The approval gate

AI drafts, the human approves, the system records it

Most of the content in this studio starts as an AI draft. No draft goes live without the human having explicitly approved it — approval is not a date, it is a state in the system.

How the gate works technically

Every piece of content carries an approval flag. While it is unset, the production build does not build the content at all — not greyed out, not hidden by a stylesheet, simply not present in the build. Approval is therefore not a declaration of intent but a build condition: a draft cannot go live by accident, because doing so would require a second, independent step that has not happened. Who approved it sits directly in the content’s metadata, with no separate logging system.

For other content types in the studio’s day-to-day operation, the same approval runs through a single central inbox rather than several parallel review loops: one channel, one decision.

Why this is a consulting question

The question when adopting AI is rarely “AI, yes or no”, but: which step needs a gate, and how low-friction can that gate be built. A gate that takes days gets bypassed. A gate anchored in the system rather than in the goodwill of the people involved does not get bypassed — it cannot be.