Rovo Work and the Artifacts App: What Longer AI Tasks Mean for Your Team

Rovo Work is a new mode in Rovo Chat that takes on complex, multi-step tasks and keeps running for minutes or hours in a secure cloud sandbox. You give it a goal, it proposes a plan, you approve it, and it does the work across Jira, Confluence and your connected tools. The Atlassian Artifacts app is where the results land: a permissioned library for AI-generated files that you can share, search and embed anywhere in Atlassian.

Together they change what you can hand to AI. Instead of asking one question at a time, your team can delegate a whole job, close the laptop and come back to a finished report, calculator or prototype. That also brings new questions about permissions, review, cost and data. This guide covers how both features work, what they are good for and how to get your team ready.

Key takeaways

• Rovo Work handles long-running AI tasks with a plan you approve first, and runs in a sandbox your admins govern.
• Each task runs as the person who started it and inherits their permissions, so it can only reach what that person can reach.
• Atlassian Artifacts stores AI-generated artifacts from Rovo, Claude, Codex, ChatGPT and other connected agents. It is in beta and available on all paid plans.
• Artifacts are private by default. Once shared, they show up in Rovo Search and the Teamwork Graph, and they can be embedded in Confluence pages and Jira work items.
• Work mode counts as a Premium AI interaction, so it uses more Rovo credits than a quick chat answer. Plan for it before extra usage billing starts on 3 December 2026.

Why trust this guide

Empyra is an Atlassian Platinum Solution Partner with 80+ certified consultants and 1,000+ customers, including Rakuten, Victoria's Secret, Houston Community College, Sally Beauty and UC Office of the President. Our Atlassian AI consulting team helps organizations roll out Rovo with the right permissions, governance and budget in place. Product facts come from Atlassian's announcement, support and licensing pages, linked throughout and checked on 9 November 2026.

What is Rovo Work?

Rovo Work is a reasoning mode in Rovo Chat built for tasks that need many steps, and one of the main Team '26 Europe announcements. Atlassian describes it as a mode that "handles complex, multi-step tasks with human oversight to review and approve".

The difference from a normal chat is scale. A chat answer takes seconds. A Work mode task can run for hours, trying tools, running scripts and correcting itself until the job is done. Atlassian says it can even teach itself new tools along the way.

Three things make this possible:

• A computer per session. Every Work mode session gets its own cloud computer, which Atlassian says starts in about 60 milliseconds. That is what lets Rovo write and run code, not just text.
• A secure sandbox. The work runs in an isolated environment in Atlassian's cloud, not on your laptop. You can close your device and the task keeps going.
• The same platform rules. The sandbox uses the same Teamwork Graph context as the rest of Rovo, runs as the user who started it, and supports enterprise settings such as data residency. Admins control network access from the sandbox.

If you used Max mode in the Rovo desktop app, Work mode is replacing it. Max mode ran tasks locally on your computer through the Rovo Dev CLI and needed Rovo Dev Standard credits (Atlassian Support). Work mode runs in Atlassian's cloud instead, and Atlassian says it is "the same Work mode you'll find across Rovo, with no separate Rovo Dev subscription needed" (Atlassian Community).

Work mode may not be on for every site yet. Atlassian runs a waitlist for early access. To check your site, open the reasoning mode menu in Rovo Chat: if Work isn't listed next to Think deeper and Deep research, it isn't on for you yet.

How does Rovo Work run a task?

Every task follows the same loop. Knowing it helps you write better requests and know when to step in.

