Atlassian adds AI requirement auto-creation to Jira, routes tasks to Claude, Copilot

Atlassian announced a suite of generative AI capabilities integrated into Jira, fundamentally reshaping how development teams create specifications and assign work. The centerpiece is Jira Planner, which accepts a high-level description of the software to be built and automatically generates requirements. Planner retrieves context from three sources—the existing codebase, Jira's historical records, and Confluence (Atlassian's knowledge-sharing tool)—and uses that context to produce structured requirement documents stored in Confluence. This approach ensures that both humans and AI agents can reference the same specification.

Once requirements are in place, Jira automatically identifies the work items (tasks) needed to execute them. These items can then be assigned to either humans or AI agents. The AI options include Claude Code, Cursor, GitHub Copilot, and a new built-in offering called Jira Coding Agent; OpenAI Codex support is planned for the near future. Critically, Jira retains the full context and serves as the coordination layer—developers can pick whichever AI agent they prefer for a given task without losing sight of the larger project.

Jira also displays the status and activity of each AI agent in a single view, showing what each is working on, whether it is waiting for human input, and more. Beyond coding tasks, Jira's automation rules builder lets teams route routine work—bug fixes, security vulnerability repairs, test generation, documentation updates—directly to AI agents. When a pull request is ready, the system notifies the assigned engineer. Atlassian frames the result as a system where humans and AI agents collaborate, with Jira as the orchestration platform.

Atlassian's update addresses a core friction point in AI-assisted development: keeping context synchronized across multiple tools. By embedding requirement generation directly into Jira and anchoring all task assignments to that single source of truth, teams avoid the context loss that occurs when jumping between Jira, GitHub, IDE-based tools (Copilot, Cursor), and external LLMs. The integration with Claude Code, GitHub Copilot, and Cursor reflects the current fragmentation of AI coding tools—developers often prefer different agents for different tasks—and Jira's new design allows that flexibility without breaking the workflow.

The system also formalizes the role of AI agents as assignees, not just suggestions. This suggests Atlassian sees future development as genuinely mixed human-AI teams, where task routing becomes a first-class concern. The addition of automated routine work (tests, docs, vulnerability patches) via the automation rules builder further reduces friction, since these tasks can now run without explicit assignment. The notification of engineers when pull requests are ready completes the loop—AI does the heavy lifting, humans gate the final decision.

No comments yet. Be the first to share your thoughts!