AI Foundry – Page 2 – GM Dynamics & Agentic AI Consulting

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:

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).
Model Selection: Within Copilot Studio, the developer selects Claude Opus 4.5 as the reasoning engine for their agent.
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.”
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.
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:

Feature	Capability	Benefit for Agents
Sharper Reasoning	Advanced logical deduction.	Solves “impossible” problems that stump smaller models.
Long-Context	Handles massive datasets.	Reads hundreds of pages in a single turn without losing focus.
Multi-Step Tasks	Stronger instruction following.	Executes complex sequences of actions with high accuracy.
Model Choice	Multi-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, copilot-studio, Customer-service, Power Apps

Revolutionizing customer support with autonomous case resolution in Dynamics 365

November 22, 2025May 3, 2026 Goloknath

Revolutionizing customer support with autonomous case resolution

(The introduction of the Case Management Agent, an autonomous AI agent that manages the end-to-end lifecycle of a support ticket without human intervention.)

Real-time User Journey

This journey showcases how an autonomous agent handles a complex, multi-step service request:

Trigger: A customer sends an email regarding a faulty device. The system automatically creates a case in Dynamics 365.
Agent Assignment: Instead of a human queue, the Case Management Agent is assigned. It immediately reads the email, identifies the product, and checks the customer’s warranty status in the database.
Autonomous Investigation: The agent identifies that a diagnostic test is needed. It sends an automated (but personalized) reply to the customer with instructions.
Data Processing: The customer replies with the test results. The agent parses the data, compares it against the “Troubleshooting Knowledge Base,” and determines a replacement is required.
Action Execution: The agent interacts with the ERP system to initiate a replacement order and generates a return shipping label for the customer.
Resolution & Summary: The agent sends the final confirmation to the customer and closes the case. It then writes a concise summary of all actions taken in the timeline for human audit purposes.

Step-by-Step: How to Enable This Feature

The Case Management Agent is part of the new “Agentic” capabilities in Dynamics 365.

Step 1: Access the Agent Hub

Log in to the Customer Service admin center or Contact Center admin center. Navigate to Operations > Agent Hub.

Step 2: Activate the Case Management Agent

Find the Case Management Agent card. Toggle the status to Enabled.

Step 3: Define the “Knowledge Source”

Under the agent settings, point the agent to your high-quality data sources (SharePoint, Dataverse tables, or specific websites) so it knows how to resolve issues based on your company policy.

Step 4: Set Safety Guardrails

Configure the “Human-in-the-loop” settings. You can specify that the agent can resolve “Low” and “Medium” priority cases autonomously but must route “High” priority cases to a human.

Step 5: Configure Connection Actions

Use Power Automate connectors to allow the agent to perform actions in external systems (like SAP, Salesforce, or internal shipping tools) to complete the resolution.

Step 6: Monitor via Activity Feed

Ensure the AI Agent Activity Feed is added to your supervisor’s workspace so they can monitor the agent’s reasoning and step-by-step progress in real-time.

Infographic: Manual vs. Autonomous Case Resolution

Feature	Manual Case Management	Autonomous Case Resolution
Response Time	Hours to Days (Queue dependent).	Seconds/Minutes (Instant).
Investigation	Manual searching of docs/ERP.	Instantaneous cross-system data lookup.
Consistency	Varies by agent experience.	100% adherence to company policy.
Availability	Restricted to business hours.	24/7/365 operations.
Cost	High (Labor intensive).	Scalable (Minimal per-case cost).

References

Revolutionizing customer support with autonomous case resolution

AI Foundry, contact-centre, copilot-studio, Customer-service, Power Apps

Use custom multisession apps in Copilot service workspace in Dynamics 365

November 9, 2025May 3, 2026 Goloknath

Use custom multisession apps in Copilot service workspace

(The ability to embed and launch custom third-party or internal line-of-business applications directly within the tabbed, multisession environment of Dynamics 365).

