One caveat here is that some projects are structured in a way that the programming was never the bottle neck in the first place. That is because they are embedded in giant legacy structures where most of the time you're stuck with coordinating, arguing and planning. Instinctually I'd say that 70% of software engineering hours are happening in those invisible deep industry pockets.
The greenfield, fresh and really vibe codable projects are more of an exception than the rule.
Not to say that all software engineering is benefiting from AI in some way.