Microsoft Adds Human Approval for Copilot Studio AI Actions

Microsoft plans to roll out changes to Copilot Studio starting in September, adding stricter human approval for AI agents. The platform lets developers build and manage agents that connect to email, tickets, and payment tools. After the update, agents must get human approval before using certain tools.

Developers can set approval switches for each tool and agent. Any action needing approval will pause and show a request, letting users approve or reject it. Microsoft listed examples like sending emails, closing tickets, and handling payments, ensuring these key steps can't be done by AI alone.

This move aims to give businesses more control over AI actions, preventing unauthorized operations and boosting security and compliance. Tech news site NeoWin reported that the feature will roll out gradually, helping organizations balance automation with risk management.