News from Atlassian Team ’26 Europe: New AI capabilities for regulated industries 

At Atlassian Team ’26 Europe in Amsterdam, Atlassian unveiled the next stage of collaboration between people and AI agents. The spotlight was on developments particularly relevant to regulated industries: shared work context, agent identity management, and language model processing within the EU. These innovations pave the way for broader process automation while presenting organizations with new challenges related to AI security and oversight.  

Financial services, healthcare, pharmaceuticals, energy and telecommunications operate in different environments, but face a common challenge: how to improve work with AI while maintaining control over information access, output quality and accountability for decisions. The announcements in Amsterdam suggest that these considerations are becoming an integral part of Atlassian’s platform development. 

Bringing agents into teams and workflows 

A central announcement at the conference was the Agentic Multiplayer Protocol (AMP). Atlassian describes it as a foundation for collaboration in which agents have their own identities, defined permissions and shared context with the team, and their outputs can be evaluated. The company also announced agent accounts and a registry of non-human identities to help administrators manage access for agents, applications and service accounts. 

For regulated organizations, the ability to connect AI activity to a specific process and its controls is particularly valuable. Implementation also requires organizational decisions: assigning an owner to each agent, defining the actions it can perform, and identifying when human approval is required. 

Shared context and control over data 

The evolving Teamwork Graph connects information across tasks, documentation, communications, code, and assets. The Artifacts app, meanwhile, provides a persistent home for AI outputs that can be shared across the team’s tools. Atlassian also announced Local EU AI Inference, which restricts language model processing to models hosted in the EU. 

Teamwork Graph can help teams draw on organizational knowledge, if sources are up to date and permissions are configured appropriately. Local EU AI Inference addresses requirements around processing location, although location alone does not establish compliance for the solution. Organizations should confirm the availability and scope of these capabilities for their specific environment. 

Retaining an AI output does not create a complete audit trail. Organizations should check whether they can establish who, or which agent, performed an action, what information was provided about it, and who approved it. 

From announcements to a measurable pilot 

A pilot focused on a single process allows an organization to assess the benefits before expanding automation. Useful measures include the time needed to prepare material for approval, the proportion of outputs requiring substantial correction, and the completeness of records documenting actions and decisions. 

“In regulated industries, a good starting point is a process where knowledge sources, agent permissions and approval points are clearly defined. Exceptions need particular attention: missing data, conflicting sources, or tasks that fall outside the agent’s permissions. The pilot should establish when AI needs to hand the task over to a person,” says Maciej Szostek, Competency Center Director at Sii Poland. 

This approach helps organizations develop their Atlassian environment around security and compliance requirements, as well as the needs of service owners. It also allows them to test the value of new capabilities in practice before extending automation to more processes. 

Want to explore which new capabilities could support your organization? Contact the Atlassian Competency Center at Sii Poland to discuss a suitable process, the scope of a pilot and how to measure its results.