Muse for IT Admins: Ticket Triage Drafts, Outage Comms, and Asset Tracking That Stay Current
An IT admin setup for Muse: a morning triage board that separates category from priority, outage updates drafted from confirmed facts only, and a license…

IT admin work fails in the gaps between systems. A ticket arrives with a calm subject line and a company-wide cause. An outage update goes out late because nobody wants to write it while fixing the problem. A license renewal arrives with a count nobody can explain. Muse does the sorting, drafting, and reconciling for those jobs, and a human decides what gets sent, closed, or bought.
This setup runs three jobs. Muse drafts the morning triage board from the tickets you paste in, drafts outage updates only from facts you mark as confirmed, and keeps an asset and license board that compares what you bought, what is installed, and what people actually use. You keep approval on every external message, every priority change on a security ticket, and every purchase.
The priority rule underneath the triage work comes from ITIL practice. Category and priority are different things: category says what the ticket is about, for example account access or email, and priority comes from impact and urgency together. Impact is how broadly work is affected, and urgency is how quickly harm grows if the ticket waits. A single user who cannot print is low impact, while a department locked out of a shared system is high impact. A ticket about email can be low or high priority depending on who is affected and whether a workaround exists.
The short version
- Keep category, impact, urgency, and priority separate. Category names the service, impact and urgency set priority, and a calm tone never lowers a security ticket.
- Draft outage updates from confirmed facts only. Atlassian guidance is to use a status page as the source of truth, never go more than one hour without an update for a customer-facing issue, and always say when the next update will arrive.
- Reconcile licenses in three columns, entitlements, installs, and usage. A gap in one direction is a compliance risk, and a gap in the other is money you can reclaim before renewal.
- Update runbooks in the same session as the incident that proved them wrong. A runbook header needs the service, the trigger, the owner, and the last-tested date.
- The failure mode to watch for is a polished draft that states a cause you have not confirmed yet, and the fix is a confirmed-facts list Muse is not allowed to go beyond.
The Setup: The Service Desk Board
Call the whole arrangement the Service Desk Board. The example I run it against is a 140-person company with a three-person IT team, one shared inbox and help desk queue, a mix of laptops and cloud apps, and no dedicated communications person during an outage. Their inputs are ordinary: the overnight tickets exported as text, the current asset list, the license spreadsheet with purchase records, and a folder of runbooks that were accurate when they were written.
The board rests on a few pieces kept together. A standing instruction from our memory guide holds the durable facts: the services you support, the priority definitions your team agreed on, the escalation contacts by service, which systems are customer-facing, and the rule that security, access, and data tickets always get a human review before any priority is lowered. Each ticket brief follows the same discipline as a task brief written for an agent: what happened, who is affected, what was already tried, and what a finished triage note looks like. The recurring checks run as a scheduled routine, a weekday morning sweep that prepares the board before standup rather than during it.
The board borrows its queue discipline from the customer support lesson, where Muse drafts and a human sends. IT work needs the same line, with one addition. A support draft that is wrong costs a correction. An outage draft that states the wrong cause, or a triage note that buries a security ticket, costs trust and time you do not get back. The prompts below separate drafting from deciding in every step. The ops manager pattern adds one place that checks the other jobs actually ran and that their outputs match.
Prompt 1: Draft the Morning Triage Board, Category First, Then Priority
Run this at the start of the day, with the new tickets pasted in full. Do not paste only subject lines. The body is where impact usually hides, in a line like “the whole warehouse team is using one login now” under a subject that just says “password reset.”
Use the four statuses and the priority levels your team already uses, and write them into the prompt so Muse does not invent its own. In the example board, P1 is high impact and high urgency with no workaround, P2 is high impact or high urgency with a workaround, P3 is one user blocked with a workaround, and P4 is low impact and low urgency. Replace those example definitions with your own policy.
Draft the morning triage board from the tickets below. For each ticket return: (1) Category: name the service and issue type only, for example account access and sign-in, email and outbound sending, device and laptop, or network and VPN. (2) Impact: high if a department, a shared system, or a customer-facing service is affected, medium if a team is affected and a workaround exists, low if one user is affected. Use only who and what the ticket says is affected. (3) Urgency: high if work has stopped or a deadline or security risk is stated, medium if work is slowed, low if it can wait to the next business day. (4) Priority: apply this matrix only, high impact and high urgency is P1, high impact or high urgency is P2, medium impact and medium urgency is P3, low impact and low urgency is P4. If impact or urgency cannot be judged from the ticket, write unknown and list what to ask the requester. (5) Draft reply: two to four sentences that acknowledge the problem as described, state the priority you assigned, and ask for any missing fact you listed. Rules: do not lower the priority of any ticket mentioning a possible security issue, lost device, or shared credentials, flag those for human review at the top of the board regardless of tone. Do not close, merge, or reassign any ticket. Do not add symptoms or causes that are not in the ticket.
