Q&A from Getting started with Rovo and Find and use AI agents with Rovo classes July 15th and 16thh.

[#1]

Thank you for joining last week’s public classes, Getting started with Rovo (July,15th ) and Find and use AI agents with Rovo (July 16th).

I planned to post this much sooner, but I was ill after the sessions and am only now catching up. I appreciate your patience on this!

I’ve combined both classes here, including the demos, the questions from chat, the opening that was missed in one recording, and the points I promised to check.

Where I floated an answer live and was not completely sure, I have checked the current Atlassian documentation and tested it.

Quick recap of the two classes

Getting started with Rovo

We covered:

• What Rovo is and how Search, Chat, and Agents fit together.

• How the Teamwork Graph connects work, people, and knowledge.

• Search, knowledge cards, definitions, and the browser extension.

• Chat for finding, summarising, reviewing, and creating content.

• Atlassian agents, no-code custom agents, and Forge-based agents.

• Prompting with TCREI: Task, Context, References, Evaluate, and Iterate.

• Skills, Jira work creation, and a quick reference to Max mode.

Find and use AI agents with Rovo

We covered:

• Finding agents, reading their descriptions and instructions, and starring favorites.

• Using an agent in Chat, in an editor, on a Jira work item, and in automation.

• Social media, service management, blocker detection, and work-planning examples.

• Assigning Jira work to an agent.

• Scheduled and event-driven automation patterns.

• Multi-step workflows that turn information into Jira work.

Rovo in one simple map

The features overlap, but they are not interchangeable. Choosing the smallest suitable tool usually gives you the cleanest result.

Why is it called Rovo?

The accurate answer is that Rovo is the product name. Atlassian describes it as the app that helps teams find, learn from and act on organizational knowledge.

How Rovo finds and uses your organization’s knowledge

Rovo uses Atlassian’s Teamwork Graph to connect people, work and knowledge across Jira, Confluence and other Atlassian apps.

Admins can also connect approved third-party sources such as Google Drive, Microsoft 365 or Slack. Search, Chat and agents can then use that connected knowledge.

Rovo respects existing permissions. A user does not gain access to a Jira work item, Confluence page or connected document simply because an agent can refer to that source.

This matters when you share agents or automate reports. Two people can ask the same agent the same question and receive different results because their access is different.

Does every Confluence user have Rovo?

Rovo is included with paid Jira, Confluence, Service Collection and Teamwork Collection Cloud subscriptions.

For Standard, Premium and Enterprise apps, Rovo must still be available for the app. Admin settings can also affect where its features appear.

If the Rovo button is missing, ask an organization admin to check Atlassian Administration > Rovo > Rovo access and the relevant app’s AI settings.

Your organization also needs a verified business domain. Rovo is not currently available in Atlassian Government organizations.

See Manage Rovo access and The Rovo button.

Search, knowledge cards and the browser extension

Rovo Search accepts natural-language questions, not just exact titles or work item keys.

Ask “What is Project Stackpad?” and Rovo can return relevant Jira work, pages and connected sources. When it recognizes a person, team or definition, it can display a knowledge card at the top of the results.

People cards show useful information about a colleague. Team cards summarize a team. Definition cards explain internal terms and acronyms using information you can access.

Rovo can also surface inline definitions while you work in supported Atlassian and browser contexts, so you do not need to leave the page to decode an unfamiliar term.

The Rovo browser extension provides a Rovo-powered new-tab experience. You can also type Rovo followed by a search term in the address bar to search company knowledge.

Small correction from class: the extension does not have to replace Google search.

See Rovo Search and the Rovo browser extension.

Getting more from Rovo Chat

Chat is useful for quick questions, but its real value appears when you give it clear sources and a defined output.

You can attach a file or image, add a URL, link a Confluence page or Jira work item, and include the page you are currently viewing as context. You can also remove the current page when it is not relevant.

For a targeted answer, tell Rovo which sources to use and which to ignore. For broader investigation, use a deeper reasoning or research mode and allow more time for the response.

