Write Exec Updates at Three Audience Altitudes

Workflow area: Stakeholder Communication · Tier: Foundation · Hands-on time: ~25 minutes

Audience translation is mechanical work sitting on top of judgment work. What happened, what the numbers are, what's blocked — that's the shared substrate, and models re-render it per audience quickly and well. What it means — the ask, the risk framing, the "so what" — changes with altitude in ways only you know: the CEO needs the bet status, engineering needs the why behind the priority shift, sales needs what they can promise. The model does the rendering; you write the meaning, three times.

When to use — and when not to

• Use it for the recurring cadenceweekly or biweekly product updates that go to more than one audience; this is where the time savings are largest and the stakes most forgiving.
• Use it for launch and milestone commsone ship, three stories: exec summary, eng-org detail, sales-enablement version.
• Use it when updates keep getting ignoredupdates written at the wrong altitude are why; a CEO update with sprint-level detail trains the CEO to skim.
• Skip it for bad news that lands on peoplereorgs, cancellations, missed commitments affecting careers deserve fully hand-written words; recipients can tell, and it matters that they can.
• Skip the altitude cut when audiences genuinely need the same thinga five-person startup doesn't need three versions; don't manufacture ceremony.

The recipe

1. 1.

   Write the source-of-truth update once, flat and honest. Bullets, not prose: what shipped, what moved, what's blocked, key numbers with where each came from, decisions needed. Include the bad news — this document is internal raw material, and sanitizing it here means every altitude inherits the spin.

2. 2.

   Attach sources to every number. "Activation 34% (Amplitude dashboard, Aug 1)" not "activation up." The model will carry your numbers faithfully but it can't know if they're right — and it will helpfully compute derived figures ("up 21%") you must be able to trace. Unsourced numbers in, unauditable claims out.

3. 3.

   Render the three altitudes.

   Copy

   Render this product update for three audiences. Same facts, three
   altitudes — do not add facts, numbers, or commitments not present
   in the source.
   
   1. CEO/EXEC (max 150 words): outcome and bet status. Metrics vs
      goal, the one risk that matters, the decision I need. No
      feature lists.
   2. ENGINEERING ORG (max 300 words): what shipped and what's next,
      the why behind priority changes, dependencies and unblocks.
      Respectful of detail, no marketing tone.
   3. SALES/GTM (max 200 words): what's now sellable, what to say to
      customers, what NOT to promise yet (with dates only if the
      source states them).
   
   For each version, leave the final "So what" paragraph as the
   placeholder [SO WHAT — PM WRITES THIS] — do not fill it.
   
   Source update:
   [PASTE SOURCE-OF-TRUTH BULLETS]
   

4. 4.

   Check the renders for altitude leaks and invented certainty. Good CEO-level output looks like:

   Copy

   Bulk approvals shipped to all customers Aug 1 — the Q3 bet on
   admin efficiency. Early signal: median approval time 2.1 days →
   1.4 (Amplitude, first week). Risk: enterprise adoption gated on
   the SSO fix (ETA Aug 20, eng confirmed). Decision needed: hold
   the pricing change until adoption data lands, or ship it with
   the Sept release?
   [SO WHAT — PM WRITES THIS]
   

   Hunt for two failure classes: detail leaking upward (ticket IDs in the CEO cut) and hedges quietly dropped ("we believe" becoming "will"). Sales versions especially: check the model didn't turn "beta next month" into a promised date.

5. 5.

   Write three "so what" paragraphs by hand. This is deliberately unfilled. The CEO's is the bet and the ask. Engineering's is why the work matters and what you're protecting them from. Sales's is where the money is. Sixty seconds each — and it's the paragraph each audience actually reads.

6. 6.

   Send from your voice, in your channels. Final read-through per version: would you say this sentence out loud to this audience? Fix the two that fail, then send. You're accountable for every claim in all three — the altitude cut changes framing, never truth.

Make it yours

• Claude Project per audience set: store each audience's profile — what the CEO cares about this quarter, eng-org tone norms, sales's current plays — plus two past well-received updates as exemplars; renders arrive pre-tuned and the prompt shrinks to "render this week's."
• Notion AI in your update workspace: if updates live in Notion, keep source-of-truth and three renders as one page with toggles per audience — history stays browsable and next week's update starts from this week's structure.
• Gamma for the monthly leadership review: paste the exec render into Gamma to generate the slide version — same discipline applies: check every number survived the transformation to slides, where truncation invents claims.

Quality check

• Every number traces to the source docincluding derived percentages the model computed; recompute any derivation yourself before it ships.
• No commitments were born in translationdates, promises, and "will ship" statements exist in the source or they don't exist; sales versions get double scrutiny.
• Hedges survived"early signal" stayed "early signal" at every altitude; certainty inflation is the model's default failure under a word limit.
• The bad news is present in all threeif the blocked item appears in the eng version but vanished from the exec version, you've built an information asymmetry you'll pay for later.
• Each "so what" is yoursthree placeholders, three hand-written paragraphs; if any survived as model text, the judgment part didn't happen.