Build a Foundation and Extend Microsoft 365 Copilot

Get started with Microsoft 365 Copilot

Microsoft 365 Copilot extensibility fundamentals

Explore how extending Microsoft 365 Copilot with agents, connectors, and plugins can help organizations address the challenges of increasing productivity and reducing costs by leveraging AI to orchestrate business processes more effectively.

Microsoft 365 Copilot is a powerful tool designed to enhance productivity by integrating AI capabilities into the daily workflows of users across various Microsoft 365 applications. However, the true potential of Microsoft 365 Copilot can be unlocked by extending its capabilities through the use of agents, connectors, and plugins. This extensibility allows organizations to tailor the tool to their specific needs, thereby addressing productivity challenges and reducing operational costs. Here’s a detailed explanation of how this can be achieved:

1. Agents:
   • Declarative Agents: These agents allow organizations to customize Microsoft 365 Copilot by declaring specific instructions, actions, and knowledge. Declarative agents run on the same orchestrator and foundation models that power Microsoft 365 Copilot, ensuring seamless integration and consistent performance .
   • Enhanced Collaboration and Productivity: By tailoring Copilot with declarative agents, organizations can automate complex processes such as team onboarding and customer issue resolution. This automation leads to smoother collaboration and increased productivity by providing consistent, personalized experiences .
   • Integration with Enterprise Data: Declarative agents can leverage enterprise data from Graph Connectors and SharePoint files, making it easier to engage with organizational data and streamline workflows .
2. Connectors:
   • Ingesting Enterprise Data: Microsoft Graph connectors allow organizations to enrich the organizational knowledge accessible to Copilot by ingesting enterprise data and content. This ensures that Copilot can draw from a comprehensive knowledge base, enhancing its ability to provide relevant and accurate responses .
   • Streamlined Workflows: By integrating external data sources, connectors help streamline workflows and ensure that users have access to all necessary information within the Microsoft 365 environment. This reduces the need for context switching and improves overall efficiency .
3. Plugins:
   • Custom Skills and Actions: Plugins enable organizations to extend Copilot’s skills by transforming their apps into plugins that increase user productivity across daily tasks and workflows. These plugins can retrieve data and execute tasks on external systems, further enhancing Copilot’s capabilities .
   • Seamless Integration: Developers can write plugins once and run them across the growing list of Microsoft 365 host applications. This ensures that the integration with Copilot is the last one they’ll ever need to perform, simplifying the development process and reducing costs .
   • Security and Compliance: Plugins, like declarative agents, inherit all data protections provided by Microsoft 365 Copilot, ensuring that security, privacy, and compliance are maintained. This allows business decision-makers and IT admins to confidently adopt and manage these technologies .
4. Overall Benefits:
   • Increased Productivity: By leveraging AI to automate and streamline business processes, organizations can significantly increase productivity. Users can stay in the flow of their work, using the apps and services that support their workflow as skills and knowledge accessible through Copilot’s natural language interface .
   • Reduced Costs: The ability to integrate and manage enterprise-grade technology centrally reduces the need for multiple disparate systems, leading to cost savings. Additionally, the automation of complex processes reduces the time and resources required to complete tasks .
   • Enhanced User Experience: The integration of agents, connectors, and plugins ensures that users have access to all necessary tools and information within a single interface. This enhances the user experience by reducing the need for context switching and providing a more cohesive workflow .

In summary, extending Microsoft 365 Copilot with agents, connectors, and plugins allows organizations to leverage AI to orchestrate business processes more effectively. This leads to increased productivity, reduced operational costs, and a more streamlined and efficient workflow. By customizing Copilot to meet their specific needs, organizations can ensure that their users have access to the tools and information they need to succeed.

Refrence: https://learn.microsoft.com/en-gb/training/modules/examine-microsoft-365-copilot-design/5-extend-copilot-microsoft-graph-connectors/

Refrence: https://learn.microsoft.com/en-us/training/modules/explore-copilot-stack/4-understand-different-copilot-extensibility-options/

Refrence: https://learn.microsoft.com/en-gb/training/modules/microsoft-copilot-extensibility-fundamentals/2b-what-are-declarative-agents/

Refrence: https://learn.microsoft.com/en-gb/training/modules/choose-copilot-extensibility-development-path/1a-why-extend-copilot/

Refrence: https://learn.microsoft.com/en-gb/training/modules/examine-microsoft-365-copilot-design/3-examine-key-components/

Introduction to Copilot Studio actions

Explore the core considerations and requirements for creating, testing, and publishing Copilot Studio actions to extend Microsoft 365 Copilot.

Introduction to Copilot Studio Actions

Explore the Core Considerations and Requirements for Creating, Testing, and Publishing Copilot Studio Actions to Extend Microsoft 365 Copilot

Creating, testing, and publishing Copilot Studio actions involves several key considerations and requirements to ensure that the actions effectively extend the capabilities of Microsoft 365 Copilot. Below is a detailed explanation of these core considerations and requirements:

