I watched a video downplaying the disruptive effect of automation on employment. The video was making the old argument that humans are more adaptable, therefore as labor is displaced by machines, the increased productivity creates new opportunities to do what machines cannot, e.g. more creative and strategic type work, then of course he brought up the Luddites and pointed out that they were wrong, that total employment has in fact grown as machines were introduced, disproving the theory.
But then I had a thought, and I think this is the crux of the matter: human labor has outpaced machines because the "adaptation rate" of the humans exceeds the "adatpation rate" of the machines. But that assumption might not hold forever. Improvements in machine learning and meta-machine learning (machines that learn to learn) will speed up the cycle immensely but cutting the machine-learning practitioner, with his slow meat-brain out of the loop.
Basically, think about a company with a bank of computers with AIs in them who's sole job is to come up with new ways to generate value, and design other single-purpose AIs to carry out those tasks. Those AIs could each have a "safety valve" in which if they find data they can't understand they pass it back up to the higher-level AIs to sort out. I say "generate value" rather than "make money", because in a post-AI / post-work world, there will still be value even if normal people have no money. Wealth generation will be measured not in currency but in your share of physical global resources.