TechPrufer
Get started
Atlassian
Atlassian logoAtlassian · Feature trail

AI Governance Platform

8 posts · Jan 2025 – Sep 2026

What this trail covers

Atlassian's AI Governance Platform, including Rovo Search, is enhanced with enterprise-ready AI-driven search, interactive chat, and intelligent agents. This post details the platform's approach to AI governance, focusing on securing underlying data, setting boundaries for autonomous behavior, and maintaining a record of AI actions. It introduces controls for data classification, content scanning, agent permissions, tool controls, and audit logs, aiming to provide centralized oversight and risk management. BearQ for Jira is introduced as an assignable agent that integrates directly into the Jira workflow to perform intelligent, end-to-end testing. BearQ parses Jira work items to author test cases, explore user journeys, and adapt tests as applications evolve. It creates Jira issues with reproduction steps and evidence when regressions or issues are found, aiming to scale QA capacity while retaining team oversight.

All Atlassian trails
  1. Autonomous testing with Jira and BearQIntroduces BearQ for Jira, an assignable AI agent that integrates with Jira work items to perform end-to-end testing. BearQ parses acceptance criteria and requirements from Jira tickets to author test cases, explore application behavior, and adapt tests dynamically. When regressions or issues are detected, BearQ creates actionable Jira issues with reproduction steps and evidence. This extends the AI Governance Platform by providing an agentic testing capability directly within the delivery workflow.
  2. Don't Build AI Governance From Scratch. Just Turn It On.This post details Atlassian's strategy and platform capabilities for AI governance, emphasizing data security, agent behavior boundaries, and auditability. It introduces Atlassian Guard for data classification and scanning, Rovo connector and chat security for prompt/response scanning, and agent-specific controls like permissions, identities (upcoming), and MCP tool controls. It also highlights audit logs and Rovo Trends for visibility and reporting. The core contribution is framing AI governance as an integrated part of the Atlassian platform, leveraging existing controls and introducing new agent-specific ones.
  3. Rovo Search: Why Rovo is a Leading Enterprise AI Search ToolThis post details several key technical advancements in Rovo Search: 1. AI-generated answers with citations, synthesizing information from multiple sources to answer natural language questions. 2. Intent-aware routing within Jira, directing users to either traditional Jira search or Rovo Search based on query type. 3. Expansion of the connector ecosystem to include GitHub, Google Drive, and others, enabling cross-tool search. 4. Richer UI components and result cards for improved context and actionability. 5. Deeper workflow integration, embedding search capabilities directly into Jira and Confluence.
  4. Secure AI adoption with data loss prevention (DLP)This post details the integration of Data Loss Prevention (DLP) into Atlassian's AI governance platform. It explains how DLP is used to discover sensitive data through content scanning (full-site scans and real-time), classify data with labels (Public, Internal, Confidential, Restricted) using automatic rules, and control data exposure through policies. Specifically for AI, it highlights how Rovo Chat security blocks sensitive data from prompts/responses and how connector data security scans third-party data before ingestion into the Teamwork Graph. AI controls are in early access.
  5. How We Brought Enterprise-Managed Authorization to Rovo MCP with XAA and ID-JAGThis post details the implementation of Enterprise-Managed Authorization for Atlassian Rovo MCP using Cross-App Access (XAA) and Identity Assertion JWT Authorization Grant (ID-JAG). The core technical contribution is the integration of an enterprise IdP to broker authorization decisions upfront, eliminating per-user OAuth consent screens. The flow involves the client exchanging its identity assertion with the IdP for an ID-JAG, which is then presented to Atlassian OAuth via the JWT bearer grant. Atlassian OAuth validates the ID-JAG (signature, issuer, type, audience, client continuity, freshness, resource/scope) and mints Atlassian access tokens. Key implementation details include maintaining protocol boundaries, correctly validating ID-JAGs, resolving users across trust domains by mapping verified identity claims to Atlassian accounts, and providing central administration for connector and client authorization.
  7. Unraveling Rovo search relevanceThis post details the technical implementation of Rovo search relevance, including the use of OpenSearch with BM25 and KNN for initial retrieval, a Gradient Boosted Decision Tree model for L1 reranking, and a multi-task trained cross-encoder and DCN model for L2 semantic and behavioral ranking. It also describes the L3 interleaving mechanism based on product affinity and query intent. The Smart Answers workflow is explained, including query classification and rewriting with enriched context. The post also touches upon the federated search approach for certain third-party applications.
  8. How Rovo Deep Research works - Inside Atlassian How Rovo Deep Research worksThis post introduces 'Deep Research', a novel feature within Rovo Chat that enhances the AI Governance Platform by enabling in-depth information retrieval and comprehensive report generation for complex queries. It details a two-stage RAG architecture: query-dependent retrieval and content-grounded answer generation. The system decomposes complex tasks into sub-tasks for concurrent information retrieval and implements structured report generation by creating an outline, developing sections in parallel, and integrating them. It leverages Atlassian's Teamwork Graph for contextual data and uses a diverse array of LLMs (Llama 3.2 8B, GPT-4o, GPT-4.1, Claude 3.7 Sonnet) for different stages of the process. The post also outlines Atlassian's quality measurement framework using automated side-by-side and reference-free evaluations, and details how challenges like limited insight depth and insufficient source attribution were addressed through structured report generation and inline citations.