1. Understanding the Types of Actions:
   • Prompt Actions: These actions return generative responses to user prompts within Microsoft 365 Copilot. They are designed to provide users with relevant information or perform specific tasks based on the input provided.
   • Connector Actions: These actions expand the knowledge and capabilities of Microsoft 365 Copilot by integrating external data sources. They enable Copilot to access and utilize data from various systems and applications.
   • Conversational Actions: These actions are tailored for predictable scenarios that require specific responses. They help in building structured conversations that guide users through predefined workflows or processes.
2. Development Tools:
   • Copilot Studio: A graphical, low-code tool that allows users to create custom agents and extend Microsoft 365 Copilot with enterprise data and scenarios. It simplifies the process of building and deploying actions.
   • Teams Toolkit: A development toolkit designed for building and deploying Teams applications, including declarative and custom engine agents.
   • SharePoint: Agents in SharePoint can answer questions about the content on any SharePoint site or document library. Users with edit permissions can create agents for specific tasks and share them with their team.
   • Visual Studio Code: For pro-code development, Visual Studio Code can be used to build declarative agents with more complex requirements.
3. Security and Trust:
   • Trusted Knowledge Sources: Ensure that the declarative agents use trusted knowledge sources and connect to trusted REST APIs. This helps in preventing potential security risks associated with untrusted data sources.
   • Security Controls: Microsoft 365 provides extensive controls governing who can acquire and use integrated apps. Admins can govern connectors used for both knowledge and custom actions to maintain security and compliance.
4. Testing and Validation:
   • In-Context Testing: Agents can be tested by @ mentioning them from Microsoft 365 Copilot business chat (BizChat). This allows for real-time validation of the agent’s responses and behavior.
   • Immersive Testing: Agents can also be accessed from BizChat for 1 to 1 conversations, providing a more immersive testing environment to ensure the agent performs as expected.
5. Publishing and Deployment:
   • Admin Controls: Admins have the ability to control the deployment of declarative agents within the organization. This includes setting permissions and ensuring that only authorized users can deploy and use the agents.
   • Continuous Monitoring: Once deployed, it is important to continuously monitor the performance and behavior of the agents to ensure they are functioning correctly and providing accurate responses.

By following these core considerations and requirements, organizations can effectively create, test, and publish Copilot Studio actions that extend the capabilities of Microsoft 365 Copilot, enhancing productivity and streamlining business processes.

References:

Refrence: https://learn.microsoft.com/en-gb/training/modules/microsoft-copilot-extensibility-fundamentals/4-add-skills-copilot-plugins/

Refrence: https://learn.microsoft.com/en-gb/training/modules/choose-copilot-extensibility-development-path/5-apply-data-privacy-security-practices-developing-plugin-connector/

Refrence: https://learn.microsoft.com/en-gb/training/modules/copilot-declarative-agent-intro/2-what-are-declarative-agents/

Create prompt actions for Microsoft 365 Copilot in Copilot Studio

Use Copilot Studio to create prompt actions that return generative responses to user prompts in Microsoft 365 Copilot.

To create prompt actions for Microsoft 365 Copilot in Copilot Studio that return generative responses to user prompts, follow these detailed steps:

Step 1: Understand the Role of Prompt Actions

Prompt actions in Microsoft 365 Copilot are designed to generate responses based on user inputs. These actions leverage the capabilities of AI to provide relevant and context-aware answers, enhancing the user experience by making interactions more intuitive and efficient.

Step 2: Set Up Your Environment

Before you begin creating prompt actions, ensure that you have the necessary tools and permissions: - Microsoft Copilot Studio: This is the primary tool for creating and managing prompt actions. - Microsoft Admin Center: Used for managing permissions and settings. - SharePoint Online: May be used for storing and managing data. - Visual Studio Code: Optional, for more advanced coding and customization.

Step 3: Define the Prompt Action

1. Identify the Use Case: Determine the specific scenario where the prompt action will be used. This could be anything from answering common customer queries to providing detailed explanations of product features.
2. Create a New Action: In Copilot Studio, navigate to the section for creating new actions. Select the option to create a prompt action.
3. Configure the Action: Define the parameters and settings for the prompt action. This includes specifying the input types, expected outputs, and any conditions that need to be met for the action to be triggered.

Step 4: Develop the Generative Response

1. Write the Prompt: Craft the initial prompt that will be used to generate the response. This should be clear and concise, providing enough context for the AI to generate a relevant answer.
2. Leverage AI Capabilities: Utilize the AI capabilities of Microsoft 365 Copilot to generate responses. This involves integrating with the underlying AI models that power Copilot, ensuring that the responses are accurate and contextually appropriate.
3. Test the Response: Before deploying the prompt action, thoroughly test it to ensure that it generates the desired responses. This may involve running multiple test scenarios and refining the prompt as needed.

Step 5: Deploy and Monitor the Action

1. Deploy the Action: Once the prompt action is ready, deploy it within your Microsoft 365 environment. This involves publishing the action and making it available to users.
2. Monitor Performance: After deployment, monitor the performance of the prompt action. Collect feedback from users and analyze the responses to ensure that the action is functioning as intended.
3. Refine and Improve: Based on the feedback and performance data, make any necessary adjustments to the prompt action. This may involve refining the prompt, adjusting the parameters, or updating the AI models.

Step 6: Ensure Security and Compliance

1. Use Trusted Data Sources: Ensure that the prompt action only uses data from trusted sources to prevent security risks. Avoid using untrusted data sources that could compromise the integrity of the responses.
2. Implement Security Controls: Utilize the security controls provided by Microsoft 365 to manage access and permissions for the prompt action. This includes setting up appropriate user roles and ensuring that sensitive operations require human intervention.

By following these steps, you can effectively create prompt actions in Copilot Studio that return generative responses to user prompts, enhancing the functionality and user experience of Microsoft 365 Copilot.

