Atlassian's Mike Cannon-Brookes: AI agents need a human owner

Cannon-Brookes frames the difficulty of workplace AI less as a model problem than an interface and accountability problem. His comparison is Microsoft Word's toolbar: if only keyboard shortcuts remained, he argues, everyone's documents would end up looking like plain text — most people do very basic things because the interface is poor. Atlassian's answer, the Agentic Multiplayer Protocol, is meant to make an agent's edits appear in a document's history the way a colleague's would, so "you have to be able to see who did what when."

The concern is concrete. Cannon-Brookes describes an employee using company documents to build a visual chart, warning that if it draws on internal material you must be careful who receives it, since recipients may see the same documents and information could leak. That is why he separates consumer agents from workplace ones: in a workplace there is a lot of conversation around permissions, security, and compliance.

The context is Atlassian's own pivot. The Sydney-based company, known for Jira and Confluence, launched the in-house assistant Rovo in 2024 and in March cut its workforce by 10% — roughly 1,600 employees — as part of a shift toward AI and enterprise sales. Cannon-Brookes also casts Australia's role ambitiously: at roughly 1% of world GDP, he says the country needs to be 1% of the world's technology production.