Atlassian
Jun 2026
Atlassian Agents
────────────────────────────────────────── · · ✿✿✿ · · ──────────────────────────────────────────
.overview
Managing AI agents from one place
Agents is a feature in Atlassian Administration that helps admins manage AI agents across their organization. Agents are AI-powered virtual teammates that automate tasks, solve complex problems, and drive productivity across the Atlassian platform. While Atlassian teams are building new AI agent experiences across Atlassian products, Enterprise agent adoption is slowing due to the lack of admin controls.
.my role
I was the lead designer responsible for designing the end-to-end agent management experience, ensuring admins have the tools and confidence to deploy, govern, and scale agents securely and efficiently. I worked alongside Content design, Product, Engineering, Central AI, Studio, and Identity teams and I owned the research, product design, and final delivery of admin experience for agent management.
.impact
Enterprise adoption
Delivered features and controls required for adoption in regulated and security-conscious industries, unlocking new market segments for Atlassian.
Platform Extensibility
Enabled integration of out-of-the-box, custom, and third-party agents, supporting Atlassian’s System of Work and Collections strategies.
Centralized Agent Management
Designed the admin experience for managing all types of agents (Rovo, Studio, Forge, JSM) in Admin Hub, providing visibility, access controls, and governance at both org and unit levels.
Granular Permissions & Security
Defined agent identity and permission models, supporting both autonomous and user-context agents, with robust controls for app access, group membership, and delegated consent.
────────────────────────────────────────── · · ✿✿✿ · · ──────────────────────────────────────────
.problem
Governing agents at the organization level was not possible
33% of enterprise customers have turned “off” some or all of their AI reflecting AI trust & safety is of particular concern. Enterprises require robust governance and the lack of fine-grained controls is a major blocker for enterprise rollout. With the implementation of AI agents across the Atlassian platform, it is critical for admins to govern, monitor, and evaluate agents operating across their organization at scale.
01
Fragmented visibility and weak controls
Without unified visibility, admins can’t effectively monitor agent activity, identify potential risks, or ensure compliance
02
Insufficient controls over what agents can do
Governance is essential for admins ot manage agent permissions, data access, and actions agents can perform
03
Agent sprawl
Admins require tools that reduce manual configuration, cost management, and repetitive review in order to manage agents at scale
────────────────────────────────────────── · · ✿✿✿ · · ──────────────────────────────────────────
.designs
Creating an agent management experience centrally managed at the organization level
Summary statement walking through the key core flows
Our strategy for this feature is made up of 3 pillars: visibility to address agent discoverability, governance to empower admins to set boundaries and controls, and scale to ensure minimal increase of admin overhead.
Agent directory
A single pane of glass for admins to view all agents across all agent types, Rovo, Studio, Forge, and JSM, at both organization and unit levels. This centralized directory provides organization level visibility over all agents.
Default access controls
Admins can manage access at scale by configuring the default access to apps and groups for future agents. I applied existing access management patterns, resulting in a consistent, intuitive experience extended to agents.
Agent profile
Admins can configure granular access per agent via the agent profile by managing app and group access or adding default access controls.
────────────────────────────────────────── · · ✿✿✿ · · ──────────────────────────────────────────
.key decisions
Defining how agent access should work
We needed to determine how AI agents gain access to apps and groups. In early explorations, we adopted an existing pattern access controls called Automatic access to apply access to all apps and groups to agents automatically. However, this design came with a number of risks.
Admins pushed back against allowing blanket access to all apps for agents as it posed major security and compliance concerns for enterprise customers. Admins wanted governance controls that allowed them to specify what agents had access to within their Atlassian environment.
As an enterprise.Someone managing a tool set for an enterprise, that's great you're adding these new features, and we'd love to turn them on, but let us assess each one as they come, and I don't want that on by default… AI's a bit of a problem for us, it actually needs more governance and control. We need toggled features to be able to determine what parts of this we want to turn on and off.
Interview participant
Enterprise Organization Admin
Initial design using automatic access to provision agent access to apps and groups
Our resolution was to implement guardrails to prevent unauthorized access to apps and sensitive data, allowing admins to control agents' access by configuring default access. This enables admins to establish agent access to a specific scope of apps and groups, providing admins with the granular control they expect for AI configuration.
Future proofing for lifecycle management
This project required scheduling alignment with the Studio team, who were responsible for building the agent builder experience. The Studio team wasn't able to deliver agent lifecycle management in their experience, prompting our team to focus on establishing the model for agent management that was platform extensible. Our engineers built in statuses that can hook up to Studio's backend services once they've completed implementation of lifecycle actions on their end.
Agent list and profile displaying status
────────────────────────────────────────── · · ✿✿✿ · · ──────────────────────────────────────────
.lessons
Building for clarity and cohesive agent workflows
AI programs span multiple teams that are building fast and sometimes with reckless abandon. A major learning from this project was to request a program/project manager as soon as possible and a core review team to ensure a cohesive experience across Atlassian. With so many teams building agentic experiences, there was a lot of time spent trying to wrangle who was building what, align on patterns, terminology, strategy, and timelines.
────────────────────────────────────────── · · ✿✿✿ · · ──────────────────────────────────────────