Depending on your organization’s settings, Chat and agents can use company knowledge, connected apps and optional web search. Name the source boundary when it matters.

The current Chat modes have different jobs and different credit costs:

• Quick Answer is best for fast, focused questions.

• Think Deeper plans and reasons through a more complex problem.

• Deep Research searches and synthesises a larger body of information with references.

Try:

Summarize this page for a new team member. Use five bullets covering the goal, owner, decisions, open questions and next action. Link each decision to its source. Ignore archived content.

That is more reliable than simply asking, “Can you summarize this?”

See Rovo Chat capabilities.

Skills: where to find them and how to use them

Skills are reusable capabilities that let Rovo do something in an app, not just generate text.

Think of Chat as the conversation and skills as the tools Rovo can use during that conversation.

To find the skills available to you:

1. Open Rovo Chat.

2. Select the plus (+) button.

3. Browse or search the available skills.

4. Alternatively, type / in the prompt box to see slash commands.

If you are already on a Jira work item or Confluence page, open the Rovo button at the bottom right. It offers skills relevant to whether you are reading or editing.

Some skills also respond to natural language. Atlassian currently documents this for creating work items, updating work items and writing status updates.

Can I create my own skills?

This depends on the Rovo surface you are using.

In standard Rovo Chat, use the approved skills directory. organization-built and third-party tools appear only after they have been approved and enabled for the site.

Rovo Dev supports Agent Skills separately. A shared Confluence prompt library is still a great team habit, but a saved prompt is not the same as a first-class Rovo Chat skill.

See What are skills? and Use skills in a chat.

Agents: where to find them and what makes one useful

Rovo includes more than 20 Atlassian-built agents. You can also create a no-code custom agent in Studio or build more advanced agent capabilities with Forge.

Open Chat, select Agents, then browse or search. An agent profile shows its purpose, instructions, knowledge sources, tools and conversation starters.

Read the description first. Then inspect the instructions and tools before relying on the agent for an important workflow.

You can star a useful agent, copy a link to its profile, duplicate some agents, or edit one you own. Studio is the main place for configuring custom agents and automations.

In Studio, you can describe the agent you want in everyday language and let Rovo prepare a draft. You can also use manual setup when you want direct control over instructions, knowledge, tools and conversation starters.

An effective agent has a narrow job, clear instructions, trusted knowledge and only the tools it needs.

Examples from the classes

Agent or pattern	What it can help with
Social Media Writer	Draft for each platform, add hooks, hashtags and calls to action, then reshape a post into a short campaign
Service Request Helper	Summarize a request, find similar resolutions, and draft a customer reply and internal steps
Work Item Planner	Turn Confluence goals into epics and stories, then add completion criteria and team tags
Work Item Organizer	Find, group and update Jira work, including parent and sprint suggestions
Blocker Checker	Review comments for blocker signals, then let automation apply the agreed status, label or comment
Jira Delivery Agent	Find stale work and risks, then produce stand-ups or leadership updates
Customer Insights	Find themes and signals in approved customer feedback

PRD means Product Requirements Document. Rovo’s Product Requirements Guide agent can help create or review one.

Agents can be used in Chat, in Jira and Confluence editors, in automation rules, and on Jira work items.

Assigning work to an agent is useful when it has a defined process to run, needs to ask for missing context, or should leave structured findings on the work item.

See the current out-of-the-box agents and browse Rovo agents.

Creating and updating Jira work items

Julianne asked whether creating or updating Jira work should be done by an agent. It can be, but it does not need to be.

Option 1: use Chat and the create work items skill

For a one-off job, open Rovo Chat in Jira and enter /create-work-items, followed by the outcome, target space and useful context.

Rovo suggests a set of work items on an AI-powered canvas. You can review, split, rewrite or add acceptance criteria before selecting Create all.

You can link an existing work item, Confluence page or Loom video as context.

