CEO Communications
A message in a leader's name is judged twice. First on whether it says the right thing, and second on whether it sounds like the person whose name is on it. The second judgement happens faster and is harder to recover from. An audience that reads a paragraph and thinks "she would never say that" stops reading for content and starts reading for who actually wrote it, and from that point the message is a piece of evidence about the communications function rather than a message from the leader.
The costlier failure is quieter. Most communications from leaders answer the question the sender wanted to answer. The audience arrived with a different question, almost always some version of what this means for me, and if the first quarter of the message does not touch it they stop listening and fill the gap themselves. What they fill it with is worse than the truth, because the space between an announcement and its explanation is where a rumour gets its shape. A reorganisation announced without saying whose reporting line changes produces a week of speculation that costs more attention than the reorganisation itself.
Both failures are avoidable and both are avoided before the drafting starts.
When to use this, and when not to
Use it for anything delivered under a leader's name to an audience larger than a handful of people: all-hands scripts, company-wide announcements of reorganisations, leadership changes, strategy shifts, acquisitions or redundancies, internal memos that argue for a direction, messages after a public win or a visible setback, external talking points for an interview or a panel, and opinion pieces published under the leader's byline.
Do not use it for a live incident where people are affected right now and the sequencing is governed by harm and by statutory notification. That is crisis-and-incident-comms, which has a different first hour and a different order of stakeholders. Use this skill for the reflective message that comes afterwards.
Do not use it to make a decision. A communication that is really an unmade decision in narrative form reads as evasion, and the audience can tell. Send it back to decision-memo and draft once the decision exists.
Do not use it for the routine operating update, which is weekly-status-update, or for the quarterly note to investors, which is investor-update and has its own conventions. Do not use it to build the board's materials, which is board-deck. For a long-form public essay under the leader's name, the drafting craft belongs to newsletter-post-writer and this skill supplies only the voice note and the message decision.
What you need before starting
Three to five samples of the leader's unedited output. Emails they typed themselves, a transcript of them speaking without a script, a message sent from their phone. Polished pieces that a team already worked on are the wrong input, because they encode the team's voice rather than the leader's. Missing: ask for the sent folder rather than the published archive. Where nothing unedited exists, interview the leader for twenty minutes on the subject of the message, record it, and use the transcript as the sample. Never draft in a voice you have not sampled and then present it as calibrated; mark it uncalibrated in the delivery note.
The decision or the fact, settled. What is actually true, what has actually been decided, and by whom. Missing: stop and find out. The most common cause of a rewritten announcement is that the draft was commissioned while the decision was still moving.
The audience list, with what each part already knows. People hear a message against what they have already heard, and the same sentence lands differently for a team that has been waiting three weeks for news. Missing: ask the leader's direct reports what their teams currently believe. This takes an hour and reliably changes the draft.
The hard part, named explicitly. The fact or consequence the leader would prefer to soften. There is one in almost every consequential message. Missing: ask directly, in the form "what is the sentence you are hoping not to have to write". The answer is usually immediate, and it is usually the sentence the audience most needs.
The constraints on what can be said. Legal restrictions, market disclosure rules, employment law where individuals are affected, a customer contract, a deal under an agreement not to disclose. Missing: ask legal and human resources before drafting, not after. Redrafting around a constraint discovered at review costs more than the check.
The sequence and the timing. Who must hear before whom, what is already in a calendar invitation, whether an external announcement has a fixed hour. Missing: propose a sequence and get it confirmed. Managers before their teams, teams before the whole company, the company before the public, and each gap as short as it can be, because the gaps are where leaks live.
The channel and the length. Whether this is spoken, written, or spoken from a written script, and roughly how long. Missing: assume written, assume shorter than the requester expects, and state the assumption. A memo written for the eye and delivered aloud fails audibly.
The method
Build the voice note before writing a word. Read the samples and record, in one page: typical sentence length and whether the rhythm is short declaratives or longer reasoning; words the leader uses and words they demonstrably never use; their own names for teams, products and priorities; how they open and how they close, including whether they thank people, whether they joke, and whether they reach for numbers; what they emphasise when they have a choice, which is usually one of customers, craft, speed, people or mission; and how their register shifts between internal and external. Keep the note; it is reusable and it improves with each message.
Write the message decision, in three lines, before drafting. What the audience must know, what they must feel, and what they must do. If the second line is hard to write, the message is an information transfer and should be shorter than planned. If the third is empty, say so explicitly rather than inventing a call to action, because a manufactured action is the clearest possible signal that the message had no purpose.
