AI Foundry, copilot-studio

What’s New in Copilot Studio: July 2025 – Advancing Agentic Capabilities

The primary focus of this update was the introduction of Agentic Actions and Dynamic Chaining, shifting the platform from simple conversation-based triggers to reasoning-based autonomous execution.

Real-time User Journey: Dynamic Chaining

The core “journey” updated in July 2025 involves how an agent handles a multi-step request without manual “if/then” branching:

  1. Complex Prompt: A user asks, “Find the latest sales report for the Northeast region, summarize the top three concerns, and email that summary to the Regional Manager.”
  2. Reasoning: Instead of looking for a specific pre-built “topic,” the agent uses its Generative AI orchestrator to identify the necessary steps.
  3. Dynamic Action: The agent autonomously “chains” together three separate tools:
    • Tool 1 (SharePoint): Searches for and retrieves the specific PDF/Excel file.
    • Tool 2 (LLM Reasoning): Analyzes the content of that file to extract the top three concerns.
    • Tool 3 (Outlook): Drafts and sends the email to the specific contact.
  4. Completion: The agent confirms the email has been sent and provides a copy of the summary to the user in the chat.

Step-by-Step: How to Enable Dynamic Chaining

To move your agent from “Classic” mode to the new “Agentic” reasoning mode:

  • Step 1: Open Agent Settings: In Microsoft Copilot Studio, select your agent and go to the Generative AI page in the side navigation.
  • Step 2: Switch Orchestration: Under the “Orchestration” section, select Dynamic Chaining (Preview). This allows the agent to use generative AI to choose the best topic or action to fulfill a request.
  • Step 3: Add Actions: Navigate to the Actions tab. Add the specific “skills” the agent can use (e.g., pre-built connectors for Outlook, SharePoint, or custom Power Automate flows).
  • Step 4: Refine Descriptions: For every action, provide a clear, natural language Description. This is critical, as the agent uses these descriptions to “know” when to trigger that specific tool.
  • Step 5: Test & Publish: Use the Test Pane to verify the agent can correctly chain these tools together before publishing to your users.

Infographic: The Shift to Agentic AI

The July 2025 update defined a clear evolution of the platform’s architecture:

FeatureTraditional Bot (Legacy)Agentic AI (New)
Logic EngineFixed Topic Triggering (Keywords)Generative AI Orchestration
WorkflowLinear/Pre-defined PathsDynamic Chaining (Reasoning-based)
Tool IntegrationManual API CallsAutonomous Tool Use via Descriptions
User InputsSpecific Phrases RequiredComplex, Multi-intent Prompts

References

Uncategorized

2025 Release Wave 2: Transitioning to Proactive, Agent-Based AI Ecosystems

2025 Release Wave 2: Transitioning to Proactive, Agent-Based AI Ecosystems The core theme of this release (covering October 2025 to March 2026) is the shift from “Assistive AI” (waiting for user prompts) to “Agentic AI”—autonomous agents that act as proactive partners to drive business outcomes across Sales, Service, Finance, and Supply Chain.

Real-time User Journey: The Autonomous Supplier Communications Agent

A flagship journey introduced in this wave involves the Supplier Communications Agent within Dynamics 365 Supply Chain Management:

  1. Detection: The agent monitors inventory levels and identifies a potential stockout for a critical component due to a sudden demand spike.
  2. Autonomous Outreach: Without human intervention, the agent drafts and sends an inquiry to preferred vendors to check for lead times and pricing.
  3. Negotiation: The agent receives vendor replies, compares them against historical data and budget constraints, and identifies the best option.
  4. Drafting for Approval: The agent presents a “prepared action” to the Procurement Manager: “I’ve found three vendors who can fulfill this; Vendor X is the cheapest and fastest. Click ‘Approve’ to issue the PO.”
  5. Closing the Loop: Once approved, the agent handles the order entry and updates the demand forecast automatically.

Step-by-Step: How to Enable Wave 2 Features

Microsoft has modernized the rollout process by replacing the legacy “Early Access” toggle with a streamlined “Release Channel” system:

  • Step 1: Access Admin Center: Log into the Power Platform Admin Center (admin.powerapps.com).
  • Step 2: Create a Sandbox: It is highly recommended to create a copy of your production environment for testing before enabling new features.
  • Step 3: Switch Release Channel: Navigate to Environments > [Your Environment] > Settings > Product > Behavior.
  • Step 4: Opt-in to Monthly Channel: Change the “Release channel” setting to “Monthly channel”. This immediately enables the Wave 2 preview features for your environment.
  • Step 5: Configure Specific Agents: For features like the Sales Research Agent or Finance Reconciliation Agent, go into the respective app settings (e.g., Copilot Studio) to ground the agent in your specific Dataverse tables.