Real-time User Journey

This journey illustrates how a “Tier 2” support agent uses custom apps to resolve a complex logistics issue without leaving their primary workspace:

Contextual Intake: An agent accepts a high-priority chat regarding a delayed shipment.
Copilot Insight: Copilot summarizes the case and notes that the package is held in a custom third-party warehouse system not native to Dynamics 365.
Seamless Launch: Instead of minimizing Dynamics and opening a browser, the agent clicks a custom tab labeled “Warehouse Manager” within their active session.
In-App Action: The Custom Multisession App opens in a new tab within the same window. The agent updates the shipping manifest and releases the hold.
Data Synchronization: Because the app is integrated into the workspace, the agent can easily copy-paste or link the confirmation ID back to the Dynamics 365 record.
Resolution: The agent closes the warehouse tab and the chat session simultaneously, keeping their workspace clean and focused for the next incoming request.

Step-by-Step: How to Enable This Feature

Administrators can configure these apps through the App Management settings:

Step 1: Admin Center Access

Step 2: Navigate to App Profiles

Go to Agent Experience > Workspaces > Agent experience profiles.

Step 3: Edit the Profile

Select the profile used by your agents (e.g., “Customer Service Representative”) and click Edit.

Step 4: Add Application Tab Template

Navigate to the Application tab templates section. Click Add and select New Template.

Step 5: Define the Custom App
- Set the Template Type to “Third-party website” or “Custom App.”
- Enter the Target URL of your internal tool or third-party software.
- Configure Parameters to ensure the app opens with the correct context (e.g., passing the Customer ID in the URL).
Step 6: Assign to Sidebar/Tab

Choose where the app should appear (e.g., as a default tab in every new session or as an option in the productivity pane). Save and Publish the profile.

Infographic: Single-Session vs. Custom Multisession Workspace

Feature	Single-Session/Standard Browser	Custom Multisession Workspace
App Switching	Manual (Alt+Tab between windows).	Native Tabs within the workspace.
Context Retention	Easy to lose track of which window belongs to which customer.	Session-Linked; apps stay tied to the specific customer record.
Workflow Speed	Slower due to window management.	Instant Access to internal/external tools.
Security	Harder to govern external browser usage.	Governed Access through the Admin Center.
Focus	High risk of distraction/clutter.	Unified Interface for all service tools.

References

Use custom multisession apps in Copilot Service workspace

AI Foundry, contact-centre, copilot-studio, Customer-service

Elevate support excellence with Quality Evaluation Agent in Dynamics 365

November 1, 2025May 3, 2026 Goloknath

Based on the Microsoft Dynamics 365 blog post from October 30, 2025, here is the breakdown of the Quality Evaluation Agent.

Elevate support excellence with Quality Evaluation Agent

(An autonomous AI agent designed to automate the Quality Assurance (QA) process by evaluating 100% of customer interactions against defined business rubrics).

Real-time User Journey: From Manual Sampling to Total Oversight

This journey highlights how a Quality Manager shifts from “hunting for errors” to “strategic coaching”:

Continuous Analysis: Throughout the day, thousands of calls and chats occur. In the background, the Quality Evaluation Agent automatically analyzes every single transcript.
Rubric Application: The agent applies the company’s specific “Excellence Rubric,” checking for compliance, empathy, accurate product knowledge, and adherence to greeting protocols.
Anomaly Detection: The agent identifies a specific agent who is consistently providing high-quality service but struggling with a new “return policy” detail. It also flags a wider trend where customer sentiment drops whenever “shipping delays” are mentioned.
Proactive Reporting: Instead of the manager listening to random tapes, they open their Insights Dashboard. They see a prioritized list of interactions that require human attention (e.g., high-risk calls or those with low AI-generated quality scores).
Targeted Coaching: The manager clicks on a flagged call. The agent provides a Reasoning Summary: “Score: 65/100. Agent missed the mandatory privacy disclosure at 02:14.”
Loop Closure: The manager uses these precise insights to provide a 5-minute targeted coaching session. The agent continues to monitor the agent’s next calls to track improvement in real-time.

Step-by-Step: How to Enable This Feature

The Quality Evaluation Agent is part of the “Agentic” suite in Dynamics 365 Customer Service.

Step 1: Access the Agent Hub

Step 2: Activate the Quality Evaluation Agent

Locate the Quality Evaluation Agent card and select Manage. Toggle the status to Enabled.