Try:

/create-work-items Create an epic and delivery stories from this product brief. Use space ABC.

Add a clear outcome, acceptance criteria, owner placeholder and dependency notes to every story: [link]

The canvas feature is rolling out gradually. Your organization admin may need to enable beta AI features, and you need permission to create work in the target Jira space.

Chat also has a built-in capability for creating a single work item. The /create-work-items skill is the better route for a set of related work or an epic structure.

See Create work items with Rovo.

Option 2: use an agent

Use an agent when you want repeatable judgement or a consistent planning method before work is created.

The Work Item Planner can break a large initiative into manageable work. A custom agent can use your planning rules, definitions and approved knowledge sources.

Interactive agents ask for confirmation before taking action. Current Atlassian documentation says bulk Jira actions performed by an agent are limited to 20 work items at a time.

That replaces my informal “10 to 15 works well” comment from class. Ten to fifteen may be a sensible batch size, but it is not the documented limit.

See Rovo agent tools.

Option 3: use an automation flow

Use automation when creation or updates should happen because of an event or schedule.

Inside an automation, the agent returns text. It does not use its own interactive tools.

Pass the result through {{agentResponse}}, then let a later automation action create or update a Jira work item, add a comment, publish a page or send a message.

Another class example was meeting preparation: run a weekly flow, ask an agent to assemble the agenda, current risks and required decisions, then publish the result and send a Slack notification.

For example:

1. When: a new request is created.

2. Then: use a triage agent to classify it and propose delivery work.

3. Then: create the Jira work using the structured agent response.

4. Then: comment with what was created and why.

See Automating Rovo agents.

Option 4: use Max mode

Max is useful when the job is genuinely multi-step: read several documents, reconcile a plan, edit a list of work items and create a report from the result.

It is more capable, but it also has extra installation, subscription and safety requirements. Those are covered below.

Your agent questions

Can an agent act as a virtual project manager or Scrum Master and create scheduled reports?

Yes. This is a strong Rovo use case, but the scheduled part belongs in Atlassian Automation.

The Jira Delivery Agent can identify stale work, blockers and slipping dates, produce stand-up digests, and tailor updates for teams or executives.

Combine a scheduled trigger with the agent, then publish {{agentResponse}} to Confluence, Slack or email.

Try:

Automation Mode. Review space ABC for completed work, work in progress, stale items, blockers, dependencies and dates at risk.

Produce a five-bullet team update and a three-bullet leadership summary. Cite the work items behind every risk.

See Work with Jira Delivery Agent.

Does an agent run in the background 24/7?

Not by itself. An agent becomes autonomous when an admin places it in an enabled automation flow.

The flow can run on a schedule or react to events such as a work item, comment or status change. The automation watches; the agent reasons when the rule invokes it.

Scheduled Jira flows can run at a fixed interval or use a cron expression. See Jira automation triggers.

Can I share an agent with a lead instead of sending a static report?

Yes. In Chat, open Agents, find the agent, select More actions, and copy the link to its profile.

The person receiving the link can start a full-screen chat with the agent. The agent does not give them extra access; they can only retrieve information they already have permission to see.

For self-service, add clear conversation starters such as “Give me this week’s delivery risks” or “Summarize roadmap changes since Monday”.

I would pair the shared agent with a scheduled automation when leaders also need a dependable weekly digest.

See Share a Rovo agent.

Can Rovo summarise roadmaps across several Jira Product Discovery spaces?

Yes, provided the user and agent can access the relevant spaces and the prompt defines the scope clearly.

Start manually. Give Chat or a custom agent the exact spaces, saved views or JQL, the date range, and the fields that matter. Ask for sources so every conclusion can be checked.

Try:

Compare the active roadmaps in JPD spaces A, B and C. Group ideas by strategic theme.

Highlight duplicated bets, conflicting dates, missing owners, dependencies and changes from the last 14 days. Link every finding to its source.

If this becomes a regular need, move the tested prompt into an agent and then into an automation.

