Breakout
Thursday, November 26, 2026, 4:00 PM - 4:30 PM at Pennar

The next move : AI-powered work intelligence with Atlassian

When an engineering change order lands on the board, the change manager has to answer three questions fast: what will this change break, who needs to approve it, and what does it cost if that analysis is wrong. Today those answers are assembled manually over days, and the awkward ones surface at the end of execution — a last-time-buy window nobody tracked, committed inventory that is now obsolete, a quality plan that no longer matches what is being built.

Change Ripple AI answers those three questions before the change order is approved. It is built on Atlassian Forge, with a deliberate split: deterministic computation runs as Forge functions, and a model layer handles inference and approver derivation, traversing the Teamwork Graph to find where the superseded part is actually used. The underlying structure is ordinary Atlassian — a Product Master project and a Change Control project in Jira with work types for change orders, products and purchase orders, and Confluence as the golden source for part specifications, certification records, quality inspection plans and service manuals. The change manager runs it from a panel on the change control board, or as an agent in Rovo Studio, and the analysis writes derived approvers back onto the Jira ticket rather than leaving them in a report nobody actions.

  • Why the deterministic/inference split is the design decision that matters. Forge functions compute what the data can prove; the model layer infers only what it cannot. This is what makes an AI agent acceptable on a regulated change process rather than merely impressive.
  • How to structure change control in Atlassian so prediction becomes possible. Jira work types for change orders, products and purchase orders, with Confluence as the golden data layer for specifications, certifications and quality plans — the data model is the prerequisite, not the AI.
  • How to collapse days of manual impact analysis into an answer at the point of approval. Including derived approvers written straight back to the Jira ticket, so the people who need to act are named rather than hunted down.
Tags
Products & Solutions
Teamwork Collection, Teamwork Graph, Rovo, Jira, Rovo Dev