Exposure is not the same as replacement
Being affected by AI and being replaced by it are different things.
Most of the alarming headline numbers measure exposure: the share of tasks a model could touch. Exposure tells you where change is coming. It does not tell you whether that change means a tool in your hand, a smaller team, or a missing job. Reading exposure as replacement is the most common mistake in this conversation.
What the evidence shows
The 'GPTs are GPTs' study found around 80% of US workers could have at least 10% of their tasks affected by LLMs, and about 19% could see half their tasks affected. The authors are explicit that this is exposure, not a forecast of job loss.
The ILO's global analysis concluded most exposed jobs are more likely to be augmented than automated, with clerical work the most exposed category.
Higher-paid knowledge work turns out to be more exposed than manual work, reversing the usual assumption about who automation hits first.
Exposure can still end badly. High exposure plus weak bargaining power plus an employer chasing cost is exactly how augmentation quietly becomes a smaller payroll. Exposure is a warning light, not an all-clear.
Treat a high exposure score as a prompt to move, not a verdict. The question is whether you shape the change or it shapes you.