Muse for Launch Ops: Launch Checklists, Announcement Drafts, and Cross-Team Updates That Hold

A launch ops setup for Muse: a checklist with a go or no-go gate 24 hours out, announcements drafted from approved facts only, and a retro that ends with owned actions.

Jolly the cream-furred creature at a launch control desk in a warm library room, with a checklist board, a countdown clock, and announcement drafts on the desk, soft-3D, warm cinematic lighting, no text

Launch work fails in the handoffs between teams. Engineering ships on time, marketing announces to everyone on day one, support meets a feature it has never seen, and nobody can say who was allowed to call a pause. Muse does the assembling work that keeps those pieces in sync, turning the plan, the readiness evidence, and the approved facts into a checklist, an announcement set, and status updates a human reviews and sends.

This setup runs three jobs. Muse builds the launch checklist from your template and the evidence each owner supplies, and checks it at a go or no-go gate 24 hours before launch. Muse drafts every announcement version from one approved facts list, so the blog post, the email, the in-product note, and the support reply cannot drift apart. Muse then turns the launch-day log and the first results into cross-team status updates and a post-launch retro draft. You keep approval on the go call, on every external message, and on any rollback or pause decision.

Atlassian splits a launch into three phases, pre-launch, launch, and post-launch, and puts 70 to 80 percent of the timeline in pre-launch. That split is the working rule here. Most of what Muse does in this setup happens before launch day, when a missing help article or an unnamed rollback owner is still cheap to fix.

The short version

  • Run one checklist with an owner, a status, and evidence for every item. Productboard groups readiness into product and engineering, analytics, documentation, marketing, sales enablement, customer success, legal, launch day operations, and post-launch, and that grouping works for most launches.
  • Hold a real go or no-go gate 24 hours before launch. Every item is done or not applicable, exceptions need a named sign-off, and the rollback owner and trigger conditions are written down before the gate.
  • Draft every announcement from one approved facts list. The list states what shipped, who gets it and when, the exact rollout percentage, and what is excluded. No version may add a claim that is not on it.
  • Keep one status thread on launch day. Updates go out on a set cadence from the launch log, the stakeholder update lands within 48 hours, and the formal review happens one to two weeks after launch.
  • The failure mode to watch for is an announcement that describes the plan instead of the shipped state, and the fix is a facts list the launch lead signs after engineering confirms what is live.

The Setup: The Launch Board

Call the whole arrangement the Launch Board. The example I run it against is a 22-person software team launching a new self-serve plan with a pricing page change, with one product manager as launch lead, two engineers, one marketer, one support lead, and a phased rollout that starts at 10 percent of accounts. Their inputs are ordinary: a launch plan, a readiness checklist from the last launch, staging test results, a draft help article, and a shared launch log that starts on launch morning.

The board keeps a few pieces together. A standing instruction from our memory guide holds the durable facts for each launch: the launch lead and the rollback decision owner by name, the channels where status updates go, the success metric and its baseline, and the rule that no external message goes out without the launch lead’s approval. Each launch brief follows the same discipline as a task brief written for an agent: what is launching, who gets it first, what done looks like, and what would trigger a pause. The recurring checks run as a scheduled routine, a daily readiness sweep in the final week that lists blocked items by owner so a person decides who to chase, rather than Muse messaging owners on its own.

Jolly the cream-furred creature at a desk with a product roadmap and sticky notes Muse Academy Muse for Product Managers

The board borrows its run-of-show discipline from the event planners lesson, where the day runs from a timed sheet with a named owner for each segment and a central coordinator who tracks progress. A launch day needs the same shape, an hour-by-hour runbook and one coordinator, which is the pattern Atlassian recommends for launch day itself. It also borrows from the meetings lesson, where the brief goes out before the call and the recap goes out after it, and from the ops manager pattern, where one place checks that the other jobs actually ran. The prompts below keep drafting with Muse and deciding with a named person in every step.

Prompt 1: Build the Checklist and Hold the Go or No-Go Gate

Run this when the launch plan is first written, then run the gate version 24 hours before launch. Productboard’s readiness checklist covers the same ground in nine groups. It includes acceptance criteria passing in staging, a feature flag tested on and off, monitoring and alerting set up, and a rollback plan documented and tested, plus analytics events firing and a dashboard built for the primary success metric, a help article published, release notes written, support briefed with an escalation path, and legal sign-off where the launch needs it. Google’s Launch Coordination Engineering team uses a checklist the same way, to assess a launch and give the team action items. It keeps the list from growing without control by requiring each new question to be justified, ideally by a previous launch problem.

Your checklist does not need nine groups for a small launch. It needs every item to carry an owner, a status, and the evidence that proves it, and it needs the gate items separated from the items that can finish after launch day.

