My experience is that if I manage to describe a problem in enough detail for a junior or LLM to be able to solve it, it would have been faster to do it myself.

Prior to LLMs the idea was to involve juniors in the engineering process to give them an opportunity to learn rather than necessarily to improve the team's immediate productivity. Some companies famously (and consciously) refused to hire juniors to avoid the performance hit even prior to genAI (eg Netflix).

Involving LLMs in our engineering processes has very suspect implications for both productivity and quality of our output, since unlike juniors the LLMs don't even learn.