Infographic: 2025 Wave 2 Strategic Pillars

This release wave focuses on four key pillars of the “Agentic Era”:

PillarFeature HighlightImpact
Agentic Power AppsAgent FeedA hub for humans to supervise and interact with multiple agents working in the background.
Autonomous ERPPayables & Sales AgentsMoves Business Central from manual entry to automated order-taking and payment reconciliation.
Intelligent PortalsPower Pages Security AgentProactively scans external-facing sites for vulnerabilities and suggests fixes to makers.
Unified GovernanceAutomation CenterA single dashboard to monitor the health, security, and ROI of all agents and flows across the tenant.

Reference

Uncategorized

What’s New in Copilot Studio: June 2025 – Enhancing Agent Lifecycle and Enterprise Governance

What’s New in Copilot Studio: June 2025 – Enhancing Agent Lifecycle and Enterprise Governance

The June update focused on maturing the “Pro-dev” experience, specifically introducing Agent Environments, Managed Identities, and Advanced Analytics to help organizations move agents from prototypes to secure, governed production workloads.

Real-time User Journey: Managed Identity Authentication

One of the most significant “real-time” journeys introduced is how agents access secure company data without managing complex user secrets:

  1. Request: An employee asks the agent, “What is the status of the confidential project Alpha in our Azure SQL database?”
  2. Silent Authentication: Instead of prompting the user for a login or using a shared service account password, the agent uses its Managed Identity (a secure, Entra-based ID assigned to the agent itself).
  3. Secure Access: The agent authenticates directly with the Azure SQL database. The database recognizes the agent’s identity and grants access based on the specific permissions set by the IT admin.
  4. Reasoning & Response: The agent retrieves the data, summarizes it using the LLM, and provides the answer—all while ensuring no passwords were ever exposed or stored in the code.

Step-by-Step: How to Enable Agent Environments

This feature allows teams to separate development work from production use:

  • Step 1: Admin Center Setup: Log into the Power Platform Admin Center and navigate to the Environments tab.
  • Step 2: Create “Dev” and “Prod” Environments: Create two separate environments specifically for your Copilot agents to keep experimental features away from live users.
  • Step 3: Move Agents via Solutions: In Copilot Studio, wrap your agent in a Solution (a container for all its topics and tools).
  • Step 4: Export/Import: Export the solution from the Dev environment and import it into the Prod environment.
  • Step 5: Connection Reference Update: Once imported, update the Connection References (e.g., SharePoint or Outlook links) to point to the production-level data sources.

Infographic: Enterprise Agent Readiness

The June update introduced a “checklist” of capabilities that define an enterprise-ready agent:

PillarFeatureValue
SecurityManaged IdentitiesEliminates the need for hard-coded secrets or credentials.
LifecycleSolution IntegrationEnables standard DevOps pipelines (ALM) for agent deployment.
InsightsConversation TranscriptsGA of full transcript logs for auditing and debugging.
ConnectivityEnhanced ConnectorsNew UI for managing 1,400+ connectors with better error handling.
DiscoveryAgent GalleryA central place for employees to find approved agents.

References

Uncategorized

What’s New in Copilot Studio: May 2025 – The Era of Autonomous Agents

What’s New in Copilot Studio: May 2025 – The Era of Autonomous Agents

The theme of this update was the transition from “Copilots” (assistants that wait for prompts) to “Agents” (autonomous entities that can be triggered by business events to work in the background).

Real-time User Journey: Event-Driven Autonomy

The core feature introduced in May 2025 was the ability for agents to act without a user starting a chat.

User Journey Example: Automated Onboarding

  1. Trigger: A new record is created in a HR system (e.g., Workday) indicating a new hire has signed their offer letter.
  2. Autonomous Analysis: The agent is “woken up” by this event. It automatically reviews the new hire’s role and department.
  3. Cross-App Orchestration: Without human intervention, the agent:
    • Requests a laptop via the IT ticketing system (ServiceNow).
    • Sends a welcome email with a customized first-week agenda to the new hire via Outlook.
    • Pins a “New Joiner” announcement in the specific department’s Microsoft Teams channel.
  4. Human Summary: Once finished, the agent sends a notification to the HR Manager saying, “I’ve completed the onboarding setup for [Employee Name]. Here is the summary of actions taken.”

