies themselves are already hiring highly paid tastemakers, creating the blandness and insuring against it at the same time. And one 2026 report found that skills atrophy shockingly fast once AI is introduced.
In one line: if every restaurant in town is cooking from the same recipe, what exactly are you selling — and how many of the cooks still remember how to cook?
Why it matters
AI speeds up output, not distinctiveness: when every competitor runs on the same models, the advantage turns negative. The real warning is that firms are cutting the judgment they paid for while quietly losing the thing that made them valuable, and the bill usually shows up years later.
Curated from high-quality sources, with concise summaries and key takeaways.
