Unofficial update digest
An unofficial daily digest that organizes public Microsoft 365 Copilot, Copilot Studio, and Agent Builder updates.
Between October 27 and 31, 2025, GPT‑4o will be retired in Copilot Studio for agents using generative orchestration, except for GCC customers who can continue using GPT‑4o. The new default model is GPT‑4.1, which delivers improved performance, reliability, and consistency across...
(Preview) Group related user questions into themes and drill down into analytics to uncover patterns and gain deeper insights.
Track time and cost savings for both autonomous and conversational agents to measure ROI and optimize performance.
(Preview) Evaluate your agents using customizable test sets—whether uploaded, manually created, or AI-generated. Test sets can include test cases using different test methods (graders) measured against defined reference answers, helping teams identify strengths and areas for imp...
(Preview) Accelerate flow execution to minimize timeouts and deliver a faster, smoother user experience.
Between October 27 and 31, 2025, GPT‑4o will be retired in Copilot Studio for agents using generative orchestration, except for GCC customers who can continue using GPT‑4o. The new default model is GPT‑4.1, which delivers improved performance, reliability, and consistency across...
(Preview) Group related user questions into themes and drill down into analytics to uncover patterns and gain deeper insights.
Track time and cost savings for both autonomous and conversational agents to measure ROI and optimize performance.
(Preview) Evaluate your agents using customizable test sets—whether uploaded, manually created, or AI-generated. Test sets can include test cases using different test methods (graders) measured against defined reference answers, helping teams identify strengths and areas for imp...
(Preview) Accelerate flow execution to minimize timeouts and deliver a faster, smoother user experience.
Learn about Copilot Studio Kit, a suite of tools developed by the Power Customer Advisory Team (Power CAT) to help test custom agents, validate AI-generated content, analyze conversation key performance indicators, and more.
Access a unified activity and transcript view, pin sessions, and submit feedback for faster, more effective troubleshooting.
Use the Model Context Protocol (MCP) server to access dynamic, real-time content—such as files, database records, and API responses—for richer context and improved agent responses.