It is tasks that get automated, not jobs
The single most important distinction in this whole debate.
A job is a bundle of tasks. AI does not arrive and take the whole bundle; it takes some tasks, leaves others, and changes the shape of what is left. The research that scores whole occupations tends to overstate the threat. The research that scores tasks gives a calmer, more useful picture: most jobs change, fewer disappear.
What the evidence shows
Frey & Osborne's headline that 47% of US jobs were at high risk treated each occupation as a single thing that is automatable or not.
Reanalyzing the same question task by task, the OECD found only about 9% of jobs at high risk, because almost every job mixes automatable and stubbornly human tasks.
David Autor's long-run view: automating a task often raises the value of the tasks around it that only a person can do, which is why technology keeps changing jobs without ending them.
This is not a reason to relax. A job can survive as a title while most of its day-to-day tasks are hollowed out, the pay falls, or the number of people needed drops. 'The job still exists' and 'your job is safe' are not the same sentence.
Stop asking whether your job will be automated. Ask which of your tasks will be, and what that leaves you doing.