Name the audience's first question and decide where it gets answered. For an announcement it is almost always what this means for me, and the rule is that it is answered inside the first quarter of the length. Not hinted at, answered. Where the honest answer is that nothing changes for most people, say that in those words, early, because it is the most reassuring sentence available and it is almost always omitted.
Decide consciously what is not being said, and why. List it. Legal, timing, not yet decided, or someone else's news to tell. Then decide for each whether to acknowledge the omission. The rule: where the audience will notice the gap, name the gap and say when it closes. Where they will not, stay silent. An unacknowledged obvious omission is read as concealment, and the reader supplies the worst available explanation.
State the hard part plainly, early, in one sentence. Then the reasoning, then the support available. Softened bad news does not stay softened; it produces a second communication a week later that has to be harder than the first would have been, and by then the leader has spent the credibility that would have carried it.
Draft in the voice note, not in the communications register. Every sentence should be one the leader could say without stumbling. The specific test: read it aloud in their cadence. Words that trip you will trip them, and a leader stumbling over their own words in front of the company is the visible form of this failure.
Replace every generality with a name, a number or a date. "Significant progress" becomes the figure. "The team" becomes the three people. "Soon" becomes the date, or an honest statement that there is no date yet and when there will be. Generalities in a leader's message are read as evasion even when they are only laziness, and the reader cannot tell which.
Write the question and answer sheet, starting with the questions the leader hopes nobody asks. Those are the ones that will be asked, either in the room or in the private channel afterwards. Each gets a real answer, prepared, in the leader's voice, including the answer that is "I do not know yet, and here is when I will". An unprepared honest answer usually comes out worse than a prepared one.
Write the sequencing plan as a table with times. Who hears, from whom, through what channel, at what hour, and what they are given to say. Managers receive their briefing before the company message lands, not simultaneously, because a manager who learns alongside their team cannot answer for it. The rule for the gap: long enough to read, short enough that nobody has time to forward it.
Have the leader read it aloud before it goes. This is the single highest-yield step and it takes ten minutes. Anything they stumble on, hesitate over, or instinctively reword is rewritten in the words they reached for. Their instinctive rewording is the voice note working, and it should be captured back into the note.
Formats, and what each one must carry
All-hands script. Open on something concrete and recent: a customer, a number, a person by name. Then the state of the business in three to five points, each with a figure. Then the one theme the leader wants remembered, stated once and returned to at the close. Then what changes for the people in the room. Then recognition, specific to work rather than to effort. Then a close that comes back to the opening. Ten to fifteen minutes of script, written for the ear, with pauses marked, plus a question and answer sheet. Nothing that reads a slide aloud.
Company announcement. The news in the first sentence, with no preamble. The reason in the second paragraph. What it means for the reader in the third, with the reporting-line consequences stated rather than implied. What happens next, with dates. Where to ask questions, with a named person. Then one honest paragraph in the leader's own stance. Coordinated with human resources, legal, and the managers who will field the questions, all of whom are briefed first.
Internal memo on strategy. The situation, the decision, the reasoning including the alternatives considered and rejected, what the organisation will do and what it will stop doing, and how progress will be visible. Written in prose. A memo that is bullets is a slide, and a slide cannot carry reasoning.
Message after a setback. What happened, stated fully. What the leader takes responsibility for, in the first person and without conditionals. What changes. What does not change. Short, and sent quickly; a careful message on day four is worth less than a plain one on day one.
Message after a win. The specific people and the specific work. The customer or the outcome. One sentence on what it proves about how the organisation works. The next thing. Not longer than the setback message, because a leader who writes at length about wins and briefly about failures teaches the organisation exactly that ratio.
External talking points. Three messages, one proof point each. The bridge from each likely question back to those messages. Written-out answers to the hard questions. The list of things not to say, with the reason. One page, because anything longer will not be read in the car.
Worked example
Situation. A 240-person logistics software company merging two product organisations into one after eighteen months of duplicated roadmaps. Nineteen people would change reporting line, two director-level roles would be removed with those directors offered different roles, and no redundancies were planned. The chief executive, who had founded the company and communicated in short, blunt, number-heavy emails, was three days from an all-hands. The communications draft prepared by the marketing team opened with a paragraph about the company's exciting next chapter and did not mention reporting lines until the ninth paragraph.
Task. A script and an announcement email that the chief executive would actually deliver, that answered the reporting-line question before people invented an answer, and that did not require a second clarifying message. Good meant no all-company speculation thread within twenty-four hours.