References

• : Discusses the importance of using trusted data sources and implementing security controls.
• : Provides an overview of how Copilot Studio actions leverage Microsoft business applications.
• : Explains the use of declarative agents and the tools available for building them.
• : Highlights the integration of agents, plugins, and connectors within the Microsoft 365 ecosystem.
• : Describes the role of plugins and their interaction with other systems in real-time.

Create connector actions for Microsoft 365 Copilot in Copilot Studio

Use Copilot Studio to create connector actions that expand the knowledge and capabilities of Microsoft 365 Copilot with external data.

To create connector actions for Microsoft 365 Copilot in Copilot Studio that expand its knowledge and capabilities with external data, follow these detailed steps:

Step-by-Step Guide to Creating Connector Actions

1. Understand the Role of Connectors:
   • Connectors allow developers and IT admins to integrate external data sources into Microsoft 365 Copilot, enhancing its ability to reason over a broader range of organizational knowledge .
   • Microsoft Graph connectors enable the ingestion of unstructured, line-of-business data into Microsoft Graph, facilitating unified data access and insights across Microsoft 365 and other services .
2. Set Up Your Development Environment:
   • Ensure you have access to Copilot Studio and the necessary permissions to create and manage connectors.
   • Familiarize yourself with the Microsoft Graph connectors APIs, which are essential for building custom connections that import content from external data sources into Microsoft Graph .
3. Define the External Data Source:
   • Identify the external data sources you want to integrate with Microsoft 365 Copilot. These could include databases, cloud services, or APIs from various platforms such as Salesforce, ServiceNow, or Google services .
   • Determine the type of data and the specific business functions it will support, such as sales, customer service, or operations .
4. Create and Register the Connector:
   • Use the Microsoft Graph connectors APIs to create and manage external connections. This involves defining and registering the schema of external data types and ingesting external data items into Microsoft Graph .
   • Follow the API documentation to ensure proper setup and configuration of the connector.
5. Ingest External Data:
   • Utilize Graph connectors to ingest data from various sources into Microsoft Graph. This process enables Copilot to access and summarize diverse datasets, providing more comprehensive insights and responses .
   • Ensure the data is correctly mapped and synchronized with Microsoft Graph to maintain data integrity and accessibility.
6. Leverage Power Platform Components:
   • Integrate Power Platform components such as Power Automate flows and certified connectors to define specific business behaviors that Copilot can address .
   • Use these components to automate workflows, trigger actions, and facilitate real-time data exchange between Copilot and external services .
7. Test the Connector Action:
   • Thoroughly test the connector action to ensure it functions as expected. This includes verifying data ingestion, synchronization, and the ability of Copilot to reason over the integrated data.
   • Use test scenarios that reflect real-world use cases to validate the connector’s performance and reliability.
8. Publish and Monitor the Connector:
   • Once testing is complete, publish the connector action in Copilot Studio.
   • Monitor its performance and gather feedback to make necessary adjustments and improvements.

By following these steps, you can effectively create connector actions in Copilot Studio that expand the knowledge and capabilities of Microsoft 365 Copilot with external data, enabling it to provide more comprehensive and insightful responses to user prompts.

Refrence: https://learn.microsoft.com/en-gb/training/modules/choose-copilot-extensibility-development-path/3-extensibility-benefits-limitations/

Refrence: https://learn.microsoft.com/en-gb/training/modules/microsoft-copilot-extensibility-fundamentals/3-add-knowledge-graph-connectors/

Create conversational actions for Microsoft 365 Copilot in Copilot Studio

Build conversational actions in Copilot Studio that extend Microsoft 365 Copilot with tailored conversations for predictable scenarios that require specific responses.

To build conversational actions in Copilot Studio that extend Microsoft 365 Copilot with tailored conversations for predictable scenarios requiring specific responses, follow these detailed steps:

Step-by-Step Guide to Building Conversational Actions

1. Understand the Use Case:
   • Identify the predictable scenarios where specific responses are required. These could be common customer service inquiries, internal team support, or any repetitive task that can be automated.
2. Set Up Copilot Studio:
   • Ensure you have access to Copilot Studio, which provides an end-to-end natural language or graphical development environment for creating and deploying copilots across multiple channels .
3. Define the Conversational Flow:
   • Map out the conversation flow, including the user prompts and the expected responses. This helps in structuring the conversation logically and ensures all possible user inputs are accounted for.
4. Create Prompt Actions:
   • Use Copilot Studio to create prompt actions that return generative responses to user prompts. This involves defining the specific business behavior that the action should address .
5. Leverage Power Platform Components:
   • Integrate Power Automate flows, certified connectors, or prompts to define the specific business behavior. These components help in managing the business functions and ensuring the actions are executed correctly .
6. Develop the Conversational Logic:
   • Implement the conversational logic using low-code or pro-code approaches. For less technical users, Copilot Studio offers a natural language or graphical development environment, while pro-code developers can utilize advanced functionality within the platform .
7. Test the Conversational Actions:
   • Thoroughly test the conversational actions to ensure they handle all predictable scenarios correctly. This includes testing for various user inputs and ensuring the responses are accurate and relevant.
8. Publish and Deploy:
   • Once the conversational actions are tested and refined, publish them within Copilot Studio. Deploy the actions across the desired channels, such as Microsoft Teams or Business Chat, to make them accessible to end-users .
9. Monitor and Optimize:
   • After deployment, monitor the performance of the conversational actions. Gather feedback from users and make necessary adjustments to improve the accuracy and relevance of the responses.

