We are approaching the age where the problem will be "what to do with legions of unwanted workers that serve no useful function". The problem of "not enough young workers and consumers" is only a problem if the timing is wrong. If AI isn't reliable enough to function in the real world in the next 20-30 years, that probably means it's a much harder nut to crack than assumed and it's not going to be solved before that demographic prob becomes a massive drag on global GDP.