Course Description
This course is an immersive, hands-on training course designed to equip IT professionals, solution architects and business technologists with the skills to design, build and manage intelligent AI agents using Microsoft Copilot Studio. Participants will gain practical experience creating both conversational and autonomous agents that integrate seamlessly with Microsoft 365, Power Platform and enterprise data sources. Through guided labs, real-world examples and advanced customisation exercises, learners will develop the confidence to deploy secure, scalable Copilot agents that enhance productivity, automate workflows and transform organizational processes. This course blends conceptual understanding with hands-on application, progressing from foundational Copilot principles to advanced orchestration, governance and autonomous agent design.
Audience Profile
This course is designed for professionals who want to build or manage intelligent agents using Microsoft Copilot Studio, including:
Power Platform Makers and Developers who wish to extend their automation capabilities with AI-driven agents.
IT Administrators and Solution Architects responsible for deploying and governing AI solutions within Microsoft 365 environments.
Business Analysts and Citizen Developers looking to create task-specific copilots without extensive coding.
AI and Automation Specialists who want to design secure, autonomous, and scalable Copilot agents for enterprise use cases.
Copilot Chat and Agents
Introduce participants to Microsoft Copilot Chat and the concept of agents within the Microsoft 365 ecosystem. Learners will explore how Copilot serves as the AI interface for users and how agents extend its capabilities to perform tasks, retrieve data and automate workflows.
1. Understanding Microsoft Copilot:
Overview of Copilot in Microsoft 365 (Teams, Word, Outlook, and more)
How Copilot Chat enhances productivity through conversational AI
“Copilot lite” vs “Full Copilot” capabilities
How Copilot integrates with organizational data and permissions
2. Introducing Agents:
Definition and purpose of Copilot Agents
Types of agents: Simple, Task-based and Autonomous
Example scenarios: IT Helpdesk Bot, HR Policy Assistant, Project Tracker
How agents interact with Copilot Chat
3. Deployment and Security:
Where agents run: Teams, web and Power Platform environments
Data governance, privacy and compliance considerations
Licensing, cost control and telemetry tracking
Lab 1: Build agents in Copilot Studio Lite Experience
Building Agents with Copilot Studio
This module provides a deep dive into the architecture, core capabilities and authoring experience of Microsoft Copilot Studio. Participants will learn how Copilot Studio operates within the Microsoft ecosystem, understand its performance and governance considerations and develop hands-on skills in building conversational agents using the authoring canvas and its core components followed by an introduction to its advanced tools, orchestration features and generative AI integration.
Through guided instruction and a practical lab, learners will design, author and deploy a fully functional conversational agent using variables, entities and structured conversation flows.
1. Copilot Architecture and Core Capabilities:
Overview of Copilot Architecture and its role in the Microsoft 365 and Power Platform ecosystem
Copilot Studio Core Capabilities – building, deploying, and managing conversational and autonomous agents
Copilots and Conversational AI – the evolution from chat-based interaction to orchestrated reasoning
Understanding conversation volumes, quotas, and limits for performance management
Performance optimization and telemetry insights for efficient operations
2. Planning and Lifecycle Management:
Planning your agent – defining purpose, scope, target users, and measurable outcomes
The Typical Copilot Studio Lifecycle:
Initiate → Prepare → Design → Build → Deploy → Operate → Optimize
Selecting the right tools for your governance requirements
Citizen, Partnered, and Pro Developer governance zones
Role-based permissions and security boundaries
Leveraging the Power Platform Center of Excellence (CoE) for governance, auditing and analytics
3. Copilot Studio Authoring Canvas
Navigating the authoring canvas and understanding topic-driven design
Using the Message Node to deliver static, dynamic, or formatted responses
Using the Question Node to capture user input and enable conditional branching
Creating rich text responses (hyperlinks, adaptive cards, lists, media)
Using variables to navigate customers to tailored content
Defining entities and slot-filling for structured data collection
