Atlassian puts team context at the centre of its AI push
Atlassian is expanding AI across planning, coding, review and incident work while keeping human judgement at the centre of its software-development pitch.
Atlassian is turning its AI-native software-development playbook into product updates across Jira, Confluence, Loom, DX and Jira Service Management, the company said during Team ’26 Europe. 1
Context is the starting point
Atlassian says its Teamwork Graph connects code, documents, decisions and standards with more than 80 third-party sources in a permissions-aware layer. The company claims internal testing showed up to a 44% improvement in answer quality and up to 48% fewer tokens used when agents have that context. 1
The company also announced Code Context, which is designed to help developers and AI coding agents understand large, multi-repository codebases. It links code relationships with related Jira work, Confluence pages, Loom videos and connected services including Slack and Google Drive. 1
From planning to production
Atlassian’s broader pitch is that AI should support the full software-delivery lifecycle, not just generate code. Its playbook describes a model in which people set direction and make judgement calls while agents handle more of the execution and teams learn from the results. 1
TechRadar’s reporting from the conference adds that Atlassian executives are framing this as “multiplayer” AI: tools designed for teams rather than isolated individual productivity. The company also discussed Loom video generation from pull requests and Rovo agents that categorise defects and reduce duplicate reports. 2
What happens next
Atlassian says Code Context is now shipping, while its Planner product is listed as coming soon. The company’s announcements do not amount to a promise that every software task can be automated; they keep human review and governance in the loop. 1
Our opinion
The interesting part of Atlassian’s strategy is not another coding assistant. It is the attempt to make context and accountability the centre of AI-assisted work. That is a more credible enterprise pitch than pretending a model can replace the team around it, although the company’s own performance figures still need independent testing before they mean much to customers.