1. You set the goal. Describe the outcome, not the steps. For example: "Build a cross-project status report for the payments program from Jira and Confluence."
2. Rovo proposes a plan. It lists the steps, sources and tools it intends to use.
3. You shape and approve the plan. Remove sources it should not use, add constraints, or change the output format. Nothing runs until you approve.
4. Rovo works in the sandbox. It gathers context, runs scripts, reads errors and adjusts its approach. This can take minutes or hours.
5. You review the output. The result is saved as an artifact you can check, edit and share.
6. Confirm availability. Check whether Work mode is enabled on your site, or join the waitlist. Confirm the Artifacts app appears in the app switcher.
7. Clean up permissions. Audit restricted spaces, projects and Jira issue security levels, because tasks run with each user's access.
8. Set sandbox and data rules. Decide on network access, data residency needs, and which data types must not go into artifacts.
9. Agree sharing rules. Default to Restricted, and name who can publish an artifact as Open.
10. Budget for credits. Add Work mode as its own line, set a spending cap and assign someone to watch usage alerts.
11. Start with a pilot. Pick two or three repeatable tasks, such as a weekly status report, and measure time saved and credits used before you scale.
12. Train people to review plans. Show users how to read and edit a plan before approving it. A careful plan review is the simplest way to avoid wasted runs and credits. Our Atlassian training programs can cover this for your teams.

Step 3 is where most of the value and most of the risk sits. A clear plan review catches the wrong data source or the wrong scope before Rovo spends an hour and a pile of credits on it.

Rovo Work use cases: what it is good for

Work mode suits jobs that a person would normally block out an afternoon for. Of all the Rovo use cases, these are the ones that need many steps: reports, briefs, plans, calculators, prototypes and research. Atlassian uses each of them as an example.

Sources: Atlassian, Work mode and Atlassian, platform announcement.

For quick questions, a short summary or a single lookup, stay with a standard chat. Rovo Work mode is the better fit only when the task has several steps, several sources, or needs code to produce the result.

What is the Atlassian Artifacts app?

Atlassian Artifacts is a central library for the files AI agents create for you. Atlassian built it to solve a familiar problem: AI output gets lost in chat history and downloads folders, and it is often shared without the right permissions (Atlassian Support).

An artifact can be any file, including HTML, Markdown, CSV, PDF and PPTX files (Atlassian). In practice that means reports, dashboards, diagrams, calculators and prototypes. Here is how the app works:

• It collects output from many AI tools. Anything Rovo creates is saved to your Artifacts library automatically. Claude, Codex, ChatGPT and other agents can save to it too, once they are connected through the Atlassian MCP server or the Teamwork Graph CLI (Atlassian Support).
• You can upload your own files. Use Upload in the Artifacts app, which is useful for files an agent failed to save.
• Every edit adds a new version. When Rovo or another agent updates an artifact, it saves a new version. You can view, download or share older versions from Version history, and agents can't delete or revert versions. To make a change yourself, open the artifact and select Ask Rovo (Atlassian Support).
• Artifacts stay live when embedded. Copy an artifact's link, paste it into a Confluence page or Jira work item as a Smart Link, and choose Embed. Atlassian also lists Slack channels as a place to embed them. If the artifact changes, the embedded view updates too, and renaming it doesn't break the link.
• It works without Rovo. It is available on all paid plans and works independently of Confluence or Jira licenses.

The app is in beta, so features may change. There are also a few technical limits worth knowing before your team builds anything complex:

Source: Atlassian Support, Add an artifact.

How sharing and governance work for AI-generated artifacts

Artifacts are private by default. Rovo always saves a new artifact as private, but files from other agents may not be, so Atlassian advises checking their privacy settings. There are three sharing levels (Atlassian Support):

Sharing matters because a shared artifact becomes part of your organization's knowledge. It appears in Search and the Teamwork Graph, and Rovo agents can use it as context for future work. A good, accurate artifact makes later answers better. An outdated or wrong one can mislead them, so treat "Open" as publishing, not just sharing.

Two more points for admins:

• Deletion is permanent. A deleted artifact cannot be restored, so agree who can delete shared files.
• Check compliance before storing regulated data. Atlassian's compliance page for its beta platform experiences lists SOC 2, ISO 27001 and data pinning to the current region as supported, and HIPAA as not yet supported (Atlassian Support). Healthcare and public sector teams should confirm the current status for Artifacts before use. Our guide to Atlassian cloud security and compliance covers the wider controls.