Topic management – organizing and linking topics for efficient conversation flow
Using enhanced speech authoring capabilities for voice interactions
Productivity and pro-code options – integrating Power Fx, expressions, and custom logic
Lab 2: Build a conversational agent in Copilot Studio
4. Agent Tools, Knowledge, and Orchestration:
Overview of Agent Tools within Copilot Studio
Connecting Knowledge Sources for grounding answers and context
Using Tools to extend agent capabilities:
Copilot Connectors – integrate with Microsoft 365, Dynamics 365, and external APIs
Computer Use Tool – automate interactions with desktop or web apps
Code Interpreter – execute Python code directly within Copilot Studio
Advanced Approvals – automate multi-level approval processes
Leveraging the Orchestrator to coordinate multi-topic, multi-tool execution
Designing Prompts effectively for contextual, human-like responses
Generative AI and Optimisation:
Generative AI in Copilot Studio – enabling AI reasoning and dynamic responses
Generative building – AI-assisted topic creation and optimization
Generative answers – grounded, data-informed AI responses
How Retrieval-Augmented Generation (RAG) supports accuracy and reliability
Using analytics to easily optimize with data-driven insights
Tracking agent performance
Measuring engagement and response accuracy
Iterating designs based on telemetry
6. Integration, Agent Flows and Multi-Agent Orchestration:
Designing Agent Flows for end-to-end automation
Multi-agent orchestration – coordinating several specialized agents to handle complex workflows
Planning integrations with Power Automate, Azure services, and external APIs
Deploying and using agents in any system – Teams, web, Dynamics 365, or embedded applications
Lab 3: Use tools in Copilot Studio
Building Autonomous Agents
This module introduces the concept of autonomous agents within Microsoft Copilot Studio and explores how they extend beyond conversational interactions to perform independent, goal-oriented actions.
Learners will understand the difference between conversational and autonomous agents, explore the core components that enable autonomy and learn how to craft precise, ethical and effective agent instructions.
Through guided demonstrations and practical exercises, participants will design the foundational structure of an autonomous agent, integrating triggers, tools and logic to automate business processes intelligently and responsibly.
1. Understanding Autonomous Agents:
What Are Autonomous Agents?
Definition and evolution of autonomous agents within Microsoft Copilot Studio
How autonomous behavior enhances business automation and decision-making
Examples of autonomous agents in real-world enterprise scenarios (approvals, notifications, data insights)
Conversational vs Autonomous Agents
Key differences in purpose, behaviour and user interaction
Conversational agents: reactive and guided by prompts
Autonomous agents: proactive, event-driven, and task-focused
Choosing the right agent type for your business process
2. Reinventing Business Processes with Agents:
How autonomous agents can reimagine workflows across departments
Mapping manual processes to autonomous agent capabilities
Best practices for balancing automation with control and accountability
3. Crafting Autonomous Agent Instructions:
The Art of Instruction Design
Writing clear, structured agent instructions
Defining role, tone, scope, and context for autonomous behaviour
Handling ambiguity, uncertainty, and ethical constraints
Building Blocks of Writing Agent Instructions
Purpose statement
Behavioural rules and escalation paths
Contextual memory and grounding
Error handling and safety checks
Examples of well-structured vs poorly written instructions
4. Building Blocks of an Autonomous Agent:
Core Components of Autonomy
Triggers: initiating events that activate agents (manual, scheduled, event-driven)
Tools: actions and connectors that allow agents to perform work
Knowledge: content sources and grounding data
Instructions: defining how agents reason and respond
Designing multi-step workflows with Copilot Studio tools and triggers
Understanding dependencies, data flow and governance boundaries
Lab 4: Make your agent autonomous in Copilot Studio
Language and Orchestration
This module focuses on how Microsoft Copilot Studio interprets, processes and orchestrates natural language to deliver accurate, contextual and human-like responses.
Learners will explore language understanding models, orchestration types and best practices for building intelligent agents that can manage ambiguity, route topics effectively and deliver consistent communication across languages and scenarios.