Build the launch checklist from the material below. Return: (1) Checklist by group, using only the groups that apply to this launch: product and engineering, analytics and measurement, customer-facing documentation, marketing and communications, sales enablement, customer success and support, legal and compliance, launch day operations, and post-launch. For each item list the task, the owner as named below, the due date as written below, the status as done, in progress, not started, blocked, or not applicable, and the evidence required to mark it done. (2) Gate list: every item that must be done or not applicable before launch, including the rollback plan, the rollback decision owner, the rollback trigger conditions, the monitoring and alerting check, and the support briefing. If any gate item lacks an owner, a trigger condition, or evidence in the material below, list it at the top under Gate blockers. (3) Go or no-go summary, only if I write gate check below: for each gate item, done with evidence, not applicable with the reason I supplied, or open. Write the decision as go, restricted go for the audience I name, or no-go, using only the statuses below. Rules: use only the material below. Do not mark an item done unless the evidence below says it is done. Do not invent a date, an owner, a metric, or a test result. Write unknown where the material does not say.

What is launching and who gets it first, with rollout approach and percentage: [write it]. Launch date and time with timezone: [write it]. Launch lead and rollback decision owner: [name them]. Teams and owners by group: [paste them]. Checklist template or last launch checklist: [paste it]. Evidence and status updates so far: [paste them]. Success metric, baseline, and guardrail metrics: [write them, or leave blank].

Read the Gate blockers list first, while there is still a day to act on the risks it shows. Productboard requires the decision 24 hours before launch, with every item done or not applicable before a go is called and exceptions requiring explicit executive sign-off. If the same blocker appears at the next launch, add it to the template that week, with the justification Google’s team asks for, the launch where its absence caused a problem.

Prompt 2: Draft Every Announcement From One Approved Facts List

Run this after the gate, and rerun it if the rollout plan changes. The facts list is the contract for all launch copy. It states what shipped in plain language, who can use it now, the rollout percentage and the date the next stage is planned, the price or plan change if there is one, the known limits, and where to get help. The launch lead signs it after engineering confirms what is live in production.

Each version is for a different reader and uses the same facts. The blog or email version explains the change and the first step. The in-product note is two lines and a link. The support version is a short brief with the questions customers will ask and the escalation path. The internal version names the owners and the watch schedule. None of them may widen the audience, move a date forward, or turn a phased rollout into a general release.

Draft the launch announcement set from the approved facts list below. Return four versions from the same facts: (1) Customer announcement for email and blog: a headline that names the change, three to five short paragraphs covering what changed, who has it now, the first step to use it, and where to get help, plus a link placeholder for the help article. (2) In-product note: two sentences maximum, naming the change and who has it, with one call to action. (3) Support brief: the same facts as a short list, the five questions customers are most likely to ask based only on the facts below with a one-sentence answer for each, the known limits exactly as written below, and the escalation path and owner as named below. (4) Internal launch note: what is live, the rollout percentage, the watch schedule and channel, the rollback owner, and the time of the first metrics review. Rules: use only the approved facts below. Do not add a feature, a benefit, a price, a date, or an availability claim that is not in the list. If the facts list does not state the rollout percentage or who has access, stop and list that under Facts missing instead of drafting the customer version. Write the audience exactly as the facts list states it, for example 10 percent of accounts, never all customers unless the list says all customers.

Approved facts list, signed by the launch lead: [paste it]. Help article link: [write it]. Support escalation path and owner: [name them]. Watch schedule, channel, and first metrics review time: [write them].

Send the support brief before the customer announcement goes out. Productboard puts the customer success briefing and the escalation path on the readiness checklist for this reason, and Atlassian puts final support training in the launch phase itself. Support that learns about a launch from a customer ticket will give answers from the old product, and those answers become a second announcement you did not write.

Prompt 3: Turn the Launch Log Into Status Updates and a Retro Draft

Run this on launch day at each update point, again for the stakeholder update within 48 hours, and in full for the review one to two weeks after launch. Keep one launch log from the morning of launch day: deploy times, the rollout percentage at each step, metrics checks, support volume and the top questions, incidents and decisions, and who made each call. The log is the only source for status copy, the same confirmed-facts discipline the IT admins lesson uses for outage updates.

The retro follows the post-launch retrospective format: what went well, challenges faced, customer feedback, and improvements for next time, ending with one to three actions, each with one owner and a date. Atlassian schedules the formal review one to two weeks after launch to assess performance against objectives and document what the team learned. Run the retro while the log is still fresh. Write notes silently first, so the first voice in the room does not set the list. Keep it blameless by asking why a step failed rather than who caused it, and start the next retro by checking whether these actions happened.

