So let's run the presumptions here two ways -- "AI will replace half the white-collar jobs" and "AI will fail to replace any material percentage of white-collar jobs (<5%)"
In the first scenario you just took half of the people at or above the median income and rendered them unemployed with no skill set applicable to the current market. This will collapse consumer spending, lead to mass mortgage and other consumer credit defaults throughout the space in people with formerly high FICO scores and ultimately collapse the asset markets all of which are levered to that debt which can no longer be serviced. At the same time federal tax revenues collapse as well since all of those people are net payers of tax at the federal and state levels, whether a state has an income tax or not (if you don't buy as much you also don't pay sales tax, etc.) The word "crash" understates what is coming if this is the case.
In the second scenario the jobs remain except in the firms currently hyping and levered to AI. Most of those firms ultimately fail and their employees are rendered unemployed. However, most of those skills do translate into other IT-related positions, so while they lose their jobs and quite-likely take large pay cuts, never mind their stock grants being worthless, they do have a skill that is marketable in the current economy. This collapses asset markets, particularly the stock market which is wildly levered to AI being "successful" in producing the first scenario, but does not generally destroy the consumer economy -- although it certainly will dent it badly and will likely take half or more off the stock market simply due to where the gains are and the wildly-outsized spending those firms and their employees with insanely-high salaries and benefits will go "poof."
The point is this: Both paths lead to ruin in the asset markets.