1. Natural Language Understanding (NLU) in Copilot Studio:
Overview of Natural Language Understanding and its role in conversational AI
How Copilot interprets intent, entities and context from user input
Comparison of language processing models:
Classic Orchestration (topic-based routing)
Generative Orchestration (AI-driven reasoning and chaining)
The evolution from rule-based intent matching to generative orchestration
2. Orchestration Models and Logic:
Classic Orchestration:
How Copilot matches user intent to predefined topics
Topic routing, priority, and fallback mechanisms
Using disambiguation prompts when multiple intents are detected
Disambiguation and Orchestration:
Strategies for managing ambiguous queries
Designing topic structures for clear and efficient disambiguation
Example: differentiating between “request access” vs. “reset password” topics
Generative Orchestration:
How Copilot uses generative AI to route across tools, topics, and data
The orchestration engine: connecting tools, knowledge sources, and actions
Combining structured (classic) and generative (adaptive) orchestration for hybrid agents
How orchestration improves responsiveness, reduces errors, and automates complex reasoning
How Tools and Topics Are Orchestrated:
Role of the orchestrator in managing flow between topics and tools
Understanding priority, context retention, and handoffs
Integrating multiple data sources or services within a single agent session
3. Designing for Clarity and Effectiveness:
Best Practices for Agent Instructions:
Writing clear, concise, and contextual instructions for accurate orchestration
Controlling tone, scope, and fallback responses
Using grounding and constraints to ensure reliable generative outputs
Best Practices for Topic Inputs & Outputs:
Defining input and output expectations for each topic
Using metadata and variable binding for consistent data transfer
Creating well-structured topic hierarchies to reduce confusion and error loops
4. Language Control and Localisation:
What Controls the Agent Language:
Role of AI models, user preferences, and system settings in language handling
Managing multilingual environments with consistent tone and structure
Auto-Detect Spoken Language:
How Copilot Studio detects and adapts to user language automatically
Best practices for supporting multilingual users (voice and text)
Configuring fallback or default language options for enterprise scenarios
AI capabilities
This module introduces learners to the power of Retrieval-Augmented Generation (RAG) and how it enhances Copilot Studio agents with accurate, grounded and secure generative responses. Participants will explore how RAG architecture operates within Copilot Studio, how to connect and manage knowledge sources effectively and how to safely infuse generative AI capabilities into topic designs while maintaining compliance and reliability.
1. How RAG Enhances AI Responses:
Understanding Retrieval-Augmented Generation (RAG) and why it’s essential for enterprise-grade AI
The limitations of pure generative AI (hallucination, context drift, factual inaccuracy)
How RAG combines retrieved data from trusted knowledge sources with generative reasoning
Benefits of RAG in Copilot Studio:
Grounded responses based on organizational data
Context retention and relevance in extended conversations
Reduced risk of misinformation or unsupported answers
Real-world examples: HR policy retrieval, compliance support and project insights
2. RAG Architecture in Copilot Studio:
Overview of RAG architecture within the Microsoft Copilot ecosystem
The retrieval layer: indexing, search and grounding data pipelines
The generation layer: contextual synthesis using large language models (LLMs)
How Copilot Studio manages context windows, relevance scoring, and ranking
Integrating with Microsoft Graph, SharePoint, Dataverse and external data repositories
Understanding caching, latency, and query optimization in RAG-based designs
3. Knowledge Sources and Generative AI:
Defining knowledge sources in Copilot Studio
SharePoint document libraries
Dataverse tables
External data via connectors or APIs
How Copilot Studio uses knowledge sources to ground generative responses
Techniques for data curation and preparation for RAG indexing
Integrating structured and unstructured data into your Copilot agent
How Generative AI interacts with these sources to produce reliable, factual answers
4. Generative AI Security and Compliance Considerations:
Data protection in generative AI workflows
How Copilot Studio handles sensitive or restricted content
Managing compliance requirements (GDPR, data residency, retention policies)
Understanding Responsible AI principles within Microsoft’s Copilot framework
Limiting generative scope through instructions, topic constraints and data access rules
Setting boundaries for autonomous or open-ended responses
5. Infusing Generative AI into Topics:
Methods for embedding generative capabilities directly within topics
Enabling Generative Building and Generative Answers in Copilot Studio
Designing hybrid topics that use both structured and generative logic
Writing grounded prompts that combine user input, retrieved knowledge and AI synthesis
Testing and refining generative topics to ensure consistency, tone and accuracy
Integrations
This module explores how to integrate Copilot Studio agents with enterprise systems and external data sources using connectors, flows, APIs and automation tools.
