IT & Systems Administration · Fundamentals
Draft outage notices people actually read
When technology fails, people don't need a miniature root-cause analysis. They need to know whether the problem affects them, what to do now, and when they'll hear from IT again. They may be on a phone, in a meeting or up against a deadline, so your notice has to earn their attention quickly.
Communicate impact, not theories
AI can turn rough technician notes into a consistent set of messages. It must not invent restoration times, user impact, vendor statements or security conclusions to make a notice sound reassuring. You don't need an engineering diagnosis before sending a first notice when the impact is already evident.
A useful outage notice has five ingredients, in this order:
- What's affected: the service users see and the symptom they notice.
- Who's affected: confirmed users or sites, or "scope still being assessed."
- What to do: a safe workaround, or "please don't keep retrying" if repeated attempts make things worse.
- What IT is doing: one sentence, truthful and non-technical.
- When the next update comes: a specific clock time with a time zone, even if the problem may not be solved by then.
Don't announce a cause based on one symptom. "We're investigating a DNS issue" is too strong unless DNS is established as the cause. "We're investigating failed connections to the shared print queue" tells people what they need to know.
Create a single source of status
Instead of drafting each channel from memory, keep one short, verified status block:
| Field | Example (fictional incident) |
|---|---|
| Service | Shared office printing |
| Status | Investigating |
| User symptom | Jobs sent to Printer A stay in the queue |
| Confirmed scope | East office shared printer; other printers unconfirmed |
| First confirmed report | 09:07 ET |
| Workaround | Use Printer B if it's already installed; don't install new drivers without guidance |
| Responsible team | IT operations |
| Next update | 10:00 ET |
| Root cause | Not confirmed |
| Resolution estimate | Unknown |
Every Teams message, email, intranet banner and leadership update should come from this block. If the workaround changes, update the block first. That's how you avoid two contradictory messages circulating at once.
Send the first notice before you know the cause
A first notice should be short and calmly specific. It isn't a confession of uncertainty, and it isn't a promise to fix things within an arbitrary time.
Teams or chat:
Printing issue: East office (investigating)
We're investigating an issue with the shared printer in the East office. Print jobs may stay in the queue instead of completing.
Workaround: if Printer B is already available to you, use it for urgent printing. Please don't keep resubmitting the same job.
IT is working on it. Next update: 10:00 AM ET, even if nothing has changed.
That message works because it gives people something to do without making an unverified diagnosis.
Email. Subject: Service advisory: East office shared printing
We're investigating a problem affecting the East office shared printer. Some jobs are staying in the print queue instead of completing.
If Printer B is already installed for you, you can use it for urgent printing. Please avoid submitting the same stalled job more than once.
Our IT team is investigating. We don't yet have a confirmed restoration time. We'll send the next update by 10:00 AM ET.
Someone who reads only the subject and the first two lines should still understand what's happening.
Ask AI for consistent versions
You are drafting internal service-status communications. Use ONLY the
verified status block below. Do not invent a cause, user count, ETA,
workaround, vendor response, security conclusion or recovery claim.
AUDIENCE: [staff / leadership / affected team]
CHANNEL: [Teams / email / status banner]
TONE: calm, professional, plain English, not defensive
VERIFIED STATUS: [paste the status block]
PREVIOUS COMMUNICATION: [if any]
DELIVER:
1. A subject or heading that names the user-visible service.
2. A short message: impact, affected audience, verified workaround,
current action, and next update time with time zone.
3. A one-line status-page version.
4. A fact-check table mapping each claim to the supplied status.
5. A list of missing facts that MUST be obtained before mentioning them.
RULES: Do not write "resolved" until recovery has been independently
verified. Do not imply every user is impacted if scope is incomplete.
Do not promise an ETA unless I explicitly supplied an approved ETA.
Do not use jargon like DNS, OAuth, RADIUS, SSO or GPO unless users
need that detail to take the correct action.
This makes the AI a consistency checker as well as a writer. If it introduces a claim that isn't in your status block, remove it and correct the block rather than improvising a new fact.
Write progress updates that say something
An update is useful even when the system is still broken. People want to know whether anything has changed and when to check again.
10:00 ET, still investigating:
Update: East office shared printing
The printing issue is ongoing. Jobs sent to the East office shared printer may still stay queued. IT has narrowed the issue to the shared print service, but the cause isn't confirmed yet. The Printer B workaround remains available for anyone who already has it installed. Next update: 10:30 AM ET.
"Working diligently to resolve this matter as quickly as possible" adds no information. Avoid progress theater.
If the team misses a promised update, confidence drops even while engineers are working hard. If you can, give the communication job to someone other than the most deeply involved troubleshooter. Treat the time of the next message as a commitment, and don't confuse it with a commitment to restore service.
Separate mitigation from resolution
A workaround that restores partial function is mitigation. Resolution needs evidence that the service is working normally again. If you've temporarily disabled a feature, say so.
10:30 ET, mitigated but not resolved:
Update: East office shared printing (workaround expanded)
We've restored printing through the alternate queue for some of the affected users. The primary East office printer remains unavailable while IT keeps investigating. If you need help switching to the alternate queue, contact the help desk. Next update: 11:00 AM ET.
11:00 ET, verified restored:
Resolved: East office shared printing
Service to the East office shared printer was restored at about 10:52 AM ET. IT confirmed successful test jobs, and previously affected users have reported normal printing.
If a job is still stuck, cancel it and submit it once more. Contact the help desk if it still fails. We'll keep monitoring and record follow-up findings in the incident review.
Don't say "everyone can print again" if you tested one workstation. Use the narrower claim you can support, and give a clear escalation route.
Worked example: vague IT notes versus a real notice
Original notes: "Exchange authentication/token error. We restarted sync thing. A bunch of users can't do calendar. Need to check AAD logs. Vendor may be down."
Weak AI draft: "We are experiencing a major Microsoft 365 outage impacting all users due to authentication issues. Engineers have identified the problem and expect full restoration shortly."
That draft adds severity, scope, a diagnosis and an ETA that the notes don't support.
Repaired draft:
Calendar access issue: under investigation
We're investigating reports that the Teams desktop app can't display calendar information for some staff. We haven't confirmed how many people are affected. If you can open your calendar in Outlook on the web, you can use that in the meantime. IT is checking the service and sign-in information. Next update: 2:30 PM ET.
Verify the web workaround for the affected group before publishing it. If it hasn't been tested, write "If Outlook on the web is already working for you, keep using it" instead of promising it will work.
Adapt by audience without changing the facts
| Audience | What they usually need | Avoid |
|---|---|---|
| General staff | What broke, what works, what to do, when the next update is. | Diagnostic internals, blame, speculative ETAs. |
| Team leads | The affected workflow and practical alternatives. | Unverified counts of affected users. |
| Executives | Business impact, decisions needed, customer exposure, the next checkpoint. | Overstated certainty and technical theater. |
| Help desk | Exact current status, workaround prerequisites, ticket tagging and escalation. | Workarounds nobody approved. |
The facts stay the same for every audience. Only the level of detail changes.
What stays with you: what you're willing to promise, and what you're not ready to say yet.