Knowledge work needs to be separated into 2 categories. There are those who earn a living from having asymmetric skills (e.g. knowing how to use Excel) or access to asymmetric information which are not difficult to acquire if time permits. These will be replaced by GenAI. And they should be replaced. There is the other category of knowledge work where it requires domain-specific critical thinking such as problem-framing, what-if scenario analysis, etc. These rely partly only on historic data but requires making very subjective future-state 'predictions'. These kinds of work may not replaceable until the advent of AGI. By lumping knowledge/creative work as a whole, we make the flawed assumption that they are cognitively equivalent, but there are many gradients of cognitive abilities required.