Does this replace HR, legal, or leadership judgment?

No. Claude drafts the runbook, the manager pack, and the announcements, but leadership, HR, and legal own the decision and the sign-off. Every word is read by a human before it sends, every placeholder is filled with a confirmed fact, and anything touching employment terms, roles, or comp is routed past legal and — regionally — the works council or labor authority before it ships. Claude gives you a shape and a sequence; the judgment stays human.

How do we keep a reorg confidential before we announce it?

Keep every affected name, reporting line, and comp detail inside a workspace your company has approved for sensitive people data, and out of any prompt that doesn't strictly need it — use roles and [placeholders] instead. The runbook and the drafts only need the shape of the change, not the private facts. And put the embargo and the approval gates on the timeline so nothing leaks early through a premature sign-off or a forwarded draft.

What's the single biggest mistake in org-change comms?

Sequencing — letting people, especially managers, hear it through the grapevine before the official, in-order rollout. A manager blindsided about their own team in front of that team is a trust failure you can't undo. The fix is the runbook: leaders briefed first, managers before their teams, all-hands last, with an owner and a time on every step so the right person delivers the news in the right order, ahead of the grapevine.

How do we handle Arabic and English without one feeling like an afterthought?

Run the whole comms suite twice — once per language, in parallel — and adapt rather than machine-translate, so the Arabic reads native instead of translated. Hold both languages to the same approval gate, and make sure an approver who actually reads Arabic reviews it, so the second language never skips the review the first one gets. Run policy changes through the plain-English rewrite in each language, not English-then-MT. See the bilingual-rollout variation.

Run an org change without losing the room

when to reach for this

Org changes — a reorg, a policy rollout, a leadership change, a return-to-office mandate — rarely fail on the decision. They fail on the communication. People hear it through the grapevine, managers get blindsided in front of their own teams, the message says three different things in three different rooms, and the trust you spent years building takes the hit in a single day. The fix is to orchestrate the change like a launch: one source-of-truth narrative, a manager enablement pack, a sequenced rollout (leaders → managers → all-hands), an honest anticipated-questions FAQ, and a real feedback loop. This is the cross-cutting capstone of the People & HR set — it pulls the sensitive-comms craft, the plain-English policy rewrite, and the competency framework into one runbook so a hard change still lands with the room intact.

gather this first

What's actually changing and the honest why — and crucially what is decided versus still open, plus what you legally or contractually cannot say yet. Type it into the chat in Claude Desktop, or drop it in the workspace as change-brief.md. The message can only be as clear as your own clarity going in.
Who is affected and how — roles, reporting lines, locations, comp — and the approval chain: leadership, HR, legal, and regionally any works council or labor authority that has to sign off before anything ships.
A Claude Desktop workspace your company has approved for sensitive people data — every affected name, reporting line, and comp detail stays inside it. Open the folder in Claude Desktop and ask in the chat; approve each read in the 'Ask permissions' prompt so you can see exactly what Claude opens.
The timeline and the non-negotiable sequencing constraints — the embargo date, the all-hands slot, and the hard rule that managers must hear before their own teams do.

the workflow

Lock the narrative and the 'what we can't say yet' list — read it back first

Before any drafting, open the folder with your change brief in Claude Desktop and have Claude read the situation back to you — no terminal needed. Sensitive rollouts go wrong by saying too much, not too little, so set the guardrails before a single message exists: who's affected, the one thing each audience most needs to know, what's decided versus open, and the explicit do-not-say list. Keep individual names out — use roles and [placeholders].

you ask

Read change-brief.md. Before drafting anything, read it back to me: who is affected and how (roles, not names), the one thing each audience — leaders, affected managers, the whole company — most needs to know, what is actually decided vs still open, and an explicit do-not-say list of what is legally off-limits, embargoed, or undecided. Keep all individual names out — use roles and [placeholders]. Don't write any announcement yet.

what you get back A short read-back — affected audiences, the core message per audience, decided-vs-open, and an explicit 'won't say' list — that catches a wrong assumption now, before it's wrapped in three careful announcements. No names, just roles and placeholders.

