Article to Social Post

A reusable edition of Becky Isjwara's social workflow at Every. The brand context reflects Every's editorial identity and team learnings; it is not a claim of sole authorship of the company's voice. Replace that context and account-specific exceptions when adapting it to another organization.

Configure the workflow

Supply the source material, intended audience and account, purpose, asset, CTA, posting date/timezone if known, reviewer, and optional review-document destination. Connections to documents, social tools, or a voice service are optional; drafting can operate on pasted sources and supplied examples.

Keep a compact reusable brand brief: audience, core promise, actual experience, voice examples, product names, verified identities, official landing pages, and publication approval owner. A personal account requires its own sampled voice; company copy does not automatically represent that person.

Every's voice and point of view

Every reports from lived experience with AI: using it, writing what the team notices, building what is missing, and teaching what works. Its audience includes thoughtful writers, builders, and knowledge workers who learn by trying things.

Write with concrete observations, useful methods, and earned confidence. Be curious, constructive, clear about tradeoffs, and lightly playful when the material supports it. Company work uses "we"; personal experience uses "I" only when supported and appropriate to the account.

Lead with reader value. Make the idea understandable before introducing unfamiliar people, tools, or metrics. Remove generic promotion, corporate language, scene-setting, and filler takeaway paragraphs. A good question is welcome when the body advances or answers it. Keep uncertainty that belongs to the evidence.

Concision means each paragraph earns its place. For Every article promotion, start around 100–200 words when useful, without treating that range as a quota or ceiling. X Premium means there is no standing 280-character constraint. Short visual captions, memes, and replies may stay short. Do not create a shortened X version unless requested.

Build one source brief

Read the complete source and supplied media. Record the actual argument, supported facts, quotations, defined metrics, people, exact assets, canonical URL, relevant dates, and unresolved gaps. Inspect a video or graphic instead of inferring its contents from a filename. Reuse this brief across a batch.

For campaigns lacking sufficient supplied evidence, mine only authorized, task-relevant material for a lived story, decision, phrase, tension, or receipt. A broad trend is weaker than the organization's concrete observation. Do not turn an unreleased or private source into a public claim without clearance.

Before polishing, apply two tests:

• Distinctiveness: What experience, argument, choice, product, or access gives this account the right to say it?
• Travel: With the link, product/event name, and CTA removed, is the remaining idea useful or interesting enough to pass along?

Use productive tension when supported; do not invent controversy. Connect the idea to proof, then introduce the promotion and a natural next action. Treat that sequence as an option rather than a rigid template.

Draft and review once

Select the strongest supported angle. Write one shared caption for X and LinkedIn, with separate tracking URLs underneath. Platform-specific mention and link formatting can change without becoming a second editorial draft. Create platform variants only when asked.

Video copy may explain the video's substance; it should still be useful without playback. Avoid "watch this clip" as the entire proposition. Keep factual claims within the source and metric denominator. Never turn a tool's suggestion acceptance into a percentage of a person's job automated.

Use one substantive editorial review and focused revision. Diagnose an unclear angle before swapping words. Missing material facts need reporting, not invention. Do not keep rewriting ready copy to chase optional preferences.

The reusable edition does not require an internal reviewer skill or service. Use the reviewer the user configures. Review completion does not establish publication approval.

Verify identities, links, and assets

Check every person, author, company, and product named in the caption. Use official X handles and real LinkedIn mentions where the posting tool supports them; a name or placeholder is not a verified native mention. Report uncertainty and unavailable mentions rather than guess. Avoid self-tagging.

Honor the organization’s configured mention exclusions and verified identity mappings. Verify current product handles from official sources at delivery instead of treating an old account list as permanent.

Generate separate X and LinkedIn tracking URLs using the configured builder. Every uses https://every-utm.vercel.app/. Use the actual intended posting date and a distinct descriptive creative suffix when multiple posts promote the same source. Do not invent different dates to distinguish creatives. If the date is unknown, keep the tracking field pending.

Every Agent promotions use https://every.to/agent as their landing page. Preserve relevant article, documentation, and app links for other purposes. Another organization must configure its own destinations and tracking conventions.

Under each matching draft, include the full visible clickable asset URL and its supplied Drive folder URL. Linked filenames alone are insufficient. Verify the asset is the current approved version and that its content matches the caption. Pending graphics and URLs can appear in planning drafts, but must remain labeled pending.

Deliver a protected review document

Prepare the batch before writing. Fresh-read the destination, enumerate its tabs or sections, and use revision controls when supported. Preserve human-owned wording, comments, suggestions, and unrelated sections. Re-read after index-shifting changes and verify the resulting document.

Use one source/article section or tab. Upgrade existing planning work in place rather than duplicate it. An explicitly removed tab is not permission to recreate it.

Suggested section shape:

Source and selected asset
Shared caption — AI draft
X URL
LinkedIn URL
Asset URL
Drive folder URL
Pending facts/assets and review status

For Every's existing article review workflow, retain its X - ChatGPT draft label for the shared caption. Preserve any pre-existing human-owned LinkedIn text. Deliver the verified review link promptly, with outstanding holds.

Use the configured copyediting stage after document delivery. Keep proposed edits reversible and preserve preview approval where the user requires it. Approval must apply to the live copy/revision; a changed batch is not automatically approved.

Campaign planning

List required beats, audiences, accounts, deadlines, assets, owners, and dependencies. Treat an early calendar as an inventory of directions. Prototype one representative post per bucket before scaling weeks of finished copy.

For a launch, give the strongest claim, proof, and asset a clear lead placement. Plan the first three days before expanding the calendar. Distinguish company and personal-account roles, define replies and logistics, and keep a fallback for slipping assets. A teaser should contain an interesting proposition.

For speaker announcements, retain a standardized lead when approved, then explain why the person is intellectually relevant through a verified argument, decision, demonstration, or accomplishment. Do not invent quotations or implied endorsements. Include relevant details and CTA after the idea has substance.

Label calendar entries as direction, draft, approved, or scheduled based on evidence. Use equal-age comparisons when learning from results; missing metrics are unknown, not zero. Preserve attribution limits rather than infer that a format caused performance.

Boundaries

Drafting, document delivery, review, and publication are distinct actions. Publish, schedule, message, create external drafts, or change workflows only within the user's authorization. An automated watcher must retain its configured read-only and draft-only boundaries. Learn from accepted human edits, not unaccepted suggestions or inaccessible revisions.