Tag: copilot-studio

copilot-studio

General Availability of the Microsoft Copilot Studio Extension for Visual Studio Code

Goloknath

with reference to above link pls provide 1. Topic name 2. Realtime user journey for this feature 3. Step by step how to enable this feature 4. Infographic

General Availability of the Microsoft Copilot Studio Extension for Visual Studio Code

This release brings “Pro-Code” capabilities to agent development, allowing developers to treat AI agents like traditional software by using an Integrated Development Environment (IDE), source control, and CI/CD pipelines.

Real-time User Journey

The extension enables a “local development loop” for building complex agents:

  1. Clone: The developer pulls an existing agent definition from the cloud (Copilot Studio) into their local VS Code workspace.
  2. Edit: Using the IDE, the developer modifies topics, tools, and triggers. They benefit from IntelliSense, syntax highlighting, and the ability to use GitHub Copilot to co-author agent logic.
  3. Review: The developer uses standard Git commands to stage changes, view “diffs” (what changed), and resolve conflicts before pushing updates.
  4. Sync: Once the local edits are ready, the developer “Applies Changes” to sync the local code back to the Copilot Studio cloud environment for testing.
  5. Deploy: The agent definition is checked into a repository (like GitHub or Azure DevOps), triggering an automated pipeline to promote the agent from Dev to Test to Production.

Step-by-Step: How to Enable

To start using the extension, follow these steps:

  • Step 1: Installation: Go to the Visual Studio Marketplace (or the Extensions view in VS Code) and search for/install the “Microsoft Copilot Studio” extension.
  • Step 2: Authentication: Sign in to your Microsoft 365 / Power Platform account through VS Code to link your environments.
  • Step 3: Connect to an Agent: Open the Copilot Studio icon in the VS Code sidebar. Browse your environments and select an agent to “Clone” locally.
  • Step 4: Develop: Open the generated folder structure. You can now edit the YAML-based agent definitions directly.
  • Step 5: Sync Back: After making changes, use the extension’s command palette or sidebar button to “Apply Changes” back to the cloud.

 Infographic: Pro-Code Agent Development

The following table summarizes the shift from “Low-Code” to “Pro-Code” with this extension:

FeatureWeb-Based (Low-Code)VS Code (Pro-Code)
InterfaceVisual Canvas (Drag-and-Drop)YAML / Text-Based IDE
CollaborationLive Presence / Shared CanvasGit, Pull Requests, Code Reviews
AI AssistanceNatural Language DescriptionsGitHub Copilot / Inline Chat
SpeedBest for quick prototypesBest for bulk edits & complex logic
DeploymentManual PublishAutomated DevOps / CI/CD Pipelines

References

Copilot Studio Extension for Visual Studio Code Is Now Generally Available
AI Foundry, contact-centre, copilot-studio, Customer Experience, Customer-service, Power Apps, Power Automate

Microsoft Ignite 2025: Transitioning to the Era of Agentic Business Transformation

Goloknath

Microsoft Ignite 2025: Transitioning to the Era of Agentic Business Transformation

The central theme for November 2025 was the evolution of Copilot Studio into a foundation for “agentic” workflows—where AI agents don’t just assist but autonomously drive business processes with human oversight.

Real-time User Journey: Human-in-the-Loop (HITL)

One of the most significant “real-time” journeys introduced is the Human-in-the-Loop capability:

  1. Autonomous Start: An agent begins a multi-step workflow (e.g., processing a high-value procurement order).
  2. Pause for Judgment: The agent reaches a step requiring approval or specific context (e.g., the order exceeds a price threshold).
  3. Real-time Request: The agent automatically pauses and sends a structured Outlook form to a designated human reviewer.
  4. Human Action: The person reviews the details and submits their input/approval via the form.
  5. Resumption: The agent receives the response and immediately resumes the workflow using the human’s input as a parameter to complete the task.

