What’s New in Copilot Studio: September 2025 – Expanding Reach, Automation, and Analytics
The September updates focused on three core areas: building richer agent experiences (specifically via Computer Use and WhatsApp), advanced developer tools like Code Interpreter, and deeper ROI/Analytics for measuring business impact.
Real-time User Journey: Computer Use (Public Preview)
The most innovative journey introduced is the ability for an agent to interact with UIs like a human:
- Scenario Discovery: A user needs to enter data into an old legacy desktop application that doesn’t have an API.
- Instruction: The user tells the agent, “Open the legacy ERP, look up invoice #123, and copy the total into this Excel sheet.”
- Vision & Reasoning: The agent uses its “vision” to see the screen and its “reasoning” to identify buttons and text fields.
- Execution: On a hosted Windows 365 browser or local device, the agent moves the virtual mouse, clicks, and types to complete the task.
- Completion: The agent confirms the data has been moved and provides a screenshot or log of the completed action.
Step-by-Step: How to Enable WhatsApp (General Availability)
With WhatsApp integration now GA, you can deploy agents to customers globally:
- Step 1: Prerequisites: Ensure you have a Meta Business Account and a verified phone number for WhatsApp.
- Step 2: Access Channels: In Copilot Studio, open your agent and go to the Channels tab.
- Step 3: Select WhatsApp: Choose WhatsApp from the list of available channels.
- Step 4: Connect to Meta: Follow the guided wizard to link your Copilot Studio agent to your Meta Business Manager.
- Step 5: Configure Interactions: Set up how the agent should handle attachments (images/files) and authentication (via phone number).
- Step 6: Go Live: Once connected, your agent is accessible to anyone messaging that verified WhatsApp number.
Infographic: Scaling and Measuring AI Impact
The September update introduced a suite of “Enterprise-Ready” tools to manage agents at scale:
| Capability | What it Does | Business Value |
| Code Interpreter (GA) | Executes Python code directly within agents. | Solves complex math, data visualizations, and CRUD operations on Dataverse. |
| ROI Analysis (GA) | Tracks savings in time or money per agent run. | Provides real-time data to justify AI investment and prioritize projects. |
| File Groups (GA) | Organizes up to 12,000 files into 25 groups. | Improves knowledge retrieval accuracy and reduces “context chaos.” |
| Agents Client SDK | Embeds agents in native Android, iOS, or Windows apps. | Keeps users in their flow of work without switching to Teams or Web. |
References
GM Dynamics & Agentic AI Consulting 