How to Build an Agent in Microsoft 365 Copilot Chat (Beginner Tutorial)

Creating an HR Policy Helper Agent

Introduction to Building the Agent

• The session focuses on creating an agent designed to assist employees in understanding policies and answering related questions.

• The starting point is the Microsoft 365 co-pilot chat experience, where users can create an agent by providing a description of its purpose.

Defining the Agent's Purpose

• Users are encouraged to describe the agent's role, which includes answering questions about policies and providing instructions on how it should behave.

• A natural language interface allows for easy input of descriptions, with templates available for learning purposes.

Crafting the Agent's Identity

• The initial description should clarify that the agent is a "policy helper," which can be renamed later if desired.

• It's important to define the agent’s purpose clearly: helping users get answers regarding human resources policies.

Establishing Persona and Tone

• The agent can be framed with a persona, such as acting as an expert in human resources familiar with all relevant policies.

• Providing examples of topics like leave, performance management, and health & safety helps define the scope of what the agent will cover.

Refining Instructions and Feedback Mechanism

• After entering a description, users receive suggestions for prompts based on their input; this aids in refining what tasks or actions the agent should perform.

How to Configure an HR Policy Advisor Agent

Handling Insufficient Information

• It's crucial for agents to clarify whether they should use general knowledge or only the provided information when answering questions. If uncertain, users are advised to contact their HR representative.

• The choice of using general knowledge can be beneficial in some scenarios but may lead to inaccuracies in others. Therefore, it's essential to restrict responses to known data.

Configuring the Agent

• Users can customize the agent's name and description based on their needs. This customization allows for a more tailored experience.

• Instructions for the agent can include providing answers in clear bullet points and step-by-step guidance, enhancing user comprehension.

Integrating Knowledge Sources

• Users can link a SharePoint site containing HR policies as a knowledge base for the agent. This ensures that all responses are grounded in relevant documentation.

• It’s important to disable web search capabilities if the goal is to limit answers strictly to internal resources.

Future Capabilities and User Experience

• Upcoming features may allow agents to perform actions like sending emails or retrieving information from other systems, expanding their functionality.

• Suggested prompts generated during setup help streamline user interactions by allowing quick access to frequently asked questions.

Testing and Sharing the Agent

• Once created, agents can be shared within organizations, making them accessible even without full Microsoft 365 co-pilot licenses.

• The ability to create specialized agents provides a solution for organizations that do not have comprehensive licensing but still need targeted assistance with specific queries.

Utilizing Suggested Prompts

HR Policy Advisor Functionality

Overview of HR Policy Queries

• The system is designed to provide straightforward answers to specific HR-related questions, such as policies on compassionate leave and the number of days allotted for it.

• Employees can query their current leave balance, showcasing the system's capability to access real-time data from integrated systems.

Handling General Employee Issues

• The system can interpret general employee concerns, like problems with a manager, and provide actionable suggestions in bullet point format.

• It is important that the system remains focused on HR-related queries and does not stray into unrelated topics.

Limitations and Testing Phase

• An example was given about asking how to make chocolate cake, illustrating that the system should only respond to HR-specific inquiries.

• Users may need to refine or correct their queries frequently during this testing phase, which is part of optimizing the user experience.

System Configuration Adjustments

• The settings allow users to restrict responses strictly based on provided knowledge; if a question falls outside this scope, the system will indicate it cannot help.

• Continuous learning and adjustments are necessary for improving response accuracy; users are encouraged to experiment with different configurations.

Iteration and Feedback Importance