Microsoft Unveils Autonomous AI Agents with Identities and Permissions

Microsoft is nearing the launch of a new generation of AI agents capable of moving beyond mere assistance to autonomously execute tasks within enterprise systems, endowed with their own identities and permissions. The company announced the development of its Copilot platform, introducing a new agent named "Autopilot." This agent is described as persistent, proactive, and personalized, with the ability to continue performing tasks in the background independently. This shift extends beyond nomenclature; the latest update to Copilot allows for the delegation of complete tasks to "Cowork," while the "Code" tool facilitates the creation of applications and dashboards using natural language commands. "Cowork" goes further than text generation, capable of performing multi-step tasks such as searching within organizational resources, preparing documents, managing files, arranging meetings, and drafting emails, contingent on user approval for sensitive actions. "Autopilot" adds an element of continuity, continuing to carry out work in the background even when the user is not actively at their computer. The most significant development is that these AI agents are no longer confined to chat interfaces. Microsoft is integrating "Agent Identities" within its Entra system, granting each agent an independent digital identity that defines its nature, the data it can access, and the actions it is authorized to perform. Enterprise administrators can monitor the agent's activity, define or revoke its permissions, and disable its identity when necessary. This presents a striking parallel to human employees, who possess a corporate account and identity, specific file access permissions, and are subject to security policies. The digital agent is moving in the same direction, though legally and technically it remains a program rather than an employee in the traditional sense. This transformation raises fundamental questions about responsibility: Who is accountable if an agent sends information to the wrong person? Who determines which files it can read? And who bears the consequences if an unintended action is taken? Consequently, Microsoft is prioritizing identity and governance alongside the enhancement of AI capabilities. The deeper question concerns the future of human employees themselves. While the first generation of AI assisted with writing and then analysis, the new agents are now moving into the phase of executing tasks autonomously. This shifts the traditional discussion about AI taking human jobs to a more pragmatic question: "What part of our jobs will we still be responsible for performing?" Although Microsoft has not declared the birth of a legal employee, granting AI an identity, permissions, and the ability to execute continuous tasks within organizations suggests that the lines between "tool" and "colleague" are becoming increasingly blurred.