In public companies a CEO is generally voted in by a board, and a board is voted in by shareholders that seek to maximize value. Situations where CEO’s hold massive amounts of stock, board members holding stock to vote for themselves, and all sorts of collusion and backroom deals for power do exist, but generally, the mass greed of the shareholders can overcome the individual greed of board members or the CEO.

I think an AI-only approach would likely not work in the long-run. Businesses would push their AI’s to be too evil and without a CEO to get arrested, the government would go after the board. However, I could easily see a company taking a chance with some configuration of human/AI decision making, and other company boards using examples of those successes to chip away at CEO pay until it’s, if not normal, then at least not ridiculous.
Heck - there will probably be a whole new profession of business LLM engineer, training these things to be psycopathic but just on the gray side of the law. (I guess they could just give business school professors comp sci classes…)

I guess my point is that I don’t think individual CEO’s would have a choice. Any world destruction would be during the ever-accelerating capitalist bacchanalia enabled by ‘AI’