IMO the top end goal is to allow capital to access skill without allowing skill to access capital. IF they can make these products work as advertised (big if, very very big if) the corpos can fire shitloads of low-end workers and give their middle managers access to these AI models and nothing else really has to change.
Workers lose their ability to earn a living, but OpenAI and Anthropic become the gatekeepers to effectively all productivity.
I know I sound tinfoil hat here but I don't see how else the numbers work. The only thing full-priced, profit-earning AI can possibly be cheaper than is, in my mind: as much of the workforce's salaries and cost of benefits as it can manage to offset, put together.
It's a great plan if you're a CEO who can't see beyond the next quarter's earnings call. Less so if you want things like, just pulling from a hat here, a functioning economy, a stable society, the ability for humans at large to live. But it's not like the owning class hasn't been undermining all of that for the better part of 50 years anyway.