Audience
Marketing / content / support leads shipping AI drafts
Friday ship windows love AI drafts. A blog post, FAQ, support macro, or ad primary leaves ChatGPT or Jasper looking polished — then lands live with a wrong price, a fake citation, or a competitor’s feature name.
Treat publish as a risk event, not a formatting step. Before anything customer-facing goes live, name one human owner who will (1) check facts against a source of truth, (2) scan for brand-voice breaks, and (3) accept the downside if it ships wrong. If nobody will put their name on it, it stays in draft.
Route by blast radius. Internal notes and brainstorming can skip the gate. Anything with numbers, legal claims, medical/finance adjacency, customer promises, or paid media needs a read — even if the model “sounds sure.” Paste the draft next to the brief and the pricing sheet; do not review from memory.
Money and risk: one wrong FAQ can burn more trust than a month of seats costs. Paid ads that invent offers invite chargebacks and policy flags. The review minute is cheaper than the cleanup thread.
Skip this tip if you are solo-prototyping throwaway copy in a private doc, or if every AI line already goes through legal/compliance with a checklist. Do not skip if “we’ll fix it if someone complains” is the current process.