Plan & Govern
Set the obligations, risk classification, and sign-off criteria each AI system must meet, before anything is built.
Six stages, running as a loop. Read them top to bottom, then notice the sixth hands straight back to the first. Each stage names what Qapitol does, in context, never as a one-time gate.
AI systems don’t hold still. They drift, they get re-prompted, the data underneath them changes, so the work of standing behind them can’t stop at a single sign-off. It runs as a continuous loop, where the last stage feeds the first.
Six stages, running as a loop. Read them top to bottom, then notice the sixth hands straight back to the first.
Run & Optimise monitors behaviour and drift in production (because AI systems change after launch) and feeds what you learn back into the loop. What you find in production reshapes how you Plan & Govern the next turn.
The last stage feeds the first.