Example Scenario

Consider a scenario where a customer service agent is required to handle common inquiries about order status. The conversational action can be built to:

• Prompt the user to enter their order number.
• Retrieve the order status from an external database using a connector.
• Provide a tailored response based on the order status, such as “Your order is being processed” or “Your order has been shipped.”

By following these steps, you can build conversational actions in Copilot Studio that extend Microsoft 365 Copilot with tailored conversations for predictable scenarios, enhancing productivity and streamlining workflows.

Choose a Microsoft 365 Copilot extensibility development path

Guide to choosing the right approach for extending Microsoft 365 Copilot.

To choose the right approach for extending Microsoft 365 Copilot, it’s essential to understand the different extensibility options available and their respective features, development efforts, and use cases. Here is a detailed guide to help you make an informed decision:

1. Extend Microsoft 365 Copilot with Connectors and Plugins

Development Effort: - Simple (No code/low code)

Data Sources: - Microsoft Graph and data from other Graph connectors - External APIs through plugins - SharePoint

User Experience: - Business Chat

Behaviors, Instructions, Foundational Model: - No direct control over behaviors, instructions, or model parameters

Multi-Model Capable: - No

Model Support: - Provided by system

Responsible AI: - Provided by system

Support for Chat History: - Provided by system

Example Scenarios: - Users in Microsoft 365 Copilot can perform actions on or gain insights from external data and services.

This approach is ideal for scenarios where you need to integrate external data and services into Microsoft 365 Copilot with minimal development effort. It leverages existing Microsoft Graph connectors and plugins to enhance the capabilities of Copilot without requiring extensive coding.

2. Extend Copilot with Declarative Agents for Tailored Use Cases

Development Effort: - Moderate (Low code/Pro code)

Data Sources: - Microsoft Graph and data from other Graph connectors - External APIs through plugins - SharePoint

User Experience: - Microsoft 365 Copilot and Business Chat

Behaviors, Instructions, Foundational Model: - Direct control over prompting behaviors

Multi-Model Capable: - No

Model Support: - Provided by system

Responsible AI: - Provided by system

Support for Chat History: - Provided by system

Example Scenarios: - Introduce a GenAI assistant in existing application surfaces to reason over customer or external data. - Present your own agent or chatbot in an existing Microsoft surface, such as Teams or Business Chat, with a separate identity and experience to Copilot.

This approach is suitable for more tailored use cases where you need some level of control over the prompting behaviors. It allows you to create a more customized experience within the Microsoft 365 ecosystem, leveraging existing data sources and APIs.

3. Build Your Own Custom Engine Agent for a Fully Customized Solution

Development Effort: - Complex (Pro code)

Data Sources: - Can span multiple data sources, services, and apps inside or outside of Microsoft tenant

User Experience: - Microsoft Teams and other supported deployment channels

Behaviors, Instructions, Foundational Model: - Direct control of model parameters such as temperature, system prompts, max tokens, etc. - Custom agent behavior

Multi-Model Capable: - Yes

Model Support: - Choice of any model from OpenAI and full model catalog

Responsible AI: - Developer responsibility with platform options in each pattern

Support for Chat History: - Developer responsibility with platform options in each pattern

Example Scenarios: - Your customers and users interact with an agent fully customized to your brand and behavior, which can reason over multiple data sets and connected systems from multiple UI choices or conversational surfaces.

This approach is best for scenarios requiring a fully customized solution with complete control over the agent’s behavior and model parameters. It is suitable for complex use cases where you need to integrate multiple data sources and provide a unique user experience.

Summary

• Connectors and Plugins: Best for simple integrations with minimal development effort, leveraging existing Microsoft Graph connectors and plugins.
• Declarative Agents: Suitable for tailored use cases with moderate development effort, providing some control over prompting behaviors.
• Custom Engine Agents: Ideal for fully customized solutions with complex requirements, offering complete control over the agent’s behavior and model parameters.

By understanding these options and their respective features, you can choose the right approach to extend Microsoft 365 Copilot based on your specific needs and development capabilities .

Refrence: https://learn.microsoft.com/en-gb/training/modules/copilot-declarative-agent-intro/4-when-to-use-declarative-agents/

Cover various extensibility options, development tools, and methods.

To cover the various extensibility options, development tools, and methods for Microsoft 365 Copilot, we need to explore the different approaches available for extending Copilot, the tools used for development, and the methods involved in the process. Here is a detailed explanation:

Extensibility Options

Microsoft 365 Copilot offers three primary extensibility approaches:

