Skip to main content
Everything is built. This step is about opening it to real people and knowing where to look when they start using it. Time: about 20 minutes, plus a week of light attention.

​
Pre-launch checks

Walk these as an ordinary employee, not as an admin. Use a second Slack account or a colleague who has not seen the setup. Two things worth checking specifically because they are quiet failures:
  • The workflow is published, not draft. Drafts never fire, and nothing tells you at submission time.
  • Knowledge folders are scoped to the channel you launched. A folder restricted to a different channel simply returns nothing, which reads as an agent that does not know anything.

​
Announce it

Say three things: where to ask, what the AI will do, and what happens when it cannot help. Skip the feature tour. Ravenna ships announcement copy and employee handbooks you can paste into Slack and adapt.

Announcement templates

Launch-day Slack copy in several tones.

Employee handbook

A short explainer to share with the whole company.
Pin the announcement in the request channel and set the channel topic to point at it. New joiners find the topic long after the message has scrolled away.
Make it easier to reach from outside Slack too. Deep links let you drop Ravenna into an Okta dashboard, an intranet page, or an email footer.
Learn more about deep linking

​
What to watch in week one

Five signals, in the order they tell you something.

Agent chat logs

The highest-value screen in your first week. Every conversation, the result the agent assigned itself, and any thumbs up or down.Read twenty conversations by hand. You will learn more from that than from any dashboard, and you will usually find one missing rule and one badly worded form field.
Learn more about chat logs
Questions the agent could not answer, clustered by topic and ranked by severity. This is your documentation backlog, written by your employees.
Learn more about closing knowledge gaps
Build a view filtered to unassigned and not resolved, and keep it open. In week one the failure mode is not a bad answer, it is a request nobody noticed.
Learn more about views
Low scores are useful, but the comments are more useful. They tell you whether people are unhappy with the answer or with how long it took, which point at completely different fixes.
Learn more about CSAT
Ravenna classifies every closed ticket an agent touched as Resolved, Assisted, Escalated, or Not Applicable. It is the honest measure of how much the AI is actually carrying.Do not judge it in week one. Get a baseline, then watch it move as you close knowledge gaps and tighten rules.
Learn more about analytics dashboards
Schedule an AI brief to your triage channel so the weekly summary arrives without anyone remembering to look.

​
Where to go from here

Day one covered one team and one desk. The next moves depend on what your week-one signals showed.

Answers are thin

Work the knowledge gaps list.

Too many manual approvals

Move them into approval templates.

People outside Slack are stuck

Launch the Portal.

No agreed response times

Set SLAs against your business hours.

Ready for the next team

Pick a department playbook and stand up desk two.

Rolling out company-wide

Follow the week-by-week plan.
Last modified on September 17, 2026