Step-by-Step: How to Enable GPT-5 Chat

GPT-5 Chat became generally available in November 2025 for US and EU customers. To enable it:

  • Step 1: Open your agent’s Overview page in Microsoft Copilot Studio.
  • Step 2: Locate the Model Selection settings.
  • Step 3: Select GPT-5 Chat from the dropdown menu to set it as the primary model.
  • Step 4 (Experimental): For US customers wanting to test the latest reasoning, you can toggle on GPT-5.2 (Experimental) in early release environments for improved code generation and multilingual support.

Infographic: The New Agentic Capabilities

The November 2025 update expanded the platform across three core pillars:

Intelligence (Models)Action (Skills)Governance (Control)
GPT-5 GA: Higher accuracy and instruction-following.Human-in-the-Loop: Agents pause for human approval.Microsoft Agent 365: The unified control plane for agents.
Model Choice: Multi-model support (OpenAI, Anthropic, xAI).Action Groups: One-click sets for Outlook/SharePoint tools.Entra Agent ID: Identity-based security for every agent.
MCP Integration: Connects to 1,400+ external systems.Producer Skills: Generate Word, Excel, and PPT files.Defender for Agents: Real-time threat protection.

References

What’s new in Microsoft Copilot Studio: November 2025
AI Foundry, copilot-studio

Announcing Claude Opus 4.5 in Microsoft Copilot Studio

Goloknath

Based on the Microsoft Copilot blog post from November 24, 2025, here is the information regarding the announcement of Claude Opus 4.5 in Copilot Studio:

Announcing Claude Opus 4.5 in Microsoft Copilot Studio

This update marks the integration of Anthropic’s most powerful model into the Copilot Studio platform, providing developers with advanced reasoning and long-context capabilities for building complex agents.

Real-time User Journey

The user journey for Claude Opus 4.5 centers on high-complexity, multi-step problem solving:

  1. Complexity Identification: A developer identifies a business process that requires deep reasoning over vast amounts of data (e.g., analyzing a 200-page legal contract alongside historical case files).
  2. Model Selection: Within Copilot Studio, the developer selects Claude Opus 4.5 as the reasoning engine for their agent.
  3. Interaction: The end-user provides a high-level goal, such as “Identify all conflicting clauses in these five documents and suggest a reconciliation strategy based on our internal compliance handbook.”
  4. Long-Context Reasoning: The agent utilizes the model’s massive context window to ingest all documents simultaneously, maintaining the “thread” across long, multi-step logic paths.
  5. Execution: The agent produces a highly nuanced output, citing specific sections of the context provided, and can trigger follow-up actions like drafting an email to the legal team.

Step-by-Step: How to Enable

As an experimental model, Claude Opus 4.5 requires administrative permission to use:

  • Step 1: Admin Opt-in: A Microsoft 365 Administrator must log in to the Microsoft 365 Admin Center.
  • Step 2: External Model Settings: Navigate to the settings for external AI models and explicitly opt-in to Anthropic models (if not already done).
  • Step 3: Access Copilot Studio: Once opted in, go to Microsoft Copilot Studio.
  • Step 4: Create/Open an Agent: Open an existing agent or create a new one.
  • Step 5: Select Model: Go to Settings > Generative AI. Under the model selection dropdown, choose Claude Opus 4.5 (Experimental).
  • Step 6: Test: Use the “Test your agent” pane to validate how the new model performs with your specific prompts and knowledge base.

Infographic: Claude Opus 4.5 Core Strengths

The following table highlights why Claude Opus 4.5 is positioned as the “premium” choice for complex agents:

FeatureCapabilityBenefit for Agents
Sharper ReasoningAdvanced logical deduction.Solves “impossible” problems that stump smaller models.
Long-ContextHandles massive datasets.Reads hundreds of pages in a single turn without losing focus.
Multi-Step TasksStronger instruction following.Executes complex sequences of actions with high accuracy.
Model ChoiceMulti-vendor flexibility.Allows developers to use Agent Evaluations to pick the best model for the job.

