Automation loves specificity.
If a task is easy to explain, it's easy to teach to a robot.
So, if we want to avoid competing with superhuman machines, we need to run in the opposite direction.
The more abstract the value being provided, the more resistant it is to replacement (and therefore the less likely you are to be taken out of the game).
Think about it:
At one point, we were subsistence farmers, working to meet basic physiological needs of ourselves and our families.
Agricultural advancements brought many of us to cities, where we solved new problems, including other physiological ones (like a doctor) and more abstract ones (like a banker or physicist).
The trick is to project this process into the future.
Already, robots are learning to suture. But it seems less likely that they'll be able to comfort a nervous pre-op patient.
AI models are becoming more reliable, but they don't seem to have reputations in the same way you and I do.
Automation will almost certainly decrease the number of workers needed for existing roles, in healthcare and elsewhere. But it may increase the breadth of available roles overall.
And by banking on this, we give ourselves a shot at reaping the real rewards.