The Bottleneck Moved. Most Workflows Did Not.
AI collapsed the cost of drafting content, and most editorial workflows responded by keeping their old shape with a faster first step, which is how teams end up publishing more words with less confidence than ever. The drafting bottleneck did not disappear. It moved downstream, to verification, editing, and judgment, and workflows that have not reorganized around that fact are shipping the problems at scale. Here is the editorial workflow we run and recommend for AI-assisted teams.
The Principle: AI Drafts, Humans Own
Every piece has exactly one human owner who answers for it: facts, claims, voice, links, and the decision to publish. AI is staff, not staff replacement: it drafts, outlines, rephrases, and summarizes, and its output enters the workflow as material, never as finished work. Teams that blur this line publish confidently wrong statements in their brand's name, and readers, and increasingly AI search systems, punish the genericness and the errors alike.
The Workflow, Stage by Stage
- Brief first, always. The AI-era temptation is prompting straight to draft. Resist it: a real content brief, with intent, structure, links, and metadata decided, is what separates content with a job from content with a word count. AI helps write the brief too; the choices in it are human.
- Draft with the machine, openly. Generate against the brief, in your voice guidelines, and label the draft as machine-assisted internally so the next stage knows what it is handling.
- The verification pass, the new load-bearing stage. Every fact, number, name, and claim checked by the owner. AI writes plausibly, which is precisely the danger: errors arrive pre-polished. Budget real time here; this is where the drafting savings get reinvested.
- The voice pass. Rewrite until it sounds like your business and says something only your business would say: the specifics, positions, and experience that survive the sameness pressure and give AI search a reason to cite you.
- The technical pass at publish. Links, schema, metadata, per the brief. Then measurement against the goals the brief named.
Cadence and Honesty Rules
Two governance rules keep the system healthy. Volume is set by verification capacity, not generation capacity: publish what you can stand behind, which is still more than the pre-AI pace, just not infinitely more. And keep an internal record of how each piece was made; disclosure debates aside, you cannot manage what you have not tracked, and the workflow only improves if the record exists, the same drift-prevention logic as website governance.
We run this workflow for our own content and our clients' content programs: machine speed, human ownership, and a publish button that still means something. If your team's output went up while trust in it went down, borrow the workflow. We will help you install it.
