Skip to content

Handle a Sensitive or Crisis Message

When to use this

Some messages can't be taken back. A layoff note, a response to a breach, an apology after something went wrong. One careless line here does real, lasting damage. So this is the page where you slow down on purpose.

Use it when the message is sensitive, the audience is upset or anxious, and the stakes are high. The AI is a sounding board, not the author. You own every word, and you own the decision to send.

This isn't a single prompt. It's a slow loop: get clear, draft, test how it reads, revise, then check who else must see it first.

A prompt to start

Open the loop with something like this. Then keep the conversation going — don't rush to a final draft.

I'm preparing a sensitive message and I want to get it right, not get it fast.

Here's the situation: [WHAT HAPPENED, IN PLAIN TERMS]
Who's receiving this: [AUDIENCE — and how they likely feel right now]
What I need them to understand: [THE ONE OR TWO THINGS THAT MATTER MOST]
What I can't say yet, or shouldn't: [ANYTHING LEGAL, UNCONFIRMED, OR PRIVATE]

Before we draft anything, ask me the questions you'd need to write this well.
Don't soften the situation. Don't reassure where reassurance isn't earned.

Work through it step by step

  1. Get clear on the facts and the audience. Answer its questions honestly. Be sure of what's true before you shape how to say it. Know who's reading and what they fear. A message that ignores the reader's worry won't land.

  2. Draft, slowly. Ask for a first draft, then read it as the recipient. Does it acknowledge what happened? Or does it dodge? A direct, honest opening beats a warm one that buries the point.

  3. Test how each line could be read — and misread. Go line by line. Ask: "How might an anxious or angry reader take this?" Hunt for any sentence that sounds evasive, cold, or like it's hiding something.

  4. Revise for honesty, not comfort. Cut the spin. If something needs a plain acknowledgement, give it one. Don't let the draft smooth over a hard truth just because that feels safer — it rarely is.

  5. Decide who else must see it before it goes. Most sensitive messages aren't yours alone to send. Loop in comms, HR, or legal. Ask yourself who gets blindsided if this goes out — and bring them in first.

  6. Sit with it, then decide. Read it once more, out loud if you can. If a line still nags at you, fix it or hold the message. Send only when you'd stand behind every word.

What stays your call

Every word is yours. So is the timing, and the choice to send at all.

The AI can help you see how a line lands and where you're being vague. It can't know what's actually true, what you can legally say, or what your people need to hear from you directly. It has no stake in the outcome — you do.

Use it to pressure-test and to slow down. Never let it make the call.

Watch out for

  • Keep sensitive specifics out of a tool you don't control. Names, legal exposure, unconfirmed details, anything private — don't paste those into a chat tool. Work in general terms, or take it offline.
  • For a real crisis, involve comms, legal, or counsel. This workflow helps you think. It doesn't replace the people whose job is to protect the company and the people in it.
  • AI tends to smooth and reassure. It will round off hard edges and offer comfort by default. Don't let it paper over something that needs a direct, honest acknowledgement. Sometimes the right message is the harder one.