Learners will understand key integration patterns, performance constraints and quotas, as well as how to extend Copilot capabilities through custom connectors, HTTP requests and the Model Context Protocol (MCP).
1. Integration Patterns and Considerations:
Overview of integration architecture within Copilot Studio and the Power Platform
Integration patterns:
Direct integration (connectors and HTTP actions)
Event-driven orchestration (Power Automate flows)
Data-driven integration (Dataverse, APIs, Azure Logic Apps)
Choosing the right integration model based on scalability, latency, and control
Security considerations for data flow and authentication (OAuth, managed identity, service principal)
2. Quotas, Limits, and Performance:
Understanding Copilot Studio integration quotas and limits (calls per minute, session size, data throughput)
Performance tuning strategies for efficient agent workflows
Managing API throttling, retries, and error handling in long-running tasks
Using telemetry and analytics to monitor connector and flow performance
How to design integrations that minimize cost and resource consumption
3. Agent Flows, HTTP Actions and Connectors:
Introduction to Agent Flows in Copilot Studio for orchestrating integrations
Using HTTP actions for RESTful API calls and external service integration
Working with standard connectors (SharePoint, Outlook, Teams, Azure, Dynamics 365, Dataverse, etc.)
Designing connector-based actions to extend agent functionality
Handling timeouts and long-running processes in cloud flows
Best practices for async processing and status callbacks
Managing cloud flow timeouts gracefully with user feedback loops
4. Custom Connectors and Advanced Integration:
Overview of Custom Connectors in Copilot Studio and Power Platform
How to build and register a custom connector for internal or external APIs
Using Swagger/OpenAPI definitions to define connector actions and responses
Testing and validating custom connectors in sandbox environments
Governance considerations for connector publishing and sharing
Integration with Azure Functions and Logic Apps for extensibility
5. Model Context Protocol (MCP):
Introduction to the Model Context Protocol (MCP) and its role in AI-driven integrations
How MCP enables Copilot agents to securely access external data models
Using MCP to extend Copilot with contextual understanding from multiple systems
Best practices for maintaining data integrity and compliance in MCP-connected agents
6. Computer-Using Agents (CUA) and RPA:
Understanding Computer-Using Agents (CUA) – what they are and how they work
Enabling agents to interact with desktop applications and web browsers
Comparing CUA and Robotic Process Automation (RPA):
RPA for structured, rule-based workflows
CUA for intelligent, adaptive automation through AI orchestration
Building integrated processes that combine cloud flows, connectors, and CUAs
Example use cases:
Reading data from legacy apps
Filling forms or automating reports
Coordinating actions between on-premises and cloud systems
Security, monitoring and governance
This module equips learners with the knowledge and best practices needed to secure, monitor and govern Microsoft Copilot Studio environments at scale. Participants will learn how to balance innovation and control, implement zoned governance models, enforce data loss prevention (DLP) and compliance policies and manage the security of both agents and users.
