TL;DR
AI in work management platforms does four useful things: finds information across tools, summarises and drafts, creates and breaks down work, and increasingly acts as an agent that can be assigned tasks. In Atlassian's platform that's Rovo, included in paid Cloud plans with a monthly credit allowance; extra usage is billed from 3 December 2026. AI is only as good as the data and permissions underneath it, so the work that matters most is cleaning up oversharing, connecting the right sources, and deciding which agents may act on their own.
When we first wrote about AI in work management in 2024, most of it was promise. In 2026 it's in the product: Jira and Confluence summarise, draft, plan and search, and AI agents can be assigned work like a colleague. The question is no longer whether to use it, but how to get value without creating new risks.
What AI does in work management today
1. Finds information across tools
Search that understands questions, not just keywords, across Jira, Confluence and connected tools. Rovo connectors bring in content from tools such as Google Drive, SharePoint, Microsoft Teams, HubSpot and GitHub, so "what did we decide about the pricing change?" can find the answer wherever it lives.
2. Summarises and drafts
Summaries of long work items, comment threads and pages, so nobody has to read the whole history. Drafts of status updates, release notes, meeting notes and knowledge articles that a person then edits.
3. Creates and breaks down work
In Jira, AI can create work items from a Confluence page, a Slack thread or an email, and break a large piece of work into tasks. It can also turn a plain-language question into a JQL search, or a description into an automation rule.
4. Acts as an agent
AI agents can now be assigned work in Jira, like a colleague. Atlassian provides ready-made agents, such as one that checks whether a work item is ready to start, and AI can draft status updates from chat. Rovo Studio lets admins and power users build their own agents and automation flows in natural language, with chosen knowledge sources and skills.
And in the other direction, the Rovo MCP Server, generally available since February 2026, lets external AI tools such as ChatGPT, Claude and GitHub work with Jira and Confluence content.
Use cases that earn their keep
| Team | Use case |
|---|---|
| Leadership | Weekly portfolio summaries drafted from Jira updates, for a person to check and send |
| Project and product | Turning a brief or requirements page into a set of work items; checking items are ready before a sprint |
| Marketing | Breaking a campaign brief into tasks; drafting copy variations; summarising feedback |
| IT and service | Summarising incidents and finding similar past ones; drafting knowledge articles from resolved requests |
| New starters | Asking questions in plain language and getting answers from Confluence and connected documents |
Where AI falls short
- Bad data in, bad answers out. Outdated Confluence pages and half-filled work items produce confident but wrong answers.
- It sees what users can see. Rovo respects permissions, which is right, but also means it surfaces old pages that were shared too widely and never cleaned up.
- It only knows what's connected. Without connectors or integrations, AI in Jira can't see the customer data in your CRM or the budget in your finance system.
- Agents need boundaries. An agent that can create and change work items needs clear limits and review, especially early on.
What it costs
Rovo is included in paid Atlassian Cloud plans, but deeper use is metered in credits, pooled across the organisation and reset monthly:
- Jira or Confluence: 25 credits per user per month on Standard, 70 on Premium and 150 on Enterprise.
- Teamwork Collection or Service Collection: 250, 700 and 1,500 credits per user per month on the same plans.
- Free to use: Rovo Search, in-line summaries and rewrites don't consume credits. Rovo Chat, agents and other deeper interactions do.
From 3 December 2026, usage above the allowance is billed. Before rolling out agents widely, look at your organisation's usage in admin and set expectations with teams.
Governance and data
- Models and training: Rovo uses models from OpenAI, Anthropic, Google and open-source providers, and Atlassian says prompts and responses aren't used to improve them.
- Data residency: Rovo is among the products covered by Atlassian's data residency, so Australian organisations pinning data to Sydney can include it.
- Connectors and external AI: decide which connectors to enable and which external AI clients may connect through the MCP server, and who approves new ones.
- Automated decisions: if agents make decisions that significantly affect people, the Privacy Act's automated decision rules apply from 10 December 2026.
A rollout plan
- Clean up permissions on sensitive Confluence spaces and Jira projects before switching AI on.
- Start with search and summaries, which help everyone and don't consume credits.
- Connect the sources that matter, such as SharePoint or Google Drive, so answers are complete.
- Pilot agents with one or two teams on well-defined tasks, with people reviewing outputs.
- Measure and expand: time saved, quality of outputs, credit usage. Expand what works and stop what doesn't.
Beyond Atlassian
Other platforms are moving the same way. Microsoft 365 Copilot works across Planner, Teams and SharePoint; Salesforce's Agentforce brings agents into the CRM. Each works best on its own data. The gaps between them, such as a Jira agent that can't see the customer in Salesforce, are integration problems, and that's usually where the next round of value sits.
How Great Minds helps
Great Minds is an Atlassian Silver Solution Partner and an integrator first. We roll out Rovo with the right permissions, connectors and governance, build agents for specific workflows, and connect Atlassian to the CRM, finance and other systems AI needs to see. See our Rovo consulting service.
Frequently asked questions
What is Rovo?
Atlassian's AI, built into its Cloud apps. It searches across Atlassian and connected tools, summarises and drafts, helps create and break down work, and runs agents that can be assigned tasks.
Is Rovo included in Jira and Confluence?
Yes, in paid Cloud plans, with a monthly credit allowance per user. Search and in-line summaries don't use credits. It isn't included in the Free plan.
Can AI agents do work on their own in Jira?
Yes. Agents can be assigned work items in Jira. Start with low-risk tasks and keep people reviewing outputs until you trust the results.
Is our data used to train Atlassian's AI?
Atlassian says prompts and responses aren't used to improve the models Rovo uses.
Where to start
Ask three teams what information they spend the most time hunting for, and what they write most often. Those two answers point to the first AI use cases worth piloting. We're happy to help you set them up.