Every: how an AI-native editorial team structures its workflows role by role

What the article is about

Every, a publication focused on technology and business, published its editorial AI guidelines in February 2026. Unlike generic guides about AI writing tools, this piece reveals the actual workflows each member of the team uses — from the editor-in-chief to the podcast producer. It was written in a moment when the team had moved from experimenting with AI to building stable, repeatable practices into their day-to-day work.

Context

Every covers AI closely as a topic, so the team is unusually motivated to work out what it actually means to integrate AI into editorial production. The article is not advocacy — it also surfaces points where the team disagrees or where AI has not provided clear value. That candor makes it more useful than most editorial AI coverage, which tends toward either enthusiasm or skepticism without operational specifics.

Key workflows

The editor-in-chief, Kate Lee, uses a custom Claude skill called “top-edit” that checks completed drafts for common quality problems: vague pronouns, unsourced claims, hedging phrases, and writing patterns that signal AI-generated text. This runs as a first-pass check before Lee reads the piece herself, reducing the time spent identifying mechanical issues and focusing her reading on the harder editorial judgments.

Eleanor Warnock, the managing editor, runs AI evaluations on pitches and drafts before detailed review — not to decide what to publish, but to identify where a piece needs the most work. She also runs style-check skills against the publication’s house guidelines and stores all skills in a shared GitHub repository so the whole team can use and update them.

Staff writer Katie Parrott describes her approach to AI-assisted drafting as nonlinear — more like shaping material than building from a blank page. She uses Claude projects with Monologue, a voice-to-text tool, to externalize thinking about a piece through a conversation before writing. The article includes her point that writing must “articulate truth, offer learning value, and sound authentically like her voice” — AI accelerates the process but does not change the standard.

Contributing editor Jack Cheng uses AI as what he calls “relief pitchers” when repeated editing cycles on a single piece have blunted his perception of it. He also built a Claude agent that identifies gaps between different pieces within the publication’s coverage — flagging topics referenced in one column that have never been directly addressed in another.

Anthony Scarpulla, the social media manager, built a custom tool using Claude Code that links the X API to the publication’s article database. The tool generates multiple “building blocks” — quotable sentences, proof points, concrete examples — rather than finished posts. Scarpulla then selects from those materials himself, which he frames as the part of the work where his editorial judgment creates the most value.

Key argument

The article’s clearest contribution is demonstrating that a sophisticated AI editorial integration is not a single unified system but a collection of role-specific tools built against a shared standard. The standard — what quality means, what the publication’s voice sounds like, what must always involve a human decision — is set by the editorial team. The tools accelerate work within that standard. All team members retain final human judgment over what publishes.

Who should read this

Editorial teams evaluating how to build stable AI workflows across different roles rather than adopting a single platform. Also useful for independent writers who want to see what a well-developed individual practice looks like when described with operational specificity.