1. Balancing Innovation and Governance:
The importance of governance in AI and Copilot deployment
Balancing citizen development and enterprise oversight
How governance enables safe innovation without stifling productivity
Building governance frameworks aligned with corporate IT and security policies
Common governance challenges in scaling Copilot adoption
2. The Agent Controls Model:
Understanding the Agent Controls Model: policies, permissions and oversight
Key governance layers: user, environment, tenant and data
How agent-level controls support compliance and operational transparency
Tracking agent performance and telemetry for security auditing
3. Zoned Security, Governance and Operations:
Overview of Governance Zones (1–3):
Zone 1: Citizen Development (low-risk, innovation sandbox)
Zone 2: Partnered Development (moderate risk, departmental use)
Zone 3: Pro Development (high governance, enterprise-critical)
Mapping organizational maturity to governance zones
Getting Started with Zones – setting up secure, scalable environments
How to manage transitions between zones while maintaining compliance
4. Security and Administration Controls:
Overview of security architecture in Copilot Studio and Power Platform
Security, agent and user management strategies:
Assigning roles and permissions
Enabling secure authentication (Azure AD, Entra ID)
Monitoring user actions and agent activity logs
Designing an effective environment strategy for production, test, and development
Understanding Copilot Studio security roles and least-privilege access design
5. Data Loss Prevention (DLP) and Policy Management:
Overview of DLP Policies in Power Platform and Copilot Studio
The role of DLP connectors and data classification in controlling data flow
How to manage connectors across risk profiles (Business vs. Non-Business)
DLP policies and rules per environment — and when they can be safely relaxed
Designing DLP frameworks that balance flexibility and compliance
Practical examples: blocking external connectors, auditing data access
6. Securing Copilot Studio Usage at Scale
Strategies for secure scaling across large organizations
Controlling adoption through environment boundaries and sharing rules
Using analytics and reports to monitor agent performance, cost and usage
Automating governance checks and policy enforcement through CoE Starter Kit
Prompt Injection Mitigations – preventing manipulation of agent behaviour through malicious inputs
DDoS Protection for Anonymous Chatbots – safeguarding public-facing Copilots against overload attacks
Data Residency and Compliance Management
Understanding data residency and storage in Microsoft Copilot Studio
Managing data movement restrictions across geographies and tenants
Compliance with GDPR, SOX, HIPAA, and other global standards
Designing for multi-region governance and local data processing
Tools and best practices for monitoring data movement and access patterns
Application Lifecycle Management
This module focuses on implementing Application Lifecycle Management (ALM) practices for Copilot Studio. Learners will explore how ALM ensures structured development, testing and deployment of Copilot agents across environments while maintaining governance, consistency, and control. Participants will gain practical insights into using Power Platform ALM, Azure DevOps, GitHub Actions and Power Platform Pipelines to manage agent updates and continuous delivery in enterprise environments.
1. Understanding ALM Strategy:
Defining an ALM strategy for Copilot agents within Microsoft 365 and Power Platform
Why ALM is essential for enterprise-scale development and governance
The benefits of structured lifecycle management:
Version control
Testing and quality assurance
Controlled deployment across environments
Reduced risk and rework
Aligning ALM practices with organizational change management and security policies
2. What Is ALM and Why It’s Important:
Overview of Application Lifecycle Management (ALM) concepts
Development → Testing → Staging → Production cycles
Managing agent versions and configurations
How ALM supports collaboration between makers, developers and administrators
Common pitfalls in unmanaged agent updates or direct publishing
Real-world ALM examples in Copilot Studio agent deployment
3. What “Publish” Really Does in Copilot Studio:
Understanding the Publish process in Copilot Studio
What happens behind the scenes when publishing an agent
Version creation, environment packaging, and synchronization
How publishing differs from exporting/importing solutions in Power Platform
Best practices for publishing safely without disrupting production agents
Integrating publishing into your broader ALM workflow
4. Power Platform ALM for Copilot Studio:
Overview of Power Platform ALM capabilities relevant to Copilot Studio
How solutions encapsulate Copilot agents, connections and data configurations
Managing Copilot components as part of broader Power Platform solutions
Understanding environments and their role in ALM:
Development, Test, UAT and Production
Environment permissions, data boundaries and DLP alignment
Tracking agent versions, dependencies and solution history
5. ALM with Azure DevOps:
Integrating Azure DevOps with Power Platform and Copilot Studio
Managing Copilot solution source control and version tracking
Automating deployment pipelines through Azure DevOps YAML templates
Example workflows:
Export → Validate → Deploy
Trigger-based deployments for agents or connectors
Using Azure DevOps Boards for ALM governance and change tracking
6. GitHub Actions for Microsoft Power Platform:
Introduction to GitHub Actions for CI/CD with Power Platform
Setting up a GitHub repository to manage Copilot Studio solutions
Example automation:
Exporting a solution from Dev → Importing to Test or Prod
Running validation checks before deployment
Comparing Azure DevOps Pipelines vs. GitHub Actions for ALM workflows
Security and permission considerations for GitHub integrations
7. Power Platform Pipelines for Copilot Studio:
Introduction to Power Platform Pipelines — no-code ALM for citizen and pro developers
How Pipelines simplify solution promotion across environments
Configuring automated pipelines for Copilot Studio agents
Using deployment profiles to control environment variables and data connections
Monitoring deployment success, rollback procedures and version tracking
Combining Pipelines, GitHub, and DevOps for hybrid ALM strategies
Analytics and KPIs
This module teaches learners how to measure, analyse and optimise the performance of Copilot Studio agents using data-driven insights.