Priority definitions and escalation contacts: [paste yours]. Tickets, full text: [paste them].
Read the flagged list first, not the P1 list. The most useful output in this prompt is the unknown list, because it shows which tickets your intake form failed to capture. If the same missing fact appears three mornings in a row, fix the form that week. A triage board that keeps asking for the affected team name is telling you the form should require it.
Prompt 2: Draft Outage Updates From Confirmed Facts Only
Run this when an incident is declared, and again at each update point until it is resolved. Prepare the channels and templates before an incident, not during one. Atlassian lists the main channels as a dedicated status page, embedded status, email, workplace chat, social media, and SMS, and recommends the status page as the primary source of truth that the other channels point back to. Their incident statuses are Investigating, Identified, Monitoring, and Resolved, and each status changes what you are allowed to claim. Investigating means you have been alerted and are looking. Identified means you have found the problem. You cannot write Identified language while you are still in Investigating.
NIST SP 800-61 Revision 3, published in April 2025, organizes incident response around the six Cybersecurity Framework 2.0 functions, Govern, Identify, Protect, Detect, Respond, and Recover. Preparation sits in the first three functions and the response itself in the last three. A small IT team needs the same preparation, with the contacts, templates, and severity definitions ready before an incident, so the response work is not spent inventing them.
Draft an outage update from the confirmed facts below. Return three versions from the same facts: (1) Status page: a plain title that names the affected service and the symptom users see, then three to five sentences covering what is affected, who is affected, what the team is doing now, and the time of the next update. Use the status I give you, Investigating, Identified, Monitoring, or Resolved, and do not use language from a later status. (2) Internal chat: the same facts in two sentences, plus the incident lead name and where staff should direct questions instead of filing new tickets. (3) Staff email: the status page text plus one sentence on any workaround that is confirmed below. Rules: use only the confirmed facts below. If a fact is not listed, do not state it. Write unknown for the cause unless the cause is listed under confirmed facts. Do not give a resolution time unless one is listed as confirmed. Always end the status page version with the exact time of the next update, which I will supply.
Status: [Investigating / Identified / Monitoring / Resolved]. Affected service and symptom: [write them]. Who is affected: [write it]. Confirmed facts: [list them, one per line]. Confirmed workaround, or none: [write it]. Incident lead: [name]. Time of next update: [time with timezone].
Keep the cadence even when there is nothing new. Atlassian recommends never going more than one hour without an update for an issue that still affects customers, and always indicating when the next update will come. A short update that says the team is still working and names the next update time prevents the duplicate tickets and side messages that slow the fix. Send the update from the template, then go back to the incident. The person fixing the problem should not also be the person rewriting prose, which is why this prompt separates the draft from the lead who approves it.
Prompt 3: Reconcile Assets and Licenses, Then Flag the Runbooks the Week Proved Wrong
Run this monthly for the license position and weekly for anything expiring in the next 90 days. Software asset work compares three numbers that are often kept in three places: entitlements, what you bought, installs, what is actually deployed, and usage, what people opened and used. Reconciliation maps those three together. Installs above entitlements mean compliance risk. Entitlements above active usage mean budget you can reclaim or reassign before you buy more.
Record each entitlement with its evidence, the product, the license metric, the quantity, the term, and the order or agreement number, with the contract or certificate kept with it. Set renewal work on a 90-, 60-, 30-day pattern. At 90 days flag the renewal and start the usage review, at 60 days decide whether to renew, reduce, or replace, and at 30 days confirm the order so an auto-renewal does not make the decision for you.
Reconcile the asset and license board from the material below. (1) License position: for each product, list entitlements, installs, and active users from the lists below, then the gap. Flag installs above entitlements as compliance risk, and entitlements above active users as reclaim candidates, with the number of seats in each case. If any of the three numbers is missing for a product, write unknown and do not estimate it. (2) Renewals: list every term ending in the next 90 days, grouped as 90 to 61 days, 60 to 31 days, and 30 days or less, with the owner and the next step for that group. (3) Assets: list laptops and devices with no assigned user, no last-seen date in the material, or a warranty or age flag as written below. Do not assign a device to a person who is not listed as its user. (4) Runbook updates: from the incident notes below, list each runbook that was used, each step that did not match what actually happened, and a draft replacement step based only on what the incident notes say was done. Flag any runbook with no owner or no last-tested date in the list below. Rules: use only the numbers, dates, and notes below. Do not invent a usage count, a renewal price, or a fix step that is not in the notes.