Turn the launch log below into status and review drafts. Return: (1) Launch-day status update, for the time I write below: what is live and the current rollout percentage, the success metric and guardrail metrics exactly as recorded in the log with their check times, any incident or pause with the decision and who made it, and the time of the next update. Two to four sentences for the internal channel, plus a one-sentence version for stakeholders. (2) 48-hour stakeholder update, only if I write stakeholder below: progress against the success metric and baseline as recorded in the log, support volume and the top three customer questions from the log, decisions made since launch, and the date of the formal review. (3) Retro draft, only if I write retro below: what went well, challenges faced, customer feedback, and improvements for next time, each item tied to a dated entry in the log or a metric in the log, followed by a proposed action list of one to three items with a blank owner and date for the team to assign in the meeting. Rules: use only the log below. Do not invent a metric value, a customer quote, or a cause. If the log does not record a metric check, write not yet checked. Do not assign blame to a named person.

Launch log with times: [paste it]. Success metric, baseline, and guardrail metrics: [paste them]. Update time and next update time: [write them]. Mode, status, stakeholder, or retro: [write one].

The retro draft is what improves the next launch. Close it by writing the agreed actions into the checklist template itself. A retro action that says add a pricing page check to the gate list is finished only when the template carries that check, with the launch that justified it noted beside it.

Phase When What Muse drafts Done means
Readiness Final week, daily Checklist updates and blocked items by owner Every item has an owner, a status, and required evidence
Gate 24 hours before launch Go or no-go summary from gate items only Decision recorded, rollback owner and triggers written down
Launch day Hour by hour, from the runbook Status updates from the launch log on a set cadence One status thread, next update time stated every time
First 48 hours Within 48 hours after launch Stakeholder update from log and metrics checks Metric against baseline, top support questions, review date set
Review 1 to 2 weeks after launch Retro draft in four parts, with proposed actions One to three actions assigned, template updated in the meeting

What Can Go Wrong: An Announcement That Describes the Plan, Not the Shipped State

In the example setup, the team planned a full release and drafted the announcement a week early from the launch plan. Two days before launch, engineering moved to a phased rollout at 10 percent of accounts to watch error rates, and the change went into the runbook. The announcement did not change. On launch morning the email went out saying the new plan was now available to everyone, support received tickets from accounts that could not see it, and the correction email had to explain a rollout the first email should have described. The launch itself worked, but the copy described a different launch.

The draft was accurate when it was written, but it treated the plan as the shipped state after the plan had changed. Availability is a production fact, and it expires the moment the rollout decision changes.

The fix is the facts list in Prompt 2, plus a required sign-off step. The launch lead signs the facts list after the gate, with the rollout percentage and the audience written as numbers, and any change to rollout, date, or scope after that sign-off voids the drafts and triggers a rerun. Muse refuses the customer version when the percentage is missing. A missing percentage decides whether the announcement is correct or needs a correction. Keep the assembling and the drafting with Muse, and keep the go call, the sending, and the pause or rollback call with the launch lead.

Start With the Checklist and the Gate Only

Do not rebuild the announcement set, the status cadence, and the retro in the same week. Start with Prompt 1 on your next launch, with your own template pasted in and the gate items separated, and hold the gate 24 hours out even if the decision feels obvious. If the gate produces a written decision, a named rollback owner, and trigger conditions you could act on without a meeting, the setup is working. Add the announcement prompt for the following launch, and add the retro draft once you have a launch log worth reviewing.

The three launch phases, pre-launch taking 70 to 80 percent of the timeline, hour-by-hour launch day schedules with a central coordinator, and a formal review one to two weeks after launch follow Atlassian, Product Launch Timeline. The nine readiness groups, acceptance criteria and feature flag checks, analytics events and success metric dashboard, help article and release notes, support briefing and escalation path, launch runbook, war room channel, on-call schedule, first 24-hour metrics review, rollback owner and trigger conditions, stakeholder update within 48 hours, post-launch review at two weeks, and the go or no-go rule 24 hours before launch with every item done or not applicable and exceptions requiring executive sign-off follow Productboard, Launch Readiness Checklist. The launch checklist as an assessment that produces action items, and the rule that new checklist questions must be justified, ideally by a previous launch problem, follow Google, Site Reliability Engineering, Reliable Product Launches at Scale and the Launch Coordination Checklist. The retro format, what went well, challenges faced, customer feedback, and improvements for next time, ending with one to three actions each with an owner and a date, and the blameless, silent-writing-first practice follow TeamRetro, Post-Launch Retrospective Template, and standard retrospective facilitator guidance.

Keep reading