Management and maintenance of AI systems remains a challenge for many companies, especially with the potential for agent proliferation that would expose companies to risky entry points.
Microsoft entered the observability conversation with the launch of Agent 365 at Tuesday’s annual Ignite conference. Agent 365 is described as a control plane for AI agents, serving as an observability layer for enterprises using any agent. The company said the platform “delivers unified observability” with telemetry, dashboards and alerts to track every agent used.
Agent 365 supports any agent, whether built on Microsoft or third-party platforms, including Adobe, Databricks, Cognition, and ServiceNow.
“Agent 365 marks a new chapter in how organizations build, secure and scale their agents. It marks a shift from isolated experimentation to enterprise readiness where agents operate within a unified, managed and productive system,” Microsoft president of business applications and agents Charles Lamanna said in a blog post.
What companies get
Microsoft 365 has five capabilities: enrollment, access control, visualization, interoperability, and security.
The first step in starting observability tasks in Microsoft 365 is to register the agents that can be present, which will serve as a single source of truth. The company calls this registry Entra.
“This single registry allows you to track agents for every important role in your organization – IT, developers, security leaders and business leaders. With the Agent Store, users can easily find the right agents for their roles and workflows directly in Microsoft 365 Copilot and Teams,” Microsoft said.
Access control would require agents to have a unique agent ID, allowing enterprise administrators to restrict access as needed. Organizations can set rules that agents must follow, and the tool can respond by blocking misbehaving agents.
Agent 365 also offers a visual dashboard, allowing companies to see how their agents are connected, measure performance and complete tasks.
For enterprises already struggling with agent proliferation, Microsoft’s approach stands out because it typically combines a collection of add-on tools into a single, managed control plane. Most observability options today are either siled functions or separate platforms, which add complexity. Agent 365 treats agents as part of the stack, providing IT and security teams with unified oversight when needed.
Companies like DataDogDynatrace and Splunk offer observability services for AI systems. Activation Chronosphere provided observability-like capabilities, focusing on debugging issues, and Raindrop Also introduced its own observability productivity tools.
Google has also started offering an observability dashboard in its AI Agent Builder tool.
“As the number and sophistication of agents increases, companies face a new kind of challenge: how to manage agents responsibly and at scale, without having to rebuild the trusted systems they rely on,” Lamanna said. “The best solution is to manage agents the same way you manage people, using the same infrastructure, applications and security that support your business today.”