The read-it-back-first move matters most here, because the cost of a wrong org-change message is measured in trust, not typos. Confirm Claude has the situation right — especially the do-not-say list — before it writes a word.
Sequence the rollout into a comms runbook

The sequencing IS the strategy. Decide who hears what, in what order, from whom, and when — leaders briefed first, managers before their teams, all-hands last — with an owner and a time on every step, exactly the way a launch runbook sequences its channels.

you ask

Build a rollout runbook as a table: each audience (leadership, affected managers, affected teams, the rest of the company) on its own row, with what they hear, who delivers it, when, and which step it depends on. Enforce the sequence — leaders briefed first, managers briefed before their own teams, all-hands last — and put every approval gate (HR, legal, works council / labor authority) on the timeline as its own row with an owner and a deadline, not as an afterthought. Flag any step where the timing is tight enough to risk the grapevine getting there first.

what you get back A sequenced runbook — leaders → managers → all-hands — with an owner, a time, and a dependency on every row, and the approval gates scheduled like any other step. The order is the plan: managers can never learn about their own team's change from the all-hands.

Schedule the approvals the way you schedule the announcements. At enterprise scale what slips a rollout isn't the copy — it's a works-council sign-off nobody secured or an exec who never saw the message.
Build the manager enablement pack

This is the single highest-leverage artifact. A blindsided manager can't hold the room; an enabled one can. Give every affected manager talking points, honest answers to the hard questions, and a clear can-say / can't-commit line — so they lead their own team's conversation instead of forwarding an email and hoping.

you ask

Draft a manager enablement pack for the managers of affected teams: the core narrative in their words, talking points for the team conversation, honest answers to the hardest questions people will ask (using [placeholders] where the specifics aren't mine to share yet), and an explicit can-say / can't-commit line so a manager knows exactly where the boundary is. Add a short 'if you don't know, say this' script for the questions we can't answer yet. Keep individual names out.

what you get back A pack a manager can actually hold their room with — narrative, talking points, hard-question answers, and a bright can-say / can't-commit line — so the change reaches each team from their own manager, in order, not from the grapevine or a cold all-hands.

Pressure-test this pack the hardest. The whole rollout's credibility rests on managers sounding informed and consistent in the room — a manager who's caught flat-footed turns one announcement into a hundred anxious side-conversations.
Draft the layered announcements

Now write the actual messages — the leadership note, the manager brief, and the all-hands message — all consistent with the one locked narrative so the change never says three different things in three rooms. Run each through the Draft a sensitive message you can stand behind craft; run any new or changed policy through Rewrite a policy into plain English so people understand the new rules; and if levels or roles change, reconcile the language against Build the competency framework every role inherits.

you ask

Draft the three layered announcements from the locked narrative, each consistent with the others: (1) a leadership note, (2) the manager brief, (3) the all-hands message. For each, follow our sensitive-message craft — honest about the hard part, kind without false cheer, no promise outside what's decided, names kept out, and [placeholders] for anything only I can confirm. Where the change introduces or revises a policy, rewrite that policy section into plain English so people actually understand the new rules. Where levels or roles change, keep the wording consistent with our competency framework. Flag any line that promises something not on the decided list.

what you get back Three consistent drafts — leadership note, manager brief, all-hands message — that tell one story across three audiences, each honest and free of accidental promises, with policy changes in plain English, role changes reconciled to the framework, and visible [placeholders] wherever you still owe a confirmed detail.

Consistency across the three messages is what protects trust — the affected team should never hear a fourth version. Read every announcement word-for-word, and route anything touching employment terms, roles, or comp through legal before it's anywhere near send.
Anticipate the questions and open a real feedback loop

A change that only broadcasts creates a hundred anxious DMs; a change that listens creates a conversation. Draft an honest FAQ that answers the obvious follow-ups, name a real channel with a real owner, and plan for the questions you can't yet answer — so people have somewhere to go that isn't the grapevine.

you ask

Draft an anticipated-questions FAQ for this change: the obvious follow-ups people will have, answered honestly — including the questions where the honest answer is 'not decided yet, here's when you'll know.' Then propose a feedback loop: a named channel, a named owner, an office-hours or Q&A slot on the timeline, and a plan for routing the questions we can't answer yet to someone who can. Keep it warm and specific — no corporate non-answers.

what you get back An honest FAQ that answers the real follow-ups (including the 'we don't know yet, here's when' ones) plus a concrete feedback loop — named channel, named owner, a Q&A slot on the runbook — so the change opens a conversation instead of a hundred anxious DMs.