Entitlements with order numbers and terms: [paste them]. Install list: [paste it]. Usage list with last-used dates: [paste it]. Asset list: [paste it]. Runbook list with owner and last-tested date: [paste it]. Incident notes from this period: [paste them].
The runbook part of this prompt is where the board earns its keep after an outage. A good runbook header carries the service, the specific trigger or alert that sends someone there, the severity it usually maps to, the owner, and the last-tested date. Google’s Site Reliability Engineering book reports that writing best practices down ahead of time in a playbook produced roughly a threefold improvement in mean time to recovery compared with improvising. That gain disappears when the steps no longer match the system. Updating the runbook in the same week as the incident, from the incident notes, is how the nightly audit pattern applies to IT documentation: a small recurring check beats an annual rewrite from memory.
| Job | When | What it answers |
|---|---|---|
| Triage board draft | Weekday morning, before standup | What is each ticket about, who is affected, and what priority follows from impact and urgency? |
| Outage update draft | At declaration, then at least hourly until resolved | What do users need to know now, from confirmed facts only, and when is the next update? |
| License position | Monthly, and before every renewal | Where do installs exceed entitlements, and where do entitlements exceed real usage? |
| Runbook update list | Weekly, from that period’s incident notes | Which steps proved wrong, who owns each runbook, and when was it last tested? |
What Can Go Wrong: A Confident Cause in the First Update
In the example setup, a Monday morning outage slowed sign-in for the whole company. The first tickets mentioned a vendor status page that listed an authentication issue in the region. Muse drafted a status page update that stated the vendor outage as the cause, in Identified language, because the vendor link was in the ticket text. The team posted it. Forty minutes later the vendor cleared its incident and sign-in was still slow, because the local cause was an expired certificate on an internal proxy that nobody had checked yet. The correction took longer to write than the original update, and staff who had read the first message stopped trusting the second one.
The draft was fluent and the fact was in the source material, but the error was treating a related fact as a confirmed cause. A vendor incident in the same area is a lead to check, not a cause to publish, until your own lead confirms the link.
The fix is the confirmed-facts rule in Prompt 2, plus a status gate you do not waive. Keep two lists during an incident, confirmed and suspected, and paste only the confirmed list into the prompt. Require the incident lead to move an item from suspected to confirmed in writing before it can appear in an external update. Hold the status at Investigating until that move happens, even if a plausible cause is circulating in chat. The update will read as less certain for the first hour, because it is less certain. That honesty is what makes the later Identified and Resolved updates believable.
Start With the Triage Board Only
Do not rebuild the asset list, the runbooks, and the outage templates in the same week. Start with Prompt 1 on Monday, with your own priority definitions pasted in and the security flag rule in place, and run it for five mornings. If the unknown list shrinks because you fixed the intake form, and the flagged security tickets reach a human before the rest of the board is discussed, the setup is working. Add the outage prompt before your next planned maintenance window, when you can test the template without a real incident, and add the license reconciliation at the start of next month.
Keep the sorting and the drafting with Muse, and keep the sending, the closing, and the buying with a named person on your team.
Priority definitions follow ITIL practice as summarized in current triage guides: category names the service or issue type, impact measures how broadly work is affected, urgency measures how quickly action is needed, and priority is derived from impact and urgency together, commonly as P1 to P4. Outage communication channels, the status page as source of truth, the four statuses Investigating, Identified, Monitoring, and Resolved, and the guidance to update at least hourly and always state the next update time follow Atlassian, Incident Communication Best Practices and its Statuspage incident communication tutorial. Incident response organization follows NIST SP 800-61 Revision 3, April 2025, which supersedes Revision 2 and maps incident response to the six NIST Cybersecurity Framework 2.0 functions, Govern, Identify, Protect, Detect, Respond, and Recover. The playbook and mean time to recovery figure follows Google, Site Reliability Engineering, as cited in runbook guidance summarizing the book: written playbooks produced roughly a 3x improvement compared with improvising. License reconciliation, recording entitlements with order evidence, calculating the license position regularly, and the 90-, 60-, 30-day renewal pattern follow current software asset management guidance on comparing entitlements, installs, and usage.


