> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ravenna.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Rollout checklist

> A week-by-week plan for rolling Ravenna out to one team and then the rest of the company, from setup call to steady state.

export const Checkbox = ({items = []}) => {
  const [checked, setChecked] = useState(Array(items.length).fill(false));
  const toggle = index => {
    setChecked(prev => {
      const copy = [...prev];
      copy[index] = !copy[index];
      return copy;
    });
  };
  return <div className="custom-checkbox-list">
      {items.map((text, index) => <div key={index} className="custom-checkbox-item flex items-center mb-2">
          <input type="checkbox" checked={checked[index]} onChange={() => toggle(index)} className="mr-2" />
          <span>{text}</span>
        </div>)}
    </div>;
};

A four-week plan that gets one team live properly and then repeats. It stitches together the [Day one](/guides/day-one/overview) track, the [announcement templates](/guides/user-onboarding/announcements), and the [handbooks](/guides/user-onboarding/handbook/simple-edition).

Copy it, put dates and owners on it, and cut anything that does not apply. The sequence matters more than the timing: pick one team, get it genuinely working, and use it as the reference implementation for everyone else.

<Tip>
  The most common rollout failure is going company-wide in week one. You get volume before you have knowledge, an agent with nothing to retrieve, and a first impression you have to earn back.
</Tip>

## Week 0: before the setup call

Owned by whoever is sponsoring the rollout.

<Checkbox
  items={[
"Pick the first team, usually IT or People, whichever has the highest volume of repeatable questions",
"Name the workspace admins, ideally two so nobody is a single point of failure",
"Confirm you have Slack admin rights, or line up who does",
"Decide whether the workspace should be private, and read Portal and privacy before you answer",
"List where your knowledge lives today: Notion, Confluence, Google Drive, a Slack channel, someone's head",
"Write down the five questions your team answers most often, in the words employees actually use",
"If access provisioning is in scope, identify your identity provider and who administers it"
]}
/>

<Callout icon="link" color="#6B7280">Learn more about [what to have ready](/guides/before-you-start)</Callout>

## Week 1: get requests flowing

The goal for the week is a real request arriving and being worked in Ravenna. Nothing intelligent yet.

<Checkbox
  items={[
"Create the workspace and invite the responders",
"Connect the public request channel, and post a test request in it yourself",
"Set up the triage channel and confirm mirrors land there",
"Connect an agent to the workspace's Slack channel so DMs route somewhere",
"Create 5 to 10 categories with descriptions and example phrases",
"Build one form for your highest-volume structured request",
"Tell the responders only. No company announcement yet"
]}
/>

By Friday, the responders should be working real tickets in Ravenna rather than in Slack threads. That habit is the actual dependency for everything after this.

<Callout icon="link" color="#6B7280">Learn more about [the Day one track](/guides/day-one/overview)</Callout>

## Week 2: make it smart

<Checkbox
  items={[
"Connect one knowledge folder you trust, not the whole wiki",
"Test retrieval in the Chat panel with the five questions from week 0, and fix the documents that fail",
"Configure the agent: custom prompt for environment and guardrails, rules for everything conditional",
"Write the escalation instructions and mean them",
"Run the agent in Testing Mode against real phrasings before it touches anyone",
"Build one workflow, triggered by your form, that does the assignment and the notification",
"Set one SLA on first response only",
"Publish the agent to the request channel"
]}
/>

<Callout icon="link" color="#6B7280">Learn more about [writing agent rules](/guides/best-practices/writing-agent-rules) and [knowledge for AI](/guides/best-practices/knowledge-for-ai)</Callout>

## Week 3: announce and watch

<Checkbox
  items={[
"Post the announcement in the company channel using a template",
"Share the handbook, or a trimmed version of it",
"Pin the request channel and add a deep link somewhere people already look, like an Okta tile or the intranet",
"Read the agent chat logs daily this week. Every wrong answer maps to a missing rule, an overlapping trigger, an unpublished form, or unreachable knowledge",
"Check the unassigned view every morning",
"Turn on CSAT and read the low scores rather than the average",
"Force a Knowledge Gaps sync at the end of the week and write the top one or two articles"
]}
/>

Expect the first week of real volume to expose two or three things you got wrong. That is the point of running one team first.

<Callout icon="link" color="#6B7280">Learn more about [going live](/guides/day-one/go-live)</Callout>

## Week 4: steady state and the next team

<Checkbox
  items={[
"Review the analytics: volume by channel, AI outcome mix, and resolution path",
"Build the two or three views the team actually asks for, instead of new channels",
"Set a recurring 15-minute weekly slot for Knowledge Gaps",
"Ship a weekly AI brief to the team channel so the numbers arrive without anyone opening a dashboard",
"Write down what you would do differently, then hand it to the next team",
"Repeat weeks 1 to 3 for team two, which should take days rather than weeks"
]}
/>

<Callout icon="link" color="#6B7280">Learn more about [reporting and briefs](/guides/playbooks/reporting-and-briefs) and [closing knowledge gaps](/guides/how-to/close-knowledge-gaps)</Callout>

For team two, start from that team's [department playbook](/guides/playbooks/overview) rather than repeating Day one from scratch. The workspace, categories, and forms are different, and everything you learned about knowledge and agent rules carries over.

## What good looks like at 90 days

Rough targets rather than promises, because the mix depends heavily on how much of your knowledge was written down before you started.

| Signal                                                         | Healthy                                       |
| -------------------------------------------------------------- | --------------------------------------------- |
| Requests arriving through Ravenna rather than a DM to a person | The large majority                            |
| Tickets the agent resolves with no human involved              | Rising month over month                       |
| Default category                                               | Nearly empty                                  |
| Knowledge Gaps at Critical                                     | Zero, because you worked the list weekly      |
| Channels                                                       | Still the number you started with, plus views |
| SLA breaches                                                   | Rare enough that each one gets discussed      |

If the agent's resolution rate is flat, the answer is almost always knowledge, not configuration.

## Next

<CardGroup cols={2}>
  <Card title="Announcement templates" icon="megaphone" href="/guides/user-onboarding/announcements" horizontal>
    Copy, edit, post.
  </Card>

  <Card title="Playbooks" icon="book-open" href="/guides/playbooks/overview" horizontal>
    Pick the department you are launching next.
  </Card>
</CardGroup>


## Related topics

- [Guides](/guides/overview.md)
- [Notion templates](/guides/user-onboarding/handbook/notion.md)
- [Announcements](/guides/user-onboarding/announcements.md)
- [Day one](/guides/day-one/overview.md)
- [Everyday tasks](/documentation/automate/copilot/everyday-tasks.md)