AI is killing mediocrity. That's the problem.
The disruption isn't dramatic. It's the quiet elimination of the price floor that used to protect adequate work — and the people who depended on it.
Every AI disruption headline is about the top end. Artists losing commissions. Lawyers being replaced. Coders made redundant. The fear is always about excellence being devalued.
That's the wrong place to look.
What AI is actually eliminating is the price floor on mediocre work. And that is a much bigger problem for a much larger number of people.
Before 2022, if you needed a passable 1,500-word article, a functional first draft of a contract, a working piece of boilerplate code — you hired a human. Not a great one. An adequate one. The floor was set by the cheapest human willing to do the job at the minimum acceptable quality.
That floor is gone.
MIT economists Shakked Noy and Whitney Zhang ran the numbers in 2023: professionals with access to ChatGPT completed writing tasks 37% faster and produced work rated 18% higher quality. The most significant finding wasn't the average improvement — it was that the quality gap between top and bottom writers shrank by 40%. AI helped mediocre writers more than it helped excellent ones. It compressed the distribution downward.
Fiverr's 2023 annual report showed what this looks like in a market: basic copywriting demand down 21%. Simple logo design down 17%. Data entry down 34%. These aren't creative leadership roles. These are the floor.
The standard reassurance is that technology always displaces routine work and creates new work. ATMs didn't kill bank tellers; they made branches cheaper to run, so banks opened more. The loom didn't destroy textile workers; it created mass markets.
This is true as history. It may not apply here, because the speed is different.
The loom gave the economy 40 years to absorb the transition. Children of displaced weavers found new industries. GPT-2 to GPT-4 took four years. The capability jump that took manufacturing technology four decades happened in one product cycle. The economy does not have a generational transition window.
David Autor — the economist most cited in the "technology creates new jobs" camp — updated his labour polarisation thesis in 2024 to acknowledge that generative AI targets non-routine cognitive tasks, precisely the category his earlier work said was safe. He hasn't abandoned optimism, but the hedge is significant.
What's left that AI can't do is narrow but real: genuine accountability (a model cannot stake its career on a wrong call), trust built through years of embodied professional relationship, and frontier work that extrapolates beyond the training data. These are not nothing. But they describe the top slice of every profession, not the floor.
The people who will be fine are those with genuine expertise, deep relationship capital, or the discipline to use AI as a multiplier rather than a crutch.
The people at risk are those whose professional value was priced at adequacy. There are far more of them.
The disruption won't look like mass layoffs in a single quarter. It will look like rates softening, contracts getting shorter, clients pausing retainers. The floor drops slowly, then all at once — and by the time it's obvious, the people who needed it most will have run out of options to build toward the work that replaces it.
Full essay with sources: What AI is actually replacing