02 of 04

Building an agent that runs company-wide

This started as a beta pilot I ran myself. It now processes every release note at Atlassian.

Role

Agent design, prompt engineering, rollout

Result

100% of Atlassian release notes processed by agent. 94% manual effort saved.

0%

of Atlassian release notes processed by agent

0%

manual effort saved

0+

FD tickets processed in the first month

0

release notes missed since launch

Fig. 01 — Agent coverage across every Atlassian release note

The problem

Every feature launch needed a release note. Engineers and PMs wrote them by hand. Every time.

  • The notes had to meet legal requirements for customer notification.
  • They had to match Editor's style and terminology.
  • They had to be published before every rollout.
  • Nobody had time to do it well.

What I did first

I mapped the full process before writing a single line of the prompt.

  • What a good release note required.
  • Where the information lived: Jira tickets and project posters.
  • What the legal obligation actually was.
  • What failure looked like when notes were done badly.

How it works

The agent reads the Jira ticket or project poster. It applies Editor's style, terminology, and Contentful formatting. It generates a reviewable draft in under a minute.

  • Select the agent in Rovo Chat.
  • Paste the FD ticket link or project poster.
  • Review and adjust the output.
  • Paste into the ticket.

What they said

The team adopted it the week it launched.

  • "Gina's Release Note agent has significantly enhanced our ability to generate release notes. It not only saves a considerable amount of time but also optimises the feature gate rollout process." — Leanne Huynh
  • "Gina your release note agent saved my life." — Tammy Gu, Senior Product Designer, Confluence AI
  • "Wow, that tool is so efficient and cool." — Suraj Sonawane, Engineer, Collaboration

Lasting impact

It started with Editor. Now it handles every Atlassian release note. Neither an engineer nor a redesign was required to scale it.

  • 11+ FD tickets processed in the first month.
  • Zero release notes missed since launch.
  • Next step: Jira automation trigger when release components are set.