1. Extend Microsoft 365 Copilot with Connectors and Plugins
   • Development Effort: Simple (No code/low code)
   • Data Sources: Microsoft Graph, data from other Graph connectors, external APIs through plugins, SharePoint
   • User Experience: Business Chat
   • Behaviors, Instructions, Foundational Model: No direct control over behaviors, instructions, or model parameters
   • Multi-Model Capable: No
   • Model Support: Provided by system
   • Responsible AI: Provided by system
   • Support for Chat History: Provided by system
   • Example Scenarios: Users in Microsoft 365 Copilot can perform actions on or gain insights from external data and services .
2. Extend Copilot with Declarative Agents for Tailored Use Cases
   • Development Effort: Moderate (Low code/Pro code)
   • Data Sources: Microsoft Graph, data from other Graph connectors, external APIs through plugins, SharePoint
   • User Experience: Microsoft 365 Copilot and Business Chat
   • Behaviors, Instructions, Foundational Model: Direct control over prompting behaviors
   • Multi-Model Capable: No
   • Model Support: Provided by system
   • Responsible AI: Provided by system
   • Support for Chat History: Provided by system
   • Example Scenarios: Introduce a GenAI assistant in existing application surfaces to reason over customer or external data. Present your own agent or chatbot in an existing Microsoft surface, such as Teams or Business Chat, with a separate identity and experience to Copilot .
3. Build Your Own Custom Engine Agent for a Fully Customized Solution
   • Development Effort: Complex (Pro code)
   • Data Sources: Can span multiple data sources, services, and apps inside or outside of Microsoft tenant
   • User Experience: Microsoft Teams and other supported deployment channels
   • Behaviors, Instructions, Foundational Model: Direct control of model parameters such as temperature, system prompts, max tokens, etc. Custom agent behavior
   • Multi-Model Capable: Yes
   • Model Support: Choice of any model from OpenAI and full model catalog
   • Responsible AI: Developer responsibility with platform options in each pattern
   • Support for Chat History: Developer responsibility with platform options in each pattern
   • Example Scenarios: Your customers and users interact with an agent fully customized to your brand and behavior, which can reason over multiple data sets and connected systems from multiple UI choices or conversational surfaces .

Development Tools

• Microsoft 365 Admin Center: Administrators can deploy or remove plugins for Copilot across the organization or for specific users or groups using the Microsoft 365 admin center .
• Copilot Studio: A platform for creating, testing, and publishing actions to extend Microsoft 365 Copilot. It supports creating prompt actions, connector actions, and conversational actions .
• Microsoft Graph: A key data source for connectors and plugins, providing access to a wide range of data within the Microsoft ecosystem .

Methods

1. Creating Prompt Actions: Use Copilot Studio to create actions that return generative responses to user prompts. This involves defining the prompt, specifying the expected response format, and testing the action to ensure it meets user needs .
2. Creating Connector Actions: Expand the knowledge and capabilities of Microsoft 365 Copilot by integrating external data sources through connectors. This involves setting up the connector, defining the data retrieval process, and ensuring secure data handling .
3. Creating Conversational Actions: Build tailored conversations for predictable scenarios that require specific responses. This involves scripting the conversation flow, defining the triggers, and testing the interactions to ensure they provide the desired user experience .

By leveraging these extensibility options, development tools, and methods, organizations can effectively enhance Microsoft 365 Copilot to meet their specific needs, improve productivity, and reduce costs through AI-driven business process orchestration.

Refrence: https://learn.microsoft.com/en-gb/training/modules/manage-agents-microsoft-365-copilot/3-enable-disable-copilot-extensibility/

Emphasize the importance of data privacy and security practices.

Emphasize the Importance of Data Privacy and Security Practices

When developing extensibility solutions for Microsoft 365 Copilot, it is crucial to prioritize data privacy and security practices. This ensures that the solutions not only enhance productivity but also maintain the integrity and confidentiality of organizational data. Here are the key considerations:

1. Data Privacy and Security in Microsoft 365 Copilot:
   • Microsoft 365 Copilot operates within the Microsoft 365 service boundary, ensuring that all data, including prompts, retrieved data, and generated responses, remain secure and compliant with Microsoft’s privacy and security commitments .
   • The Azure OpenAI Service, which powers Copilot, integrates seamlessly within the Microsoft 365 ecosystem, ensuring that data processing adheres to stringent security and compliance standards .
2. Encryption and Compliance:
   • Copilot ensures encrypted communication between your tenant and Copilot components, maintaining data confidentiality and security .
   • Existing security and compliance policies deployed by your organization continue to apply, ensuring that sensitive information is safeguarded .
3. Tenant Boundary and Data Protection:
   • The tenant boundary within Microsoft 365 ensures data privacy, location compliance, and adherence to security protocols. This means that your data remains within the Microsoft 365 service boundary, protected by robust security measures .
   • Declarative agents, which can be extended with plugins and actions, inherit all data protections provided by Microsoft 365 Copilot. This ensures that any data processed or actions performed by these agents are secure and compliant .
4. Visibility and Control for Enterprise Admins:
   • Enterprise admins have visibility and control over the distribution of declarative agents within their tenant via the Microsoft Admin Center. This allows for better management and oversight of how data is accessed and used within the organization .
5. External Data Integration:
   • When extending Copilot with Microsoft Graph connectors, external data is ingested into Microsoft Graph and remains within your tenant. This ensures that external data is subject to the same security and compliance measures as internal data .

By emphasizing these data privacy and security practices, organizations can confidently extend Microsoft 365 Copilot to enhance productivity while ensuring that their data remains secure and compliant with industry standards.

Refrence: https://learn.microsoft.com/en-gb/training/modules/examine-microsoft-365-copilot-design/4-explore-service-tenant-architecture/

Introduction to Graph connectors

Describe how to use Microsoft Graph connectors to enhance search within Microsoft 365 Copilot.

Introduction to Graph Connectors

Describe how to use Microsoft Graph connectors to enhance search within Microsoft 365 Copilot

Microsoft Graph connectors are a powerful feature that allows organizations to integrate external data sources into Microsoft 365. By using these connectors, you can bring content from various third-party services, such as file shares, databases, and SaaS applications, into Microsoft 365. This integration enables users to search and interact with this external content seamlessly within Microsoft 365 apps like Microsoft Search, SharePoint, and Office.com .

