TechPrufer
Sign inGet started

Blogs›Atlassian›AI-Native Software Development Capabilities in Jira

AI-Native Software Development Capabilities in Jira

AI-Native Software Development Capabilities in Jira

2
posts
2026

This post introduces Rovo Dev Code Reviewer, an AI-driven agent integrated into Bitbucket to automate parts of the code review process. It uses a zero-shot structured prompting approach with LLMs (Claude 3.5 Sonnet) and a multi-stage quality check (LLM-as-a-Judge for factual correctness, ModernBERT for actionability) to provide feedback on pull requests. The system augments human reviewers, aiming to reduce PR cycle time and human-written comments. Internal evaluations show a 30.8% reduction in PR cycle time and a 35.6% decrease in human-written comments, with a 38.70% code resolution rate for AI-generated comments.

2026

How we’re evolving Jira for AI-native software development - Inside Atlassian

7/15/2026

This post details the introduction of several new capabilities within Jira to facilitate AI-native software development. It describes how Jira Planner structures intent into technical specs, how Jira for Slack integrates conversational context, and how Loom video prompts convert visual and auditory instructions into actionable tasks. It also announces the integration of specific coding agents (Claude Code, Cursor, GitHub Copilot) directly into Jira workflows, the development of the Jira Coding Agent for automated routine fixes, and the introduction of agent session visibility. New governance features include coding agent automations and an Agentic Engineering project template, alongside DX AI cost management for financial oversight. The post emphasizes the role of the Teamwork Graph in providing essential context to AI agents, leading to improved accuracy and efficiency.

30.8% Faster PRs: How AI-Driven Rovo Dev Code Reviewer Improved the Developer Productivity at Atlassian - Inside Atlassian

4/7/2026

Details the architecture and implementation of Rovo Dev Code Reviewer, an AI-driven code review agent. It describes the zero-shot structured prompting approach, including persona definition, chain-of-thought, and review guidelines. It also explains the two-stage quality check mechanism: an LLM-as-a-Judge for factual correctness and a ModernBERT-based classifier for comment actionability, trained on a proprietary dataset. The post quantifies the impact through a large-scale online evaluation, reporting metrics on PR cycle time reduction, human comment reduction, and code resolution rates.