Step-by-Step: How to Enable “Agentic” Triggers

This feature allows you to move beyond the “Ask a question” model.

  • Step 1: Open Agent Studio: Navigate to your agent in Microsoft Copilot Studio.
  • Step 2: Add a Trigger: Go to the Triggers tab (previously just “Topics”). Select “Add a Trigger based on an external event.”
  • Step 3: Connect Data Source: Select a connector (e.g., Dataverse, SharePoint, or an External API). Choose the event type, such as “When a row is added” or “When a file is uploaded.”
  • Step 4: Define Instructions: Instead of a dialogue tree, provide the agent with Instructions on what to do when this event occurs. Use natural language to describe the logic.
  • Step 5: Configure Actions: Add the “Tools” the agent is allowed to use to fulfill the instructions (e.g., send an email, update a CRM record).
  • Step 6: Turn on Autonomy: Toggle the “Run Autonomously” switch and publish the agent.

Infographic: The New Agent Architecture

The May 2025 update fundamentally changed the “brain” of Copilot Studio agents:

ComponentOld Way (Bot)New Way (Agent)
TriggerUser must type a message.External events (Data/Time/System changes).
LogicManual “If/Else” branches.Generative AI Orchestrator (Reasoning).
MemorySession-based (Forgot quickly).Persistent Memory (Remembers across runs).
ToolsFixed API calls.Dynamic Selection of tools based on the goal.

References

Uncategorized

What’s New in Copilot Studio: April 2025 – Advancing Multi-Agent Orchestration and Connected Experiences

The April 2025 update focused on the release of the Multi-Agent Orchestration engine and the introduction of Agent-to-Agent (A2A) connectivity, allowing agents to work together as a unified system.

Real-time User Journey: Collaborative Multi-Agent Resolution

The core journey introduced this month involves “handoffs” between specialized agents to solve a complex user request:

  1. Request: A user asks a general “HR Agent” in Teams, “I’m traveling to London for a client meeting next week; can you handle my travel booking and ensure it complies with our latest policy?”
  2. Orchestration: The HR Agent realizes it doesn’t have booking permissions. Using the new Orchestration Engine, it identifies a specialized “Travel Booking Agent” and a “Policy Compliance Agent.”
  3. Collaborative Execution: * The Policy Agent retrieves the travel policy from SharePoint and provides the budget limits to the Travel Agent.
    • The Travel Agent searches for flights and hotels that fit those limits.
  4. Verification: The agents pass data back and forth to confirm the hotel choice is within the allowed budget.
  5. Completion: The primary HR Agent returns to the user with a single message: “I’ve found a flight and hotel that comply with our policy. Would you like me to book them?”

Step-by-Step: How to Enable Multi-Agent Connectivity

To allow your agents to communicate with one another using the new protocol:

  • Step 1: Open Copilot Studio: Navigate to your primary agent.
  • Step 2: Enable Orchestration: Go to Settings > Generative AI and toggle the orchestration mode to “Generative (Preview)” or “Multi-agent.”
  • Step 3: Register Sub-Agents: Go to the Actions or Tools tab and select “Add an Agent.” You can now select other agents published within your environment.
  • Step 4: Define Descriptions: Ensure each sub-agent has a highly detailed Instructional Description. The orchestrator uses these to “know” which agent is the expert in travel, finance, or technical support.
  • Step 5: Configure Permissions: Set up Delegated Authentication so the sub-agents can act on behalf of the user who started the original chat.
  • Step 6: Test & Publish: Use the “Test” pane to watch the “Conversation Map” and see how the orchestrator routes the logic between agents.

Infographic: The April 2025 Feature Set

The April update brought several tools to manage complex AI ecosystems:

FeatureWhat it DoesImpact on Business
Multi-Agent OrchestrationAllows a “Master Agent” to manage “Sub-Agents.”Reduces the need for users to know which specific bot to talk to.
A2A Protocol SupportOpen standard for agent-to-agent communication.Enables agents to talk across different platforms (e.g., Copilot to custom Azure bots).
Conversation MapsVisual debugging tool for orchestration.Helps developers see exactly where a handoff happened or failed.
Knowledge HandoffContext is passed between agents during a handoff.The user doesn’t have to repeat their information to the second agent.

References