Here is a detailed explanation of how to use Microsoft Graph connectors to enhance search within Microsoft 365 Copilot:

1. Ingest External Data:
   • Microsoft Graph connectors provide a way to ingest unstructured, line-of-business data into Microsoft Graph. This allows Microsoft 365 Copilot to reason over the entirety of your enterprise content .
   • Content ingested through Graph connectors is added to the Microsoft Graph, which unlocks semantic understanding of user prompts in Microsoft 365 Copilot .
2. Unified Data Access and Insights:
   • Graph connectors enable data ingestion from various sources to Microsoft Graph, facilitating unified data access and insights across Microsoft 365 and other services .
   • This means that external data can power Microsoft 365 intelligent experiences such as Microsoft Search, making the data discoverable and usable within Microsoft 365 Copilot .
3. Enhance Semantic Understanding:
   • To ensure that Microsoft 365 Copilot uses your content effectively, administrators should add a urlToItemResolver when creating connections. This enables the platform to detect when users share URLs from your external content with each other, increasing the likelihood of displaying relevant content .
   • Applying semantic labels to your schema helps Microsoft 365 Copilot interpret the semantic meaning of your data, enhancing the search experience .
4. Manage Permissions and Access Control:
   • Microsoft 365 Copilot uses the same underlying data access controls as other Microsoft 365 services, ensuring that Copilot only accesses content that the current user is authorized to access .
   • External data ingested into Microsoft Graph is represented as external items with unique IDs and permissions that can be managed individually by associating an access control list (ACL) with a Microsoft Entra user and group ID or an external group .
5. Build Custom Connectors:
   • If the existing connectors do not cover all the services your organization integrates with, you can use the Microsoft Graph connectors API to build custom connectors. This allows you to bring your external data into Microsoft 365 experiences within your organization .
   • The API enables you to create and manage external data connections, define and register the schema of the external data types, ingest external data items into Microsoft Graph, and sync external groups to manage permissions .

By leveraging Microsoft Graph connectors, organizations can enhance the search capabilities within Microsoft 365 Copilot, making external data easily discoverable and usable, thereby improving productivity and reducing costs.

For more detailed information, you can refer to the following resources: - Microsoft Graph connector experiences - Build your first custom Microsoft Graph connector using APIs - Build your first custom Microsoft Graph connector using the SDK

Configure connections and integrate external data sources.

Introduction to Graph Connectors

Configure Connections and Integrate External Data Sources

Graph connectors are a powerful feature in Microsoft 365 Copilot that allow organizations to integrate external data sources into Microsoft 365. By using graph connectors, you can bring content from various third-party services, such as file shares, databases, and SaaS applications, into Microsoft 365. This integration enables users to search and interact with this external content seamlessly within Microsoft 365 apps like Microsoft Search, SharePoint, and Office.com .

Steps to Configure Connections and Integrate External Data Sources

1. Access the Connector Catalog:
   • Microsoft provides built-in connectors, known as Stock Connectors, which allow you to integrate external data sources into the Microsoft Graph and Microsoft 365 experiences. These connectors can be accessed via the Connector Catalog or the Microsoft 365 admin center .
2. Create and Manage External Data Connections:
   • Use the Microsoft Graph connectors API to build custom connectors if the existing connectors do not meet your needs. The externalConnection resource is a logical container for your external data that you can manage as a single unit .
   • To learn more about creating, updating, and deleting connections, refer to the Create, update, and delete connections in Microsoft Graph documentation.
3. Define and Register the Schema of External Data Types:
   • The schema API determines how various Microsoft 365 experiences use your content. You must register the schema before ingesting items into Microsoft Graph. The schema is a flat list of all the properties that you plan to add to the connection along with their attributes, labels, and aliases .
   • For more details, see Register and update schema for the Microsoft Graph connection.
4. Ingest External Data Items into Microsoft Graph:
   • Microsoft Graph connectors offer an intuitive way to bring external data into Microsoft Graph. Each item from your data source must be represented as an externalItem in Microsoft Graph with a unique item ID. This ID is used to create, update, or delete the item from Microsoft Graph .
   • For more information, see Create, update, and delete items added by your application via Microsoft Graph connectors.
5. Sync External Groups:
   • The system can grant or deny access to items in the external service through an access control list (ACL) to different types of non-Microsoft Entra groups. When you ingest these items into Microsoft Graph, you need to honor these ACLs. The externalGroup API is used to set permissions on external items ingested into Microsoft Graph .
   • To learn more, see Use external groups to manage permissions to Microsoft Graph connectors data sources.
6. Enhance Integration with Microsoft 365 Copilot:
   • Add a urlToItemResolver when you create your connection to enable the platform to detect when users share URLs from your external content with each other. This increases the likelihood of Microsoft 365 Copilot displaying content shared with that user .
   • Apply semantic labels to your schema to help Microsoft 365 Copilot interpret the semantic meaning of your schema .
   • Add user activities on your items to assign greater importance to items that have more activities .

By following these steps, you can effectively configure connections and integrate external data sources into Microsoft 365, enhancing the capabilities of Microsoft 365 Copilot and other intelligent experiences within the platform.

Refrence: https://learn.microsoft.com/en-gb/training/modules/introduction-graph-connectors/2-add-connection/

Improve query understanding.

