Today, a live learning session is taking place at Team EU 26 titled: Get answers from across your site with Rovo. The session is aimed at showcasing two workflows that anyone can implement in their async work.
For those who landed here after the live session: thank you for attending! If you’re here just browsing or looking to learn something new, I invite you to have a look at this recap as well.
Now you may have noticed this post’s title says three workflows. That’s right, you can now register for Debrief, ask, clean up: Three Rovo workflows to improve your async work. It contains one more workflow than the Team EU 26 session.
Below you’ll find a summary of the these three practical workflows, along with the exact prompts we used. No matter if you attended the Team EU session or not, you should have enough detail here to get you started. If you’re eager to learn more, check out our public class calendar for upcoming classes.
Debrief - Use Loom to make a quick and workable meeting debrief
The problem: You finish a meeting with a transcript and mental notes, but by the next morning, half the context is gone.
The workflow: Record a quick Loom debrief right after the meeting, then let Rovo transform both the transcript and your debrief into a structured summary with decisions, stakeholders, and follow-ups clearly separated.
What makes this work: The raw meeting transcript has raw data on what was said. Your Loom debrief describes how people looked, your gut reads, and the human element invisible to Rovo. Together, these two sources give Rovo enough to build something useful.
Open Rovo Chat in any of our apps or on your desktop, and paste the following prompt:
Summarize the Meridian Events onboarding kickoff for me. Separate what was actually decided from what was only discussed or floated. Pull out follow-up actions with an owner and a brief reason why each one matters. Call out where Dana, Marco, and Priya stand: who's bought in, who's hesitant, and where the friction points are between them.
I've linked the raw meeting transcript and my personal voice debrief from right after the call. They cover the same meeting from different angles: the transcript has the facts, the debrief has my read on the room.
This is a new customer kickoff, so assume the reader has some context on SyncUp but none on Meridian yet.
< Link to Source A - Meeting transcript >
< Link to Source B - Loom debrief >
To create Jira work items, use with this follow-up prompt:
Take the actions owned by Kevin and list them. Add a summary and a description to each action in preparation for Jira work item creation.
Then use this one:
/create-work-items Create work items for all the follow-up actions you have just listed for Kevin.
Try it: Next time you finish a meeting, open Loom and talk through what you noticed for 2-3 minutes. Then prompt Rovo to take action. If you have both the meeting and your debrief in Loom, then this workflow still works!
Ask - Get context that would take hours to find
The problem: Leadership just revived a shelved project from six months ago. You need to know what was initially decided, what's blocked, and what technical risks exist, but the information is scattered across Jira, three Confluence branches with no cross-linking, and is buried design docs. Manually tracking it all down would take hours.
The workflow: Ask Rovo one question that spans all those sources. Name the people involved, specify the tools to search, and explicitly request source citations so you know where each piece came from. Rovo connects the work items, decisions, feedback, and documentation into one sourced answer.
What makes this work: The Teamwork Graph connects everything your team has built across Atlassian and connected third-party tools. Rovo uses all that context to make connections. Every claim links back to a real source (Jira work items, Confluence pages, etc.). The key is being specific: name the people, name the project, and ask for sources.
Open Rovo Chat in any of our apps or on your desktop, and paste the following prompt:
I'm joining the dashboard feature team. What's been decided so far, what's blocked, and is there anything technical I should be worried about? Show me where each piece of information came from.
I know Max Taylor owns the product side, Ryan Lee leads engineering, and Alana Grant has been collecting customer feedback on the Support side. Pull from the SYNC project in Jira and the SyncUp space in Confluence.
Rovo can provide further input, if you use this follow-up prompt:
What questions should I ask as the newcomer to this project? Which page do you think is the best starting point?
Try it: Think of a project you joined and don’t know how to bring yourself up to speed. Perhaps a cross-department initiative you joined mid-flight that’s happening across multiple time zones and teams. Might be confusing and overwhelming when you look at a totally foreign Confluence space. Ask Rovo to pull the current state, name the people involved, and request sources. Then prompt for questions and let Rovo find the gaps in what's documented. That way, you can directly start asking questions that matter knowing what you don’t know.
Clean Up - Audit a Confluence space with a Rovo agent
The problem: So imagine your team's Confluence space has been growing organically for two years. Your teammates add pages, abandon some of them, sometimes forget to finish them, and overtime some things simply become outdated. Everyone wants to go back and clean things up but nobody has time to actually do it.
The workflow: Build a custom Rovo agent in Rovo Studio that audits the space and outputs a structured report: stale content, thin pages, overlapping topics, and healthy content.
What makes this work: Agents are like job descriptions for AI. You write the instructions once, give the agent the right tools (in this case, Confluence read access), and it can run the audit repeatedly on any space with minimal new context.