Participants will explore conversation analytics, engagement metrics and key performance indicators (KPIs) that reflect business impact and user satisfaction.
By implementing a structured analytics and optimization strategy, learners will be able to continuously improve their agents’ effectiveness, refine conversation design and demonstrate ROI to stakeholders.
1. Conversation Design and Outcome Tracking:
Understanding conversation analytics in Copilot Studio
Measuring conversation flow effectiveness: intent recognition, success paths and drop-off points
Designing conversations with measurable outcomes (e.g., task completion, satisfaction, resolution rates)
Tracking end-user interactions and intent success using telemetry and built-in analytics
Mapping conversational outcomes to business objectives and performance goals
Using conversation data to refine prompts, topics and agent logic
2. Engagement and Outcomes:
Defining and measuring engagement metrics:
Total users, active sessions, conversation depth and dwell time
Repeat interactions and user satisfaction trends
Identifying key engagement drivers — tone, context, personalization and response time
Using data to segment audiences and identify high-value user scenarios
Correlating agent engagement with organizational productivity and ROI
Example KPIs:
Resolution rate per topic
Time-to-response improvement
Reduction in support tickets through automation
Business cost savings from AI adoption
Analytics Strategy:
Building a comprehensive analytics strategy for Copilot agents
Aligning analytics goals with business priorities and governance policies
Defining measurable KPIs for agent performance and value realisation
Using Microsoft analytics tools:
Copilot Studio Analytics Dashboard
Power BI integration for advanced reporting
Dataverse telemetry for raw data analysis
How to combine Copilot analytics with Power Platform CoE dashboards
Tracking metrics across environments: Dev, Test and Production
Data governance considerations in analytics — ensuring accuracy and privacy
Optimisation Strategy:
Developing an optimisation cycle for continuous improvement:
Monitor → 2. Analyse → 3. Adjust → 4. Deploy → 5. Measure again
Leveraging A/B testing for topic or prompt improvements
Applying analytics insights to refine:
Agent tone and conversational flow
Topic hierarchy and orchestration logic
Knowledge source selection and generative AI prompts
Using performance data to identify underperforming agents or topics
Creating a feedback loop with stakeholders and users for ongoing tuning
Setting thresholds and alerts for critical KPIs (e.g., low satisfaction, high failure rate
Licensing and capacity
This module provides learners with a deep understanding of how Microsoft Copilot Studio licensing and capacity consumption work within the Power Platform ecosystem.
Participants will learn how to plan, monitor and manage Copilot Studio resource usage across environments while maintaining cost efficiency and operational scalability.
1. Licensing and Capacity Overview:
Overview of Copilot Studio licensing models
Licensing through Microsoft 365, Power Platform, and standalone Copilot subscriptions
Key licensing dependencies: Power Virtual Agents, Power Automate, Dataverse
Capacity components and what they represent:
Dataverse storage (database, file, log)
Power Platform request limits
AI Builder and Copilot Studio usage entitlements
Aligning licensing strategy with your organization’s scale, user base, and governance zones
How to assign and manage licenses across tenants and environments
2. Basic Credit Consumption Scenarios:
Understanding Copilot capacity credits and how they are consumed
Common usage patterns that drive credit consumption:
Agent interactions and conversation sessions
Generative AI responses and knowledge retrieval
Tool usage (code interpreter, connectors, RAG queries)
Mapping agent types to credit requirements (simple, task-based, autonomous)
Real-world credit usage examples for typical Copilot deployments
How conversation complexity, orchestration, and integration affect credit burn
3. Agent Activity and Billing Rates:
Understanding Agent Activity metrics and their impact on billing
How billing rates differ based on:
Generative AI vs. retrieval-based responses
Tool and connector usage
Frequency of orchestration or autonomous agent actions
Reading and interpreting usage reports in the Power Platform admin centre
Using telemetry data to connect activity metrics to cost drivers
Best practices for minimizing unnecessary agent calls and redundant executions
4. Understanding Credit Burn Rate:
Definition of credit burn rate in Copilot Studio
How to monitor and project credit consumption across environments