Jira Product Discovery Premium can also consolidate ideas from multiple projects in a cross-product roadmap view. Rovo is useful as the narrative and analysis layer on top.

Can Rovo identify blockers or blocked stories?

Yes. There are two good patterns.

Use Jira Delivery Agent for broad delivery health, stale work and risk. Use Blocker Checker in an automation when comments such as “waiting on API team” should trigger a consistent check.

The safe pattern is for the agent to explain its decision and for the automation to apply the status, flag, label or comment.

Can it email a report covering risks, dependencies and blockers?

Yes. Use a scheduled automation, invoke the agent, then add Jira Automation’s Send customized email action and place {{agentResponse}} in the body.

Try:

Review space ABC for the last seven days. Report blockers, dependencies, overdue work and dates at risk.

For every finding, give the work item link, owner, evidence and suggested next action. If there is no evidence, say so.

Test the flow with a small, known scope first. See Jira automation actions.

Can it identify upsell and cross-sell opportunities?

Potentially, yes, but treat it as an opportunity-finding assistant, not an automated sales decision-maker.

Use the Customer Insights agent or a narrowly scoped custom agent. Give it approved customer feedback and work data, plus explicit rules for what counts as a signal.

Ask it to separate evidence from inference, link every recommendation to a source, avoid sensitive attributes, and require human approval before any customer action.

Try:

Find recurring customer needs that are not met by the current scope. Group the evidence by account and theme.

Suggest possible expansion conversations. Label each as evidence-backed, weak signal or insufficient data. Do not contact customers.

Rovo Max, Max mode and Rovo Desktop

The naming caused understandable confusion, so here is the clean version.

“Rovo Max” is not a separate upgrade or download. The current product name is Max mode, and it lives inside the beta Rovo Desktop app.

Max plans and executes complex, multi-step work. In Rovo Desktop it can also work with local documents, spreadsheets, PDFs and source files, and run command-line operations.

Download Rovo Desktop

1. Open Get started with Rovo Desktop.

2. Choose the Windows or macOS installer for your device’s x64 or ARM architecture.

3. Run the .MSI on Windows, or open the .DMG and drag Rovo into Applications on macOS.

4. Open Rovo Desktop and sign in with your Atlassian account.

Rovo Desktop requires an Atlassian account with Rovo access. Your organization’s SSO, network or firewall settings may also need to allow it.

Extra prerequisites for Max mode

Max mode currently requires:

• Rovo Desktop access.

• Rovo Dev activated on at least one site.

• Rovo Dev CLI installed through Atlassian CLI (ACLI).

• Rovo Dev credits through Rovo Dev Standard.

• A scoped API token and acli rovodev auth login to authenticate the CLI.

Follow Use Max mode in Rovo Desktop and Install Rovo Dev CLI.

Max can edit files and run commands locally. It asks for approval for sensitive actions by default. Treat Auto approve as an advanced setting and use it only in a tightly controlled workspace.

Rovo credits and where admins manage them

There are two separate credit systems:

• Rovo credits cover Rovo Chat, Rovo Agents, Deep Research, Think Deeper and some Teamwork Graph API calls.

• Rovo Dev credits cover Rovo Dev features, including the CLI used by Max mode.

They are not interchangeable.

Current Rovo credit allowances

As of 19 July 2026, Atlassian documents these per-user, per-month allowances. They are pooled at organization level, reset monthly and do not roll over.

Edition	Jira, Confluence or Service Collection/JSM	Teamwork Collection
Standard	25 credits	250 credits
Premium	70 credits	700 credits
Enterprise	150 credits	1,500 credits

The separate Rovo Standard beta subscription currently adds 400 Rovo credits per standalone Rovo user each month.

What currently consumes Rovo credits?

• Rovo Agent request: 10 credits.

• Rovo Chat Quick Answer request: 10 credits.

• Rovo Chat Deep Research request: 100 credits.

• Rovo Chat Think Deeper request: variable.