  9. Rovo drives a new era for conversational AIThis post details recent improvements to Rovo Chat, an AI-powered conversational interface integrated with Atlassian's Teamwork Graph. Key technical contributions include the implementation of 'Intent Routing,' which analyzes prompts to determine whether a quick search or a deep graph query is more appropriate, based on references to work, teams, and goals. This routing logic leverages the Teamwork Graph to ground answers in organizational context. Additionally, the post describes 'Personalized Summaries' that filter and group updates by team or project, and a clarification mechanism for ambiguous prompts. These features aim to improve the relevance, accuracy, and actionable nature of AI-generated responses.
  10. How Atlassian continuously improves Rovo Search qualityThis post details how Atlassian continuously improves Rovo Search quality by establishing a primary online search quality metric (composite of long clicks and AI answer engagement) and tracking supporting metrics. It highlights recent advancements including improved ranking models incorporating user signals, increased AI answer coverage and source citation, P90 latency reduction via LM/serving infrastructure upgrades, and enhanced Google Drive recall with permission checks. The post emphasizes transparency in measurement and improvement processes.
  11. Rovo Chat: AI that works where you workThis post details enhancements to Rovo Chat, an AI experience integrated across Atlassian products. New capabilities include file upload support for over 40 file types, enhanced formatting for pasted answers, and deeper integration with the Teamwork Graph for context-aware responses tailored to user roles. Rovo Chat can now perform actions like creating, assigning, editing, and transitioning Jira issues, drafting and publishing Confluence pages, and adding items to sprints, all with user confirmation. The 'Rovo Deep Research' feature provides structured, in-depth reports with citations and actionable recommendations.
  12. Introducing Rovo Deep Research, grounded on your dataIntroduces Rovo Deep Research, a new AI skill for Rovo that autonomously breaks down complex questions into research tasks, searches across Atlassian apps and connected SaaS apps (leveraging the Teamwork Graph), and generates comprehensive, cited reports. The skill aims to provide insights grounded in full business context, enabling proactive decision-making. It is integrated into the existing AI Governance Platform.
  13. Introducing Atlassian's Remote Model Context Protocol (MCP) ServerIntroduces the Atlassian Remote MCP Server, a new capability that acts as a bridge to connect Atlassian's Teamwork Graph data (Jira, Confluence) to external AI tools via the open Model Context Protocol (MCP). This enables AI assistants like Claude to interact with Atlassian data for tasks such as summarizing work items, creating issues/pages, and performing multi-step actions. The server is run by Atlassian, ensuring security with OAuth authentication and existing permission controls. It leverages Cloudflare's Agents SDK for OAuth and MCP support. This post extends the AI Governance Platform by enabling secure, permissioned access to Atlassian data from third-party AI agents.
  14. Bringing the magic of human-AI collaboration to every teamThis post announces the broad availability of Rovo, Atlassian's enterprise AI solution, across all customer tiers, starting with Premium and Enterprise. It highlights the integration of Rovo Search, Chat, and Agents into existing subscriptions. Key technical contributions include the introduction of Rovo Studio, a new app for building custom AI agents, automations, assets, and hubs with low-code/no-code options. The post also details enhancements to Rovo Search with more connectors (50+), improved relevance, and personalization powered by the Teamwork Graph. Rovo Chat gains 'Deep Research' capabilities for content generation and information gathering, and mobile integration. New Rovo Agent collections are introduced, including 'Teamwork Collection' agents for Jira and Confluence workflows, 'Dev AI agents' for developers (e.g., Code Reviewer Agent), and service delivery agents like the upcoming Root Cause Analyzer. The underlying Teamwork Graph is emphasized as the data intelligence layer connecting AI workflows and enterprise data.
  15. Enhancing Rovo Chat with Hybrid LLM ApproachThis post details the technical implementation of Rovo Chat, an AI-powered assistant, focusing on its hybrid LLM approach. Key contributions include: 1. Query Rewriting: Two-stage process (semantic and keyword queries) using smaller LLMs like LLaMA-3 8B, achieving +10% answer precision and -17% latency. 2. Accurate Information Retrieval: Multi-plugin routing using a BERT-based intent classifier and parallel searches, with LLM plugin accuracy ranging from 73.1% to 94.2%. 3. Plugin Helpfulness Verification: Employing a Small Language Model (SLM) for relevance checks, achieving 88% to 98.9% accuracy. 4. Generating Accurate Responses: Evaluating 6 LLMs for answer generation accuracy (81.3% to 93.7%) and implementing automatic prompt tuning. 5. LLM Judge for Quality Assessment: Using LLM judges for scalable response evaluation, achieving 79% to 95% agreement with human judgments.