I've been figuring out at what point a tiny AI-augmented team is guaranteed to outperform. The Coase angle helped, which explains that most headcount exists for coordination, not judgment. AI agents don't replace judgment, but they do eliminate the coordination overhead that justified hiring people in the first place. A lot of senior roles might just be expensive pattern matching.