Introduction to Graph Connectors

Improve Query Understanding

Improving query understanding is a critical aspect of leveraging Microsoft Graph connectors to enhance the capabilities of Microsoft 365 Copilot. By integrating external data sources into Microsoft 365, organizations can ensure that Copilot can reason over a broader set of enterprise content, leading to more accurate and relevant responses to user prompts. Here are the key steps and considerations for improving query understanding using Graph connectors:

1. Ingest External Data:
   • Microsoft Graph connectors allow you to ingest unstructured, line-of-business data into Microsoft Graph. This data can come from various sources such as file shares, databases, and SaaS applications .
   • By bringing this external content into Microsoft 365, users can search and interact with it seamlessly within Microsoft 365 apps like Microsoft Search, SharePoint, and Office.com .
2. Semantic Understanding:
   • Content ingested through Graph connectors is added to the Microsoft Graph, which unlocks semantic understanding of user prompts in Microsoft 365 Copilot .
   • This means that Copilot can better interpret the context and intent behind user queries, leading to more accurate and relevant responses.
3. Apply Semantic Labels:
   • To enhance query understanding, administrators should apply semantic labels to the schema of the ingested data. Semantic labels help Microsoft 365 Copilot interpret the semantic meaning of the data, making it easier to match user queries with the relevant content .
   • Applying as many semantic labels as applicable to your schema ensures that Copilot can understand and utilize the data more effectively.
4. Add User Activities:
   • Adding user activities on the ingested items can further improve query understanding. The system assigns greater importance to items that have more activities, making them more likely to be surfaced in response to user queries .
   • For a list of supported user activity types, administrators can refer to the external activity documentation .
5. Use urlToItemResolver:
   • Adding a urlToItemResolver when creating your connection enables the platform to detect when users share URLs from your external content with each other. This increases the likelihood that Microsoft 365 Copilot will display content shared with that user .
   • Including a urlToItemResolver in the activity settings when creating the connection is a crucial step to ensure effective content sharing and discovery.

By following these steps, organizations can significantly improve the query understanding capabilities of Microsoft 365 Copilot, leading to more accurate and relevant responses to user prompts. This, in turn, enhances productivity and ensures that users can easily find and interact with the information they need.

For more detailed information, administrators can refer to the additional reading on how to make Graph connectors work better with Copilot .

Refrence: https://learn.microsoft.com/en-gb/training/modules/introduction-graph-connectors/6-deploy-monitor-connector/

Introduction to declarative agents for Microsoft 365 Copilot

Determine scenarios suitable for customizing and extending Microsoft 365 Copilot with declarative agents.

To determine scenarios suitable for customizing and extending Microsoft 365 Copilot with declarative agents, consider the following factors:

1. Specific Subject or Task Optimization

Declarative agents are ideal when you have a specific subject or task that you want to optimize for. These agents work well in scenarios where the goal is clear and well-defined. For example, if you need to streamline team onboarding processes or resolve customer issues efficiently, declarative agents can be tailored to handle these specific tasks effectively .

2. Reuse of Existing Infrastructure

Declarative agents utilize the same orchestrator and foundational models as Microsoft 365 Copilot. This means they can leverage the existing infrastructure and controls, providing a consistent experience for end users. If your scenario can benefit from this consistency and does not require custom orchestration or a different large language model, declarative agents are a suitable choice .

3. Data Availability within Microsoft 365

Consider whether the information relevant to your scenario is available within Microsoft 365. Declarative agents are effective when the necessary data is already present in Microsoft 365. If you need to ingest data from external systems, you can use Graph connectors. For real-time interactions with external systems, custom actions can be employed .

4. Development Effort and Complexity

Declarative agents require a moderate level of development effort, involving low code or pro code. This makes them suitable for scenarios where you need a balance between customization and ease of development. If your scenario demands extensive customization and flexibility, you might need to consider building a custom engine agent instead .

5. User Experience and Deployment Channels

Declarative agents can be integrated into Microsoft 365 Copilot and Business Chat, providing a tailored user experience. They are suitable for scenarios where you want to introduce a GenAI assistant in existing application surfaces, such as Teams or Business Chat, with a separate identity and experience from Copilot .

Example Scenarios

• Team Onboarding: Automate the onboarding process for new team members by providing them with a personalized assistant that guides them through the necessary steps and resources.
• Customer Support: Create an agent that can handle common customer queries and issues, providing quick and consistent responses.
• Internal Knowledge Management: Develop an agent that helps employees find information and resources within the organization, improving productivity and collaboration.

By considering these factors, you can determine whether declarative agents are the right choice for your specific scenario and how they can be effectively utilized to extend Microsoft 365 Copilot .

Describe the function of custom knowledge and custom actions.

Introduction to Declarative Agents for Microsoft 365 Copilot

Describe the Function of Custom Knowledge and Custom Actions

Custom Knowledge: Custom knowledge in Microsoft 365 Copilot refers to the additional data and context that can be provided to the system to tailor its responses and actions to specific scenarios or tasks. This custom knowledge is essential for making the Copilot more relevant and reliable in its responses by grounding it with expanded knowledge that respects data access limitations from the knowledge source itself .

Custom knowledge consists of two parts: 1. Custom Instructions: These are the core set of instructions that define the behavior and capabilities of the agent. They can include specific guidelines on how the agent should process information and respond to user queries. 2. Custom Data Sources: These are the external data sources that the agent can access to retrieve information. By integrating these data sources, the agent can provide more accurate and contextually relevant responses .

Custom Actions: Custom actions are specific tasks or operations that the declarative agent can perform based on the instructions provided. These actions are invoked via REST API descriptions configured by the declarative agent. Custom actions enable the agent to automate and execute specific business processes on behalf of the user, such as retrieving information, summarizing data, sending emails, or updating records .