Factors influencing burn rate:
Agent concurrency and scaling
Number of active users or sessions
Size and complexity of AI prompts and retrieval operations
Strategies for managing and reducing burn rate:
Optimizing conversation length and efficiency
Reusing knowledge sources and cached responses
Scheduling non-critical agents during off-peak hours
Setting up alerts or dashboards to track consumption trends
5. Copilot Studio Estimator:
Introduction to the Copilot Studio Estimator Tool
How to use the estimator to forecast usage, licensing, and costs
Simulating scenarios based on:
Number of agents
Daily conversation volume
Generative AI usage patterns
Interpreting estimator outputs to guide budget and capacity planning
Integrating estimator results into business case and ROI modelling
Capacity Management
Building a capacity management strategy for Copilot Studio
Monitoring capacity in the Power Platform Admin Centre
Allocating capacity per environment and adjusting for usage growth
Using governance zones and environment strategy to balance capacity
Managing cross-tenant capacity sharing and reporting
Planning for scale: forecasting enterprise-wide usage
Coordinating with IT operations and finance teams for ongoing monitoring
Testing agents
This module teaches learners how to effectively test, validate and ensure quality for Copilot Studio agents before deployment. Participants will explore Copilot Studio Kit testing capabilities, learn how to test agents at scale and understand the types of tests supported within the Copilot Studio ecosystem. By applying structured testing practices, learners will develop the skills to identify defects, improve performance and deliver reliable, production-ready Copilot agents that align with enterprise standards.
1. Introduction to Testing in Copilot Studio:
The importance of testing and validation in the Copilot agent lifecycle
How testing fits into the Application Lifecycle Management (ALM) and deployment process
Typical challenges in testing AI-driven and conversational systems
Core goals of agent testing:
Ensuring accuracy, reliability, and usability
Verifying data access and security
Confirming proper orchestration and workflow logic
2. Overview of the Copilot Studio Kit:
Introduction to the Copilot Studio Kit and its testing capabilities
Components of the Kit:
Test automation framework
Reporting and analytics tools
Configuration and environment setup utilities
How the Kit integrates with Power Platform, Azure DevOps, or GitHub Actions pipelines
Preparing the testing environment and data sets
Using the Kit for continuous testing in multi-environment deployments
3. Testing Agents at Scale:
Strategies for scaling agent testing across multiple environments and use cases
Simulating large-scale user interactions and load testing scenarios
Managing and tracking test execution across Dev, UAT, and Production environments
Automating tests as part of CI/CD pipelines (DevOps or GitHub Actions)
Monitoring performance and response times under real-world usage conditions
Identifying and resolving issues related to:
Latency and orchestration delays
Data retrieval and grounding (RAG) errors
Tool integration failures or misconfigurations
Ensuring governance compliance during automated test runs
4. Supported Test Types in the Copilot Studio Kit:
Overview of test types supported in Copilot Studio Kit:
Unit Tests: Validate individual topics, nodes and responses
Integration Tests: Verify that connectors, triggers and tools function correctly together
Regression Tests: Ensure existing functionality remains stable after changes or updates
Performance Tests: Evaluate speed, concurrency, and response times
Security Tests: Validate DLP adherence, authentication and permissions
Conversational Flow Tests: Assess natural language understanding, disambiguation and orchestration accuracy
Best practices for selecting appropriate test types for each phase of development
How to interpret test results and generate actionable reports
Additional Labs:
The following labs are provided with the course materials in order to gain additional hands on learning experience with Copilot Studio. They augment the labs aligned to each module and should be carried out once all the previous labs have been completed.
Advanced Lab 1: Create a Monthly Business Review (MBR) Agent
Advanced Lab 2: Build an Autonomous Account News Assistant Agent
Advanced Lab 3: Track conversation outcomes and user feedback on AI responses
Advanced Lab 4: Autonomous Portfolio Lookup Agent with Computer-Using Agents (CUA)
Advanced Lab 5: Deliver high-quality, scalable agents with Copilot Studio Kit
Advanced Lab 6: Model Context Protocol (MCP) & Copilot Studio
