Release Crew AI

Ship with Confidence.

Comment

🚦🤖 Release Crew AI

An AI Technical Program Manager powered by Atlassian Rovo

Inspiration 🧠

Every release looks fine — until it isn’t.

Statuses are green 🟢, sprints are closed, and versions are approved — yet releases still slip. Why? Because the real risks aren’t in Jira fields… they’re buried inside comments, conversations, and unlinked dependencies.

Release managers and tech leads are forced to manually read dozens of tickets to answer questions like:

  • “What’s actually blocked right now?”
  • “Who is waiting on whom?”
  • “Which risks are implied but never tracked?”

Release Crew AI was built to do what humans do best — connect the dots — at machine speed. Using Atlassian Rovo, it turns raw Jira data into release intelligence.

What it does 🚀

Release Crew AI is an Atlassian Forge app powered by a Rovo Agent that acts as an AI Technical Program Manager.

It automatically:

  • 🔍 Scans Jira issues for a specific release
  • 🧠 Reads through comments and descriptions, not just statuses
  • ⚠️ Detects hidden risks like “blocked” or “waiting”
  • 🧩 Builds a dependency network across issues
  • 📄 Generates an executive-ready Release Health Check in Confluence
  • 🔗 Fixes Jira hygiene by creating missing issue links

All without changing how teams work.

How it works 🤖

Release Crew AI Architecture

Core Architecture & Rovo Intelligence

1️⃣ Jira Data Acquisition 📡

Using the Jira Cloud REST API (v3), the agent fetches all issues for a given:

  • Project
  • Release version

For each issue, it gathers:

  • Summary
  • Status
  • Assignee
  • Full comment history
  • Existing issue links

This gives Rovo the full operational context, not just metadata.

2️⃣ ADF Processing (Unlocking Human Conversations) 🔓

Jira stores comments in Atlassian Document Format (ADF) — a deeply nested JSON structure.

Release Crew implements a recursive ADF text extractor that:

  • Traverses complex nodes
  • Preserves meaning and intent
  • Produces clean plain text

This allows the Rovo Agent to reason over real human discussions, not just fields.

3️⃣ Hidden Risk Detection (Rovo Reasoning) ⚠️

This is where Rovo shines.

The agent applies a domain-specific risk rule-set and flags issues as High Risk when comments contain signals like:

  • “blocked”
  • “waiting”
  • “stuck”
  • “cannot start”

👉 Even if the Jira status says “In Progress” or “Done”.

Each risk includes:

  • Evidence from comments
  • Impact on the release
  • Visibility in the final report

This turns Rovo into an early warning system for releases.

4️⃣ Dependency Intelligence & Visualization 🧩

Release Crew builds a non-linear dependency graph capturing:

  • One-to-many relationships
  • Many-to-one bottlenecks
  • Cascading blockers

The Rovo Agent generates branching Mermaid.js diagrams embedded in Confluence using an ADF extension macro.

Risk levels are visually encoded:

  • 🔴 High Risk
  • 🔵 Normal
  • ⚪ Informational

Rovo-Powered Workflows & Actions ⚙️

🔍 get-issues

  • Fetches complete Jira issue data
  • Normalizes ADF content
  • Feeds structured input to the Rovo Agent

📄 create-confluence-page

  • Automatically creates a Release Health Check page

  • Includes:

    • Executive Summary
    • Color-coded Mermaid dependency graph
    • Detailed risk table with evidence

  • Designed for leadership and release reviews

🔗 create-jira-links

  • Parses structured dependency data
  • Programmatically creates missing Jira issueLinks
  • Converts implicit blockers into explicit Jira relationships
  • Improves long-term data hygiene

Why this is a Best Rovo App 🏆🤖

Release Crew AI uses Forge rovo:agent as an autonomous analytical system, not a chatbot.

The agent is:

  • 🚀 Triggered to analyze immediately (no manual “proceed” prompts)
  • 🧠 Designed for deep reasoning over unstructured data
  • 📄 Responsible for generating real, persistent artifacts (Confluence pages)
  • 🔧 Capable of taking corrective action (Jira link creation)

This demonstrates Rovo’s power as an embedded AI teammate that:

  • Thinks
  • Explains
  • Acts

Challenges we solved 🧗

  • Extracting signal from noisy Jira comments
  • Reasoning over ADF safely and reliably
  • Visualizing complex dependency graphs clearly
  • Making AI decisions transparent and auditable

Accomplishments that we're proud of

  • 🤖 Built a production-grade Rovo Agent that performs real analytical work
  • 🔍 Surfaced risks that traditional Jira dashboards completely miss
  • 📄 Automated executive-level reporting in Confluence
  • 🧩 Transformed unstructured conversations into structured release insights
  • 🔗 Improved long-term Jira data quality through automated linking

What we learned

  • The most valuable release signals are often unstructured
  • Rovo Agents are most powerful as reasoning systems, not chatbots
  • Explainability is critical for trust in AI-driven decisions
  • Confluence can act as a durable system of record for AI insights
  • Teams adopt AI faster when workflows remain unchanged

What's next for Release Crew AI 🚀

  • 🧠 Predictive risk scoring using historical release data
  • 🔁 Continuous monitoring of active releases
  • 📊 Release risk trend dashboards
  • 🧩 Expanded Rovo actions for guided remediation
  • 🔗 Deeper integrations across Atlassian products

Why Release Crew AI matters 🎯

Release failures rarely come from unknown problems — they come from known problems that weren’t connected in time.

Release Crew AI:

  • Surfaces hidden risks early
  • Connects conversation to action
  • Gives teams confidence before release day

It transforms Jira from a tracking system into a predictive, Rovo-powered release intelligence platform.

Share this project:

Updates