In Rovo Studio, you can create a new Rovo Agent from scratch using these inputs:
Instructions for the agent:
## Goal
You are a Confluence space auditor. Your job is to scan a Confluence space the user specifies and produce a structured health-check report identifying stale content, thin pages, potential topic overlaps, and healthy content.
## Process
1. Ask the user which Confluence space to audit (by name or URL).
2. Retrieve ALL pages in that space using the list-confluence-pages tool.
3. For each page, note: title, summary, last edited date, and approximate content length.
4. Categorize pages into four buckets:
- **Stale Content:** Pages not edited in 6+ months
- **Thin Content:** Pages with minimal content (under ~150 words or appearing to be empty templates/abandoned drafts)
- **Potential Overlaps:** Groups of pages covering similar topics that might benefit from consolidation
- **Healthy Content:** Pages that appear current, substantial, and unique
5. Create a new Confluence page in the same space titled "Space Health Check - [Today's Date]" using the create-confluence-page tool.
6. Structure the page with four sections (one per category), each with a table listing the flagged pages.
## Output Format
Create a Confluence page with this exact structure:
# Space Health Check - [Date]
This report identifies potential maintenance opportunities in the [Space Name] space. Review each section and decide which items need attention.
### 📅 Stale Content
Below the heading, add this text as a paragraph: Pages not edited in the last 6 months. Review to determine if they're still relevant or should be archived. Fill out the Action needed and Assignee columns to make Jira automation smoother.
Table columns:
| Asset title | Summary | Last edited | Days since last edit | Action needed? _(yes/no)_ | Assignee |
- "Asset title": Linked page title
- "Summary": Two-sentence description of the page's content
- "Last edited": Date in YYYY-MM-DD format
- "Days since last edit": Calculated integer
- "Action needed?" and "Assignee": Leave blank for the user to fill out
Sort rows by oldest first.
### 📄 Thin Content
Below the heading, add this text as a paragraph: Pages with minimal content that may be empty templates or abandoned drafts. Fill out the Action needed and Assignee columns to make Jira automation smoother.
Table columns:
| Asset title | Summary | Last edited | Approximate word count | Action needed? _(yes/no)_ | Assignee |
- Same rules as Stale Content, except column 4 is approximate word count instead of days since edit.
### 👯 Potential Overlaps
Below the heading, add this text as a paragraph: Groups of pages that cover similar topics and may benefit from consolidation.
Table columns:
| Overlap group | Pages involved | Justification | Action needed? _(yes/no)_ | Assignee |
- "Overlap group": A short label for the topic that spans the overlapping pages (e.g., "Goal Setting & OKRs").
- "Pages involved": List of all pages (with links) identified as overlapping for this topic.
- "Justification": Why you identified these pages as overlapping.
- "Action needed?" and "Assignee": Leave blank for the user to fill out.
- Each row represents one overlap group, not one page.
### ✅ Healthy Content
Below the heading, add this text as a paragraph: Pages that appear current, substantial, and unique. No action needed. Please have a brief look at them anyway to make sure there are no false positives included.
Table columns:
| Asset title | Summary | Last edited | Flag for review? |
- Same rules as other tables. "Flag for review?" is left blank for the user.
- List ALL healthy pages. Do not abbreviate or truncate this list.
## Rules
- Be specific. Link to each flagged page by title.
- For every page mentioned anywhere in the audit, include a two-sentence summary of its content.
- Include ALL four sections even if a category has zero findings. Show "No issues found in this category" in a single merged row rather than omitting the section. This shows the audit was thorough.
- Do not modify, archive, or delete any pages. This is a read-only audit.
- Keep the tone neutral and helpful. This is a tool for space owners, not a criticism of contributors.
- After creating the page, share the link and let the user know they can fill out the Action needed and Assignee columns, then use the Jira automation prompt to create work items.
- **Fallback:** If you are unable to create the Confluence page, output the full audit directly in the chat using the exact same structure and table format described above. Do not summarize or abbreviate. Include all sections, all tables, and all rows so the user can copy the output into a Confluence page themselves.
Tools added to the agent:
- Get page
- Create live doc
- Create work item
Knowledge sources:
- All organizational knowledge (so the agent can access any space the user has permission to view)
Try it: Build the agent once in Rovo Studio, then run it on any Confluence space that needs a cleanup. The agent creates a full audit you can share with your team. Don't worry, this Rovo agent won't start making changes unless you ask it to.
Try something new in Rovo this week
Pick the workflow that fits your current pain point:
- Debrief if you're drowning in meeting notes
- Ask if you're joining a new project
- Clean Up if your Confluence space feels like a digital junk drawer
🌟 Share your experience below: Which workflow did you try? What changed when you ran the prompt? The gap between your first attempt and your refined result is usually the most interesting part. It's a continuous improvement process!
Keep learning