References

Announcing Claude Opus 4.5 in Microsoft Copilot Studio
AI Foundry, architecture, copilot-studio

What’s New in Copilot Studio: September 2025 – Expanding Reach, Automation, and Analytics

Goloknath

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:

  1. Scenario Discovery: A user needs to enter data into an old legacy desktop application that doesn’t have an API.
  2. Instruction: The user tells the agent, “Open the legacy ERP, look up invoice #123, and copy the total into this Excel sheet.”
  3. Vision & Reasoning: The agent uses its “vision” to see the screen and its “reasoning” to identify buttons and text fields.
  4. Execution: On a hosted Windows 365 browser or local device, the agent moves the virtual mouse, clicks, and types to complete the task.
  5. 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:

CapabilityWhat it DoesBusiness 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 SDKEmbeds agents in native Android, iOS, or Windows apps.Keeps users in their flow of work without switching to Teams or Web.

References

What’s new in Copilot Studio: September 2025
AI Foundry, contact-centre, copilot-studio, Customer-service, Power Apps

Microsoft Power Platform: Leading the Agentic Transformation of Low-Code Development

Goloknath

The primary focus is Microsoft’s leadership in the LCAP market for the seventh consecutive year, emphasizing the integration of Agentic AI, Generative Pages, and Professional Developer Tools within a fully governed environment.

Real-time User Journey: “Vibe-Coding” & Agentic Assistance

The journey described in the 2025 landscape moves from building static apps to orchestrating intelligent solutions:

  1. Visionary Starting Point: A business user identifies a complex problem (e.g., global inventory rebalancing).
  2. AI-Guided Design (Plans): The user uses Plans in Power Apps to describe the problem. AI generates a solution blueprint, including data models and logic.
  3. Vibe-Coding (Generative Pages): The user uses natural language to “vibe-code” the interface. The AI creates modern, responsive pages instantly, removing the need for manual UI building.
  4. Agent Integration: The maker adds an Agent into the app. This agent doesn’t just display data; it proactively explores it and assists the end-user with complex data entry.
  5. Collaborative Use: An end-user opens the app and interacts with the Agent Feed, a central hub for human-agent collaboration, to resolve inventory discrepancies.

Step-by-Step: How to Enable Agentic Capabilities

To take advantage of the features highlighted in the Gartner report (like Plans and Agents):

  • Step 1: Access Power Apps Studio: Sign in to make.powerapps.com.
  • Step 2: Start with “Plans”: Instead of starting from a blank screen, select the “Describe to build” or “Start with a Plan” option on the home screen.
  • Step 3: Enable Copilot: Ensure the Copilot component is enabled in your environment settings (managed via the Power Platform Admin Center).
  • Step 4: Create Generative Pages: Within the app designer, use the “New Page” menu and select “Generative Page” to build UI using natural language prompts.
  • Step 5: Add an Agent: Go to the Knowledge or Agents tab within the app studio to configure a built-in assistant that will help end-users navigate the app.
  • Step 6: Publish & Govern: Publish the app to your organization. Admins should use the Managed Platform settings in the Admin Center to monitor usage and secure data via Data Loss Prevention (DLP) policies.

Infographic: The 2025 Power Platform Pillars

The Gartner report recognition is built on these three strategic differentiators:

PillarFeature / CapabilityBusiness Outcome
Agentic AIPlans in Power Apps & Agent Feed30% faster data entry and higher success for new builders.
Enterprise ScaleSingle Admin Center & Azure FoundationConfident global scaling with Managed Identities and VNet support.
Pro-Dev IntegrationNative Visual Studio Code & Custom ConnectorsSeamless “Fusion Team” collaboration between IT and business users.

References

Microsoft recognized as a Leader in the 2025 Gartner® Magic Quadrant™ for Enterprise Low-Code Application Platforms