Step 3: Define Your Evaluation Rubric

Upload or define your scoring criteria. You can set specific parameters such as:

Compliance: Did the agent verify the customer’s identity?
Soft Skills: Was the tone professional and empathetic?
Process: Did the agent follow the standard troubleshooting steps?
Step 4: Configure Sampling (Optional)

While the agent can evaluate 100% of calls, you can configure it to prioritize specific high-value queues or new trainees for more frequent reporting.

Step 5: Connect to Power BI Insights

Enable the Quality Insights Dashboard. This will pull the agent’s evaluations into visual reports showing trends across teams, time periods, and specific product categories.

Step 6: Set Up Supervisor Alerts

Configure “Critical Failure” alerts. For example, if a “Compliance Score” drops below 50 on a single call, the supervisor receives an immediate notification to review the interaction.

Infographic: Traditional QA vs. Quality Evaluation Agent

Feature	Traditional Manual QA	Quality Evaluation Agent
Coverage	1–2% of interactions (Randomized).	100% of interactions (Total).
Speed	Retrospective (Days or weeks later).	Near Real-time (Minutes after call).
Objectivity	Subject to human bias and fatigue.	Consistent & Unbiased application of rules.
Insights	Anecdotal/Small sample size.	Data-driven trends across the whole org.
Supervisor Role	Listening to tapes (Manual labor).	Coaching & Strategy (High value).

References

Elevate support excellence with Quality Evaluation Agent

AI Foundry, contact-centre, copilot-studio, Customer-service

Try rich messaging for live chat and WhatsApp in Contact Center in Dynamics 365

October 26, 2025May 3, 2026 Goloknath

Try rich messaging for live chat and WhatsApp in Contact Center (Enhancing digital customer service with interactive elements like forms, carousels, and suggested replies).

Real-time User Journey: Interactive Troubleshooting

This journey demonstrates how rich messaging replaces tedious typing with intuitive visual interactions:

Customer Initiation: A customer messages a retail brand via WhatsApp to report a missing part in a furniture delivery.
Suggested Replies: The AI agent instantly provides Suggested Replies (e.g., “Missing Part,” “Damaged Item,” “Wrong Item”). The customer taps “Missing Part” instead of typing it out.
Interactive Form: The agent sends a Rich Form directly in the chat. The customer uses a dropdown to select the furniture model and a date picker to specify the original delivery date.
Visual Confirmation (Carousel): The system displays a Carousel of parts for that specific model. The customer swipes through high-quality images and taps the exact screw or bracket that is missing.
Agent Handoff: If the customer needs a human, the conversation is routed to a live agent. The agent sees the completed form and the selected part in the Agent Workspace, allowing them to finalize the replacement order without asking repetitive questions.
Confirmation: The agent sends an Adaptive Card showing the new delivery tracking number, which the customer can track with one tap.

Step-by-Step: How to Enable This Feature

Administrators can configure rich messaging templates that work across multiple digital channels:

Step 1: Admin Center Access Sign in to the Customer Service admin center or Contact Center admin center.
Step 2: Navigate to Rich Messages Go to Workforce Management (or Customer Support) > Rich messages.
Step 3: Create a Message Template Select New and choose the message type:
- Forms: For structured data collection (text, numbers, dates).
- Suggested Replies: For quick decision buttons.
- Custom JSON: For advanced Adaptive Cards.
Step 4: Design for Channels Use the designer to build your message. The system is channel-aware, meaning you can design one template that automatically adapts its appearance for the Live Chat widget versus WhatsApp.
Step 5: Preview and Publish Use the Preview Pane to see how the message will render on different devices. Once satisfied, select Publish to make the template available to agents and bots.
Step 6: Integration with Copilot Studio In Microsoft Copilot Studio, use the Adaptive Card node to call these published templates, allowing your AI agents to send rich interactive content autonomously.

Infographic: Plain Text vs. Rich Messaging

Feature	Traditional Text Messaging	New Rich Messaging
User Input	Manual typing (high friction).	One-tap buttons & dropdowns.
Data Collection	Open-ended text (requires parsing).	Structured Forms (validated data).
Visuals	Basic attachments/links.	Interactive Carousels & Cards.
Accuracy	Prone to typos and ambiguity.	Precision through predefined choices.
Resolution Time	Slower (back-and-forth questions).	Faster (all data gathered in one card).

References

Try rich messaging for live chat and WhatsApp in Contact Center