That's a possible outcome. Another possibility is that AI will handle all of the thinking and problem solving part. So the market value of thinking will drop. The bottleneck will still be humans, but their input will be (1) doing physical, real-world stuff (2) providing data that the AI doesn't have, e.g. information about a specific problem domain or how does a user interface feel.