Publish more, staff less

A small editorial desk can produce good work or produce it frequently, and the constraint is almost never talent. It is the volume of non-writing that surrounds each piece: formatting, tagging, laying out, building the email, resizing for social, scheduling everything, and then measuring whether any of it landed. The writing is the short part of the day.

Automating that surrounding work is where AI earns its place in publishing, and it is a considerably less glamorous claim than automating the writing. Drafting and tagging get a first pass from the model, an editor approves or rewrites, and everything downstream — web publication, the email edition, social scheduling — happens from the approved piece without anyone rebuilding it three times in three formats.

The approval chain is the load-bearing part. Anything AI-generated enters as a draft with a named human between it and publication, because the alternative is a publication whose voice drifts and whose facts nobody checked. The leverage comes from removing production work, not from removing editorial judgement, and a system designed the other way round produces volume that damages the masthead.

Around that sits what a publication actually needs to operate: web publishing with SEO controls, embedded video, paywall and membership tiers, reader analytics tied to conversion, and multi-publication management for desks running more than one title.