Editorial Workflow

An editorial workflow is the defined sequence a piece of content moves through from commission to publication — brief, draft, edit, fact-check, legal or compliance review where required, approval, schedule, publish — together with who owns each step and what has to be true before it advances. In a content platform it is implemented as states on a document plus permissions on the transitions between them, so a draft cannot become published without passing whatever the organisation decided must happen in between. The reason it appears in AI writing tools at all is that generation changed where the bottleneck sits. When drafting was the expensive step, a workflow was overhead on a process that was already slow. When drafting becomes cheap, the expensive step becomes review — and a team that can now produce forty drafts a week without a review capacity to match has not gained throughput, it has gained a queue. This is why the team tiers of AI writing platforms sell workflow rather than generation: Jasper's pitch to marketing organisations is the workflow and the governance, not a better sentence. A worked example: a five-person content team defines four states. A brief is written and assigned; a writer produces a draft with an AI assistant and marks it ready; an editor either returns it with comments or advances it; a subject expert confirms the factual claims; the piece is scheduled. The permission that matters is the one preventing the writer from advancing their own piece past review, because that is the step generation makes it tempting to skip. Where this goes wrong is worth stating. A workflow with more states than the team has people becomes a process everyone routes around, and the routing is invisible until something ships unreviewed. Conversely a workflow with no fact-check step is where AI-assisted content actually fails: a language model produces fluent, confident, structurally correct text containing a claim nobody verified, and fluency is precisely what stops a reviewer from noticing. The step that matters most is therefore not the approval — it is the one where somebody checks whether the specific claims are true, and that step needs a named owner rather than an assumption that the editor will catch it.