# Atlassian CEO Says AI Agents’ Workplace Actions Need a Human Owner By Simon Yoon Canonical URL: https://www.tokenpost.com/news/technology/34660 Published: 2026-10-19T07:58:40.000Z Updated: 2026-10-19T07:58:40.000Z Section: Technology > Atlassian’s Agentic Multiplayer Protocol gives agents separate identities, limited access and activity records, with audit trails and human review steps. Atlassian CEO Mike Cannon-Brookes said companies need to make AI agents’ workplace actions visible so employees can track delegated work and remain responsible for what the systems do. “It has to be with the end user, ultimately,” Cannon-Brookes said. Atlassian introduced its Agentic Multiplayer Protocol (AMP) in October to bring agents into team workflows. The framework gives each agent a separate identity, limits its access and records its activity, with audit trails and steps for human review. Teams can trace which employee initiated an agent, what information it consulted and what changes it made in Atlassian products or connected services. The protocol builds on Atlassian’s effort to make agents governed participants in workplace collaboration, described in [earlier coverage of AMP’s role in team collaboration](). Cannon-Brookes has also warned that employees could expose internal information by sharing material based on company documents without first checking who can access it. He said interfaces that obscure controls can discourage workers from using more capable AI features. Visible permissions and activity records can help teams review agents’ work and keep responsibility with the employee directing it. ## Links in this article - [earlier coverage of AMP’s role in team collaboration](https://www.tokenpost.com/news/technology/27268)