Action. The voice note took ninety minutes on eleven of the chief executive's own sent emails. It found sentences averaging fourteen words, near-total absence of adjectives, a habit of opening with a number, and a specific tell: he never used the word "team" in the singular, always naming the actual group. He also never wrote "excited". The marketing draft used it twice in the first three lines.
The first rewrite led with the strategic reasoning and put the people consequences in the second half. It was abandoned after the audience-belief check. Two of the four directors reported that their teams already believed a reorganisation was coming and that the prevailing theory involved redundancies. Against that belief, three paragraphs of reasoning before any mention of jobs would read as three paragraphs of stalling. The order inverted: the news, then the reporting lines, then the explicit sentence that there were no redundancies, then the reasoning.
The hard part was identified by asking the chief executive what sentence he was hoping to avoid writing. The answer was that the two directors losing their roles had done nothing wrong and he did not want to say it in a way that implied otherwise. That produced the sentence that went in, close to verbatim: "Both of these roles are going because we built two of something we needed one of. That was my decision two years ago, not theirs."
The sequencing plan ran across three hours on one day. The two affected directors were told in person the previous afternoon. The nineteen people changing reporting line heard from their current manager at 09:00, with a one-paragraph brief the managers had received at 08:15. The company email went at 10:00. The all-hands ran at 11:00. Customers with a named account contact were told at 14:00 by their contact, using three prepared lines.
The question and answer sheet ran to eleven questions. Four came from asking the chief executive what he hoped nobody would ask. Two of those were asked in the room.
Result. The all-hands ran thirty-eight minutes with twenty-two of them in questions, which was longer on questions than any previous all-hands at the company and was treated as a good sign rather than a bad one. No all-company speculation thread appeared. One director-level conversation went badly and required a follow-up the next week, which the plan had not anticipated.
The measurable part: no clarifying message was needed, against a history of two of the previous three announcements requiring one. The honest limit of the claim is that the situation was benign, and no redundancies made this message far easier than most reorganisation messages are.
A second scenario, where it goes differently
The same company, eight months later, closing a regional office with fourteen redundancies. The method holds but three things change.
The constraint moves to the front. Employment law in the affected jurisdiction required a consultation period, which meant the announcement could not state outcomes as settled and had to be written in the language of a proposal without sounding evasive. That is a genuinely hard drafting problem, and the resolution was to be explicit about the process itself: what a consultation is, how long it takes, and what is genuinely not decided, in plain words rather than legal ones.
The sequence lengthens and inverts. Affected individuals hear first and individually, in conversations that are not scripts, before any group message exists. Nothing goes to the wider company until every affected person has been spoken to, which sets the timetable rather than the calendar doing it.
And the voice calibration matters in a different direction. The chief executive's blunt register, which served the reorganisation message well, reads as cold when people are losing jobs. The voice note is used to find the register he uses when the news is personal, which the samples showed was slower, longer-sentenced and more first-person. Voice calibration is not one setting but a range, and the samples have to cover it.
Output
Five artefacts, delivered together.
Voice note, one page, reused across messages:
| Attribute | Observed | Evidence | | Sentence length and rhythm | 14 words average, short declaratives | 11 sent emails, Jan to Mar | | Words used | numbers first, "the group", "shipped" | as above | | Words never used | "excited", "journey", "team" singular | absent across all samples | | Opening habit | opens on a figure | 8 of 11 | | Register shift, internal to external | more formal, longer sentences externally | 2 panel transcripts |
Message decision, before drafting:
KNOW [one line]
FEEL [one line]
DO [one line, or: nothing, and why]
FIRST QUESTION [the audience's, and where in the draft it is answered]
THE HARD PART [the sentence the leader would rather not write]
NOT BEING SAID [item, reason, whether the omission is acknowledged]
CONSTRAINTS [legal, HR, disclosure, contractual]
The draft, in the format's own shape, with pause marks and timing where it is spoken.
Question and answer sheet:
| Question | Likelihood | Answer, in the leader's voice | Who follows up |
Ordered with the questions the leader hopes nobody asks at the top.
Sequencing plan:
| Time | Audience | From whom | Channel | What they are given |
Failure modes
The communications-team voice. Recognise it by the vocabulary: "excited to announce", "journey", "double down", "north star", any of which will be in the draft and none of which will be in the samples. Fix by checking the draft's distinctive words against the never-used list in the voice note. This is a mechanical check and it catches most of it.
The buried consequence. Recognise it when the paragraph that says whose job changes sits below the fold. Fix by moving it up and accepting that the reasoning reads better after the consequence than before it, which is counter to how the writer wants to argue and is what the audience needs.
