New Copilot Studio Feature to Introduce AI Agent Building Tools
Microsoft has announced plans to roll out a November public preview of a new feature within Copilot Studio, allowing users to create autonomous AI "agents" designed to handle routine tasks. The move is part of Microsoft's broader push to help organizations automate workflows through more agile, programmable versions of its Copilot chatbot. In a blog post, Microsoft 365 chief marketing officer Jared Spataro called these agents "the new apps for an AI-powered world."
"Every organization will have a constellation of agents," Spataro predicted, "ranging from simple prompt-and-response to fully autonomous. They will work on behalf of an individual, team or function to execute and orchestrate businesses process."
Launched last year, Copilot Studio connects the Copilot AI to an organization's various data sources, from the Microsoft Graph to Dynamics 365. With November's public preview, users will be able to build their own agents within Copilot Studio and train them on those same data sources.
The agents will be able to run on a variety of AI models, including OpenAI's "o1" model, which is optimized for reasoning through complex problems. Agents are meant to run largely without human intervention. They can be scheduled to run at set intervals, or programmed to perform actions in response to specific triggers from other applications or tools.
Though agents act autonomously, Microsoft users will be able to view their internal logic for each task that they resolve. Knowing why an agent reacts to a trigger will help users reinforce correct behavior or debug undesired behavior. Additionally, an "Activity" tab will keep a running log of each time an agent is triggered into action, with details such as task completion, decisions and more.
More information about the agent-building capability in Copilot Studio is available in this blog post.