In theory, AI should lend itself to lower pricing of good & services even with lower employment, but that doesn't provide much benefit to the people who can't find a job within their scope of ability. As with anything where the pros and cons are essentially the same, reaction comes down to individual perception. It's probably safe to assume that half of the people affected by job loss in your scenerio are going to be thrilled that they have the opportunity to break out of their cubicles and get their hands dirty. The other half are going to flounder, resist, and resent. That's a lot of people to be thrown into poverty in a very short period of time. So what's a viable mitigation for that? Government assistance programs might keep people fed but they are a trap of government dependence that's already really hard to escape. And ultimately its paid for by the people who create the most value. Almost everyone is charitable but almost no one is okay with being forced to be charitable, so the maximum assistance provided is only going to be as high as value-creators allow themselves to be forced into paying for aka the minimum amount required to prevent mass suffering. So perhaps a lucrative opportunity for someone clever might be to focus on utilizing AI to help people find direction in a newly emerging environment that also isn't demeaning through good intention? With AI, integrating psychology, opportunity-discovery, and training into individualized programs should be not nearly as painstaking a process as any of the existing models for career choice & development and definitely more thorough. If it's done right it could be reformulated for the emergence of intelligent robotics. That particular hurdle seems most likely to be the catalyst for collapse if things go south. Government contracts funded through tax systems would in theory have a warmer reception by the high value-creators since it would minimize their tax liability while decreasing individual long-term government dependence. Whether governments would go for that is a bit of a wild card but if it works and there's pressure, eventually, they'll be forced to support it. Until that point, I feel pretty certain many large companies would utilize something like that to lessen the public backlash of switching to AI based productivity.