Softened bad news. Recognise it by conditionals and passives clustering in one paragraph: "some roles may be impacted". Fix by writing the sentence in the active voice with a number in it, then deciding whether it is true. If it is not true, the problem is the decision, not the draft.
The manufactured call to action. Recognise it when the last line asks people to bring their questions, stay focused, or keep doing what they are doing. Cut it. Ending on the real last substantive sentence is stronger than ending on a request nobody will act on.
Simultaneous release to managers and their teams. Recognise it in the sequencing plan when two rows share a time. Fix by inserting a gap of at least thirty minutes and giving managers something to say, not just something to read.
The message that is an undecided decision. Recognise it when the draft cannot say what happens next with a date, and the reason is not legal. Stop drafting.
Voice calibrated to one register. Recognise it when a message about people uses the leader's operational voice. Fix by sampling how they write when the subject is personal, which is a different set of samples and usually a different rhythm.
Edge cases
No samples exist, and the leader is new. Interview them for twenty minutes on the subject, transcribe it, and draft from the transcript. Deliver the draft with the sentence that it is calibrated to one conversation rather than to a body of writing, and expect a heavier edit. Capture their edits into the first voice note.
The leader edits the draft into their voice and out of the message. This is common and it is not a problem to solve by resisting. Accept the voice edits without argument and defend only the message decision: the first question still answered early, the hard part still plain. Argue over structure, never over wording.
The news leaks before the sequence completes. Move immediately to the shortest true statement, sent to everyone at once, acknowledging that people have heard this in a way that was not intended. Do not continue a sequence that has been overtaken, and do not pretend it held.
A regulator or a market disclosure rule sets the timing. The disclosure obligation governs and everything else sequences around it, which usually means the internal message goes after rather than before. Say so internally in one line, because the reversal of the normal order is itself noticed.
The leader will not say the hard part. Write both versions and show them together, with the second communication that the soft version will require attached. Where they still decline, draft the soft version, and record the decision so that the follow-up is planned rather than improvised.
The message is genuinely small. Where the audience's first question is trivially answered and there is no hard part, write three sentences in the leader's voice and stop. Applying the full sequence to a small message trains people to skim the important ones.
Quality bar
- A colleague who knows the leader well would believe the leader wrote it, and could not point to a sentence they would not say.
- The audience's first question is answered inside the first quarter of the length.
- The hard part is stated plainly, in the active voice, above the reasoning.
- Every generality has been replaced by a name, a number or a date, or is explicitly labelled as not yet known with a date for when it will be.
- The question and answer sheet contains the questions the leader hoped would not be asked.
- The sequencing plan has times, and no two audiences that depend on each other share one.
- The leader has read it aloud, and their instinctive rewordings are in the final text.
- Nothing in the draft is a decision that has not actually been made.
Adapting this to your context
The defaults come from internal announcements at commercial companies of a hundred to three hundred people in one jurisdiction. They travel less well than they look.
- Three to five unedited samples. Assumes a leader who writes their own email. Where they dictate, speak more than they write, or work in a second language, sample transcripts instead and expect the voice note to describe cadence rather than vocabulary.
- The sequencing gaps. Managers, then teams, then the company, then the public, each gap short. That assumes every manager is reachable within the hour. Shift-based and multi-time-zone organisations set the gap by when the last shift can be briefed, which may be a day.
- The all-hands length. Ten to fifteen minutes of script came from a 240-person company. At forty people the same message is three paragraphs; above a thousand, questions must be collected in writing or two confident people take the session.
- Consultation language. The redundancy scenario assumes a jurisdiction with a statutory consultation period. Where none exists the drafting problem inverts: outcomes can be stated, so the discipline becomes not stating them before the affected people have been told.
- What not to change. The audience's first question is answered inside the first quarter of the length, and the hard part is stated plainly and early.
Related skills
principal-simulator predicts how the leader will react to a draft and shortens the edit cycle, and is worth running before the first version goes to them. decision-memo produces the settled decision this skill announces; drafting before it exists is the most common way this work is wasted. crisis-and-incident-comms takes over when something has gone wrong and people are affected now, and hands back to this skill for the reflective message afterwards. meeting-to-decisions supplies what was actually agreed when the message follows a leadership discussion. investor-update and board-and-investor-management carry the same news to investors on their own conventions. board-deck builds the board's version. newsletter-post-writer and linkedin-post-writer handle long-form and social pieces under the leader's byline, using the voice note this skill produces. human-voice-editor strips the generic register from a draft that has acquired one.