Rovo Work mode vs other Rovo Chat reasoning modes

Rovo Chat now has four reasoning modes, and each one takes a different amount of time and credits. Atlassian lists Quick Answers (Basic chat) at 10 credits per billable event. Think deeper and Work mode are Premium AI interactions with variable pricing (Atlassian). Deep research isn't listed by name on the credits page, so check its line in your usage export. Actions that write to your Atlassian apps, such as creating a Jira work item from chat, don't use extra credits (Atlassian Support).

Sources: Compare Rovo Chat reasoning modes, Atlassian Rovo plans and Atlassian, platform announcement.

Because Premium pricing is variable, a team that runs long-running AI tasks every week can use its monthly pool much faster than chat alone. Extra usage billing starts on 3 December 2026, so size Work mode as its own line in your budget. Our guide to Rovo credits per user has a step-by-step sizing method, and our breakdown of Atlassian AI costs shows where credits typically go. For a quick first estimate, try the Rovo credits calculator.

What Atlassian hasn't published yet

Some details your team will want before a wide rollout are not public yet, so plan around them rather than guess.

• The credit cost of a Work mode task. Atlassian lists it as variable Premium pricing, with no per-task estimate.
• Admin settings for the sandbox. Atlassian says admins govern the sandbox, but we couldn't find documentation for the settings yet.
• General availability dates. Work mode has a waitlist, and Atlassian Artifacts is in beta.
• Full compliance coverage for Artifacts. Check Atlassian's Trust Center before you store regulated data.

We will update this guide as Atlassian publishes these details.

What longer AI tasks mean for your team

Long-running AI tasks shift the work from prompting to delegating. Each role feels that shift differently.

For admins. Work mode inherits each user's permissions, so permission hygiene now matters more than ever. If a person can see a restricted space by mistake, a task they start can read it too. Review space and project permissions before you turn Work mode on widely, and decide what network access the sandbox should have.

For team leads. Output arrives as finished work, which makes it easy to forward without checking. Agree a simple rule: a person reviews every artifact before it is shared as Open or embedded in a page others rely on.

For finance and budget owners. Premium interactions are variable, so costs follow usage, not seat count. Set a spending cap and watch the 80% usage alert in the first months.

For knowledge managers. Artifacts become a new content type next to Confluence pages. Decide where they belong, how they are named, and when old ones are archived or deleted, so your Teamwork Graph stays accurate.

If your organization is also adopting third-party agents, our guide to Atlassian AI agents and AMP explains how agent identity and scoped authority fit around these features.

How to get your team ready for Rovo Work

Use this checklist before you open both features to everyone.

1. Confirm availability. Check whether Work mode is enabled on your site, or join the waitlist. Confirm the Artifacts app appears in the app switcher.
2. Clean up permissions. Audit restricted spaces, projects and Jira issue security levels, because tasks run with each user's access.
3. Set sandbox and data rules. Decide on network access, data residency needs, and which data types must not go into artifacts.
4. Agree sharing rules. Default to Restricted, and name who can publish an artifact as Open.
5. Budget for credits. Add Work mode as its own line, set a spending cap and assign someone to watch usage alerts.
6. Start with a pilot. Pick two or three repeatable tasks, such as a weekly status report, and measure time saved and credits used before you scale.
7. Train people to review plans. Show users how to read and edit a plan before approving it. A careful plan review is the simplest way to avoid wasted runs and credits. Our Atlassian training programs can cover this for your teams.

How Empyra helps

Empyra helps you adopt both features with the controls in place from day one. Our Atlassian Rovo services cover:

• Permission audits across Jira, Confluence and Jira Service Management before AI rollout
• Governance rules for sharing, embedding and deleting AI-generated artifacts
• Pilot design with measurable use cases and success criteria
• Credit forecasting and spend controls through our Atlassian usage optimization services
• Custom Rovo agents that work alongside Work mode

Talk to an Empyra consultant about a readiness review for Work mode and Artifacts.

Conclusion

Rovo Work and the Atlassian Artifacts app move Atlassian AI from answering questions to finishing jobs. Work mode does the long, multi-step work in a governed sandbox, and Artifacts gives the output a permanent, permissioned home. The teams that benefit most will be the ones that clean up permissions, agree sharing rules and budget for credits before usage grows. If you want help planning that rollout, contact Empyra.