Why is Learn and Adapt at the centre?#
Because it changes all four stages. What the organisation learns from a judged outcome should alter which outcomes it pursues, where it draws the delegation boundary, how agents are configured and what counts as good enough. Treat learning as a fifth step and it happens once per cycle; put it at the centre and it happens after every outcome. John Boyd's OODA loop works the same way: speed up the act step, as AI does, and the load shifts onto orienting and deciding.
Which decisions stay with people?#
Two: what is worth building, and whether to accept the risk of releasing it. Automate the first and you build the wrong thing efficiently. Automate the second and nobody is accountable when it fails.
Everything in between is a question of evidence, which can be automated once it has proved reliable.