The 'here's when you'll know' answers matter as much as the confident ones. People can sit with an honest unknown far better than with silence — silence is what the grapevine fills.

make it your own

Sensitive changes (RIF / restructure with departures): put legal and counsel on the runbook as named steps with their own deadlines, tighten every message with the Draft a sensitive message you can stand behind craft, and pair the rollout with Run a respectful exit and keep what you learn so the people leaving get the same care as the people staying. Keep every affected name out of any prompt that doesn't strictly need it.
Bilingual rollout (Arabic + English): run the whole comms suite twice — once per language, in parallel — and adapt, don't machine-translate, so the Arabic reads native rather than translated. Hold both languages to the same approval gate, and find an approver who actually reads Arabic so the second language never skips review. Run any policy change through Rewrite a policy into plain English in each language, not just English-then-MT — the new rules have to be as clear in Arabic as in English.
Reuse the runbook (Power Track): save the read-back → sequence → enablement → announce → feedback flow as a reusable /rollout custom command or skill (see the Playbook's Features tab) so every org change runs the same sequenced, gated process instead of being improvised under deadline. Custom commands and skills are the opt-in Power Track — on Desktop the whole arc runs fine as a shared set of saved prompts each rollout works through in order.

watch out for

Confidentiality before the announcement is the whole job, not a footnote. Embargoed news, named people, and unreleased details stay inside the workspace your company has approved for people data — and out of any prompt that doesn't strictly need them. Use roles and [placeholders]; the runbook and the drafts only need the shape of the change, not the private facts behind it.
Legal and works-council sign-off comes before the message ships, not after a complaint. Anything touching employment terms, roles, locations, or comp goes past HR, legal, and — regionally — the works council or labor authority before it goes out. Put those gates on the runbook with owners and deadlines so a missing sign-off can't quietly slip the whole rollout.
Sequencing is everything. A manager who learns about their own team's change from the all-hands is a trust failure you can't take back — so brief leaders first, managers before their teams, and the all-hands last. The grapevine is the enemy; the only way to beat it is to be first, in order, from the right person.
Claude drafts the runbook and the messages — but leadership, HR, and legal own the decision and the sign-off. A human reads every word before it sends, fills every placeholder with a confirmed fact, and takes responsibility for how the change lands. Claude gives you a shape and a sequence; it never owns the call or the consequences.

you'll end up with An org change rolled out like a launch instead of leaked like a rumor: one honest narrative every audience hears the same way, managers enabled before their teams, sequenced announcements from leaders down to all-hands, an honest anticipated-questions FAQ, and a real feedback loop with a named owner — so a hard change lands with the room, and the trust, intact, and the human owns every decision behind it.

Questions people ask

Does this replace HR, legal, or leadership judgment?: No. Claude drafts the runbook, the manager pack, and the announcements, but leadership, HR, and legal own the decision and the sign-off. Every word is read by a human before it sends, every placeholder is filled with a confirmed fact, and anything touching employment terms, roles, or comp is routed past legal and — regionally — the works council or labor authority before it ships. Claude gives you a shape and a sequence; the judgment stays human.
How do we keep a reorg confidential before we announce it?: Keep every affected name, reporting line, and comp detail inside a workspace your company has approved for sensitive people data, and out of any prompt that doesn't strictly need it — use roles and `[placeholders]` instead. The runbook and the drafts only need the shape of the change, not the private facts. And put the embargo and the approval gates on the timeline so nothing leaks early through a premature sign-off or a forwarded draft.
What's the single biggest mistake in org-change comms?: Sequencing — letting people, especially managers, hear it through the grapevine before the official, in-order rollout. A manager blindsided about their own team in front of that team is a trust failure you can't undo. The fix is the runbook: leaders briefed first, managers before their teams, all-hands last, with an owner and a time on every step so the right person delivers the news in the right order, ahead of the grapevine.
How do we handle Arabic and English without one feeling like an afterthought?: Run the whole comms suite twice — once per language, in parallel — and adapt rather than machine-translate, so the Arabic reads native instead of translated. Hold both languages to the same approval gate, and make sure an approver who actually reads Arabic reviews it, so the second language never skips the review the first one gets. Run policy changes through the plain-English rewrite in each language, not English-then-MT. See the bilingual-rollout variation.