The problem is that the first is easy, the later is hard, nigh impossible.

Software development is notoriously hard to predict. Specially features against time and cost of development. But video games are even harder to predict. It's impossible to know how many copies a game will sell, you might as well hire a tarot reader. Specially if the game doesn't exist yet.

This is not a justification of the AAA practices. Quite the contrary, things are this way because mid and high management refuse to do their job or plainly suck at it. I guess that the adage still rings true: I want smaller games, with worse graphics, made by well paid developers who work less hours. But this games have never sold billions of dollars or sparked billion dollar game as a service IPs. So executives think it can't be done and keep expending more in a desperate chase of the golden eggs goose.