Microsoft introduces approval process for copilot AI agents

Microsoft is rolling out a new approval mechanism for Copilot AI agents starting in September. The update, part of Copilot Studio, requires human approval before AI agents can perform specific actions. The platform allows developers to set up approval processes for each tool and agent, pausing automated tasks and prompting users to approve, reject, or continue the action. The change applies to three key functions: sending emails, closing tickets, and processing payments. Once configured, AI agents cannot proceed without human intervention. The update aims to enhance control and security over AI-driven workflows. IT Home reported the news, citing a post from NeoWin on September 3. Microsoft’s move reflects a growing emphasis on governance in AI deployment, ensuring critical operations are verified before execution. The approval system is designed to provide transparency and oversight, allowing users to understand the intent behind AI actions. By requiring manual verification, Microsoft seeks to mitigate risks associated with unmonitored automation. This development aligns with broader industry trends toward responsible AI practices. The change is expected to impact businesses relying on Copilot for task automation, as they will need to adjust workflows to incorporate the new approval steps. Microsoft has not provided specific timelines for full implementation, but the update is set to begin in September.