• Teamwork Graph API call: variable. Write actions such as creating a Jira work item are currently free.

Rovo Search, in-product summaries, definitions and Atlassian Analytics chart insights do not currently consume Rovo credits.

Beta Rovo features are currently free unless Atlassian explicitly says otherwise.

Atlassian does not publish a separate automation rate for agent requests, so use the documented 10 credits per agent request when planning.

See the live Rovo usage allowance before relying on the figures, as credit policy can change.

Where can an admin see usage?

An organization admin can go to:

Atlassian Administration > Insights > Platform usage > Rovo credits

The page shows allowance, actual and forecast usage, refresh date, history, and breakdowns by app or user. Admins can also export Rovo credit data as CSV. The export is currently in beta.

Rovo > Rovo Insights provides a different view of adoption and agent activity. Its activity figures will not always match credit usage because some Rovo features do not consume credits.

See Platform usage.

For adoption data, see Rovo AI activity insights.

How many credits did the class demos use?

I cannot give an exact total from the recording alone because it does not capture every backend request or whether a feature was still in beta.

Rovo Dev has its own allowance and billing controls. See How billing works for Rovo Dev.

Prompting tips that make a real difference

TCREI stands for Task, Context, References, Evaluate and Iterate.

For agents and automations, add four practical details:

• Define the audience and output format.

• Add the required tone and a short example when style matters.

• State the time period and exact Jira scope.

• Require links to evidence.

• Define what the agent should do when evidence is missing.

Also tell Rovo what to ignore, break complex work into steps, and keep a human in the loop before content reaches customers, executives or production workflows.

Follow your organization’s data-handling and classification policy when attaching files or connecting sources. Do not assume that a document is safe to upload simply because an AI tool accepts it.

Try this status prompt:

Create a Friday delivery update for senior leaders. Review Jira spaces ABC and XYZ for the last seven days.

Cover progress, blockers, dependencies and dates at risk. Use five bullets, link every claim, and say “insufficient evidence” when the data does not support a conclusion.

For a reusable agent, add role and behavior:

Act as a delivery analyst for teams ABC and XYZ. Use only the listed Jira spaces and linked Confluence pages.

Always separate confirmed facts, reasonable inferences and missing evidence. Never change work or send a message without confirmation.

Prompt to Agent:

Test a prompt in Chat first. Once the result is reliable, move it into an agent. Once the agent is reliable, invoke it from an automation.

Useful links

Community and learning

• Training and Certification community

• Atlassian learning

Rovo fundamentals

• What is Rovo?

• Rovo Search

• Rovo Chat capabilities

• Use skills in Rovo Chat

• Browse Rovo agents

• Out-of-the-box agents

• Rovo Admin Course

• Get the Most out of Rovo Learning Path

Jira actions and automation

• Create work items with Rovo

• Rovo agent tools

• Automate Rovo agents

• Jira Delivery Agent

Desktop, Max and credits

• Download Rovo Desktop

• Use Max mode

• Install Rovo Dev CLI

• Rovo usage allowance

• Platform usage

Recordings

• Getting started with Rovo

• Find and use AI agents with Rovo

If you try any of these patterns, especially scheduled reports, Work Item Planner or /create-work-items, I would love to hear what worked and what didn’t.

Drop a comment, join another public class, and keep an eye on the schedule for Sven’s advanced agent-building sessions.

Thanks again for attending and for the excellent questions.

[Esther Ortega]

Thank you so much, Emma! What an incredible compilation effort. Bringing together in one place the demos, the questions from the chat, and above all the points you verified afterwards against the documentation is enormously valuable.

I'm saving this as a reference.

And I hope you've now fully recovered! Thanks again for taking the time to leave everything so well organized.

[Emma Wolstencroft]

Thanks, @Esther Ortega   , it's actually really useful for me too, as I review all the support docs, check out some features, think of examples, and try out the interesting use cases that people in the class suggest!

I am feeling much better, thank you!