I have written many times that current levels of AI are good enough to radically change work. They are not drop-in replacements for people (do we really want them to be?) but they are good enough It would be helpful to take this as a given and start thinking through the outcomes

Assuming current levels of jaggedness hold, this will take place gradually over multiple years as systems and processes are adjusted for AI. If the AI companies are seriously able to deliver a continuously learning AI system that beats humans at every task, it might happen faster