Custom actions are designed to be secure and controlled, ensuring that they only use trusted knowledge sources and connect to trusted REST APIs. This is crucial to prevent potential security risks, such as an attacker crafting a message payload to manipulate the agent’s behavior .

In summary, custom knowledge and custom actions are fundamental components of declarative agents in Microsoft 365 Copilot. Custom knowledge provides the necessary data and context for the agent to operate effectively, while custom actions enable the agent to perform specific tasks and automate business processes, enhancing productivity and efficiency within the organization.

Refrence: https://learn.microsoft.com/en-gb/training/modules/microsoft-copilot-extensibility-fundamentals/2-extension-fundamentals/

Refrence: https://learn.microsoft.com/en-us/training/modules/utilize-azure-openai-model-create-app/5-bring-azure-open-ai-model-data-teams/

Refrence: https://learn.microsoft.com/en-gb/training/modules/copilot-declarative-agent-intro/3-how-declarative-agent-work/

Manage agents for Microsoft 365 Copilot

Manage agents using the Microsoft 365 admin center.

To manage agents for Microsoft 365 Copilot using the Microsoft 365 admin center, follow these detailed steps:

1. Access the Microsoft 365 Admin Center:
   • Navigate to the Microsoft 365 admin center. This is the central hub for managing all aspects of Microsoft 365, including Copilot agents.
2. Enable or Disable Copilot Plugin Extensibility:
   • In the Integrated Apps section of the Microsoft 365 admin center, you can enable or disable Copilot plugin extensibility at the tenant level. This setting controls who can access Copilot plugins within your organization. The setting has three options:
     • No one: No one in your organization can use Copilot agents, and the third-party extensions do not appear in the Copilot plugin pane. This choice also hides the apps that have Copilot agents from the Available apps and Deployed apps lists in the Microsoft 365 admin center .
3. Deploy or Remove Agents:
   • To deploy or remove agents, go to the Available apps or Deployed apps list in the Microsoft 365 admin center. Select the app that works with Copilot and choose Deploy or Remove. You can decide to add or remove the app for everyone in the organization or for specific users or groups .
   • Note that adding or removing an app will change how the app functions in Copilot and other Microsoft 365 products like Outlook, Teams, etc. .
4. Manage Security, Privacy, and Compliance:
   • Declarative agents are built on a secure foundation, inheriting all data protections provided by Microsoft 365 Copilot. Enterprise admins have visibility and control over the distribution of declarative agents within their tenant via the Microsoft admin center. This ensures that security, privacy, and compliance requirements are met .

By following these steps, you can effectively manage agents for Microsoft 365 Copilot, ensuring that the right users have access to the necessary tools while maintaining security and compliance standards.

Tailor Copilot by limiting or granting permissions to data based on organizational needs.

To tailor Microsoft 365 Copilot by limiting or granting permissions to data based on organizational needs, you need to understand and leverage the data access controls and permissions management features provided by Microsoft 365 and Microsoft Graph. Here is a detailed explanation of how to manage these permissions effectively:

Understanding Data Access Controls

1. User Identity-Based Access:
   • Microsoft 365 Copilot uses the same underlying data access controls as all other Microsoft 365 services. This means that Copilot has the same access to data as the user submitting the prompt .
   • Microsoft Graph honors the user identity-based access boundary, ensuring that Copilot only accesses content that the current user is authorized to access .
2. Microsoft Graph API:
   • Microsoft Graph is the connective tissue that binds all your Microsoft 365 services and data together. It synthesizes and searches content from multiple sources within your tenant, such as emails, chats, documents, and meetings .
   • The API brings more context from user signals into the prompt, ensuring that users don’t need to navigate away or switch apps. It enables Microsoft 365 Copilot to bring the relevant information to you while respecting user permissions, data security, and compliance policies .

Managing Permissions

1. Ingesting External Data:
   • When you ingest external data into Microsoft Graph with a Graph connector, the data is represented as external items with a unique ID and permissions that can be managed individually .
   • You can associate an access control list (ACL) with a Microsoft Entra user and group ID or an external group to manage permissions for these external items .
2. Customizing Permissions:
   • You can customize permissions to tailor Copilot’s access to data based on organizational needs. This involves setting up and managing ACLs to ensure that only authorized users can access specific data.
   • Permissions can be granted or limited based on roles, departments, or specific user needs, ensuring that sensitive information is only accessible to those who need it.

Ensuring Compliance and Security

1. Data Security and Compliance:
   • Microsoft 365 Copilot takes into account Microsoft 365 user permissions, data security, and compliance policies. It only generates responses based on the information the user has permission to access .
   • Prompts, responses, and data accessed through Microsoft Graph are not used to train foundation Large Language Models (LLMs), including those used by Microsoft 365 Copilot, ensuring data privacy and security .

Practical Steps to Tailor Permissions

1. Using Copilot Studio:
   • Copilot Studio is a graphical, low-code tool for creating custom agents and extending Microsoft 365 Copilot with your own enterprise data and scenarios .
   • You can use Copilot Studio to define and manage permissions for the data that your custom agents can access, ensuring that they operate within the boundaries set by your organization.
2. Creating and Managing Agents:
   • Use the Copilot Studio agent builder to easily create and customize agents that can be used with Microsoft 365 Copilot .
   • Agents can be tailored to answer questions about specific content, such as SharePoint sites or document libraries, based on the permissions of the asker .

By understanding and leveraging these data access controls and permissions management features, you can effectively tailor Microsoft 365 Copilot to meet your organizational needs, ensuring that data is accessed securely and in compliance with your policies.