AI is being used as an excuse to jack up workloads and cut staff.

If your workload is jacked up and you're forced to use AI, the only way that works is if you rubberstamp everything the AI does.

The human is there to take the blame for when the AI inevitably fails, an "accountability sink" as Corey Doctorow says. If the human actually tries to review everything the AI does, then he will be less productive than his peers who rubberstamp everything the AI does, and will be fired for poor performance.

I don't see how AI coding can be anything other than a disaster in the long run. If every programmer in the world is using AI for everything, who is going to write the training data for the AIs of the future? These LLMs are just fancy Markov chains "predict the next word in the sentence". They are not true AI. That's why you can tell it "don't do X", and it'll screw up and do X anyway.