In the past three years I have read approximately two hundred predictions about AI and employment. I have noticed a consistent pattern: the people making the most alarming predictions are either selling AI tools (in which case fear drives adoption) or selling commentary on AI tools (in which case fear drives clicks). The people with the most direct knowledge of how specific industries actually work — the operators, the practitioners, the people managing teams — tend to give far more qualified and considerably less dramatic assessments.

This does not mean the concerns are without foundation. Some jobs are changing significantly. Some functions that previously required human labour are being automated. These are real shifts with real consequences for real people. I am not dismissing them. What I am questioning is the quality of the analysis being offered, and the motivations of the people offering it.

The "AI will take your job" narrative is an extremely blunt instrument applied to an extremely varied problem. It treats a call centre operator in Manchester, a radiologist in Boston, and a junior copywriter in Singapore as if they face identical risks from identical technology. They do not. The specificity is everything, and the specificity is almost always absent from the coverage.