Overview
This module introduces learners to AI agents and how they can be designed to support specific business tasks, retrieve information and automate selected activities.
Learners will begin by understanding how effective instructions, contextual grounding and structured prompting influence an agent's responses. They will then explore different types of AI agents and learn how to build custom agents using Microsoft Copilot Studio, OpenAI Custom GPTs, Claude and Gemini Pro.
The module includes practical work on creating Retrieval-Augmented Generation (RAG) agents, connecting them to business knowledge sources such as SharePoint and OneDrive, and testing their responses for accuracy and relevance.
By the end of this module, learners will understand how to create, test and deploy AI agents that can work with business information and support specific organisational requirements.
Tools Covered: Instructional Prompting | Microsoft Copilot Studio | Google AI Studio | Claude Sonnet & Haiku | OpenAI Custom GPTs | Gemini Pro | SharePoint | OneDrive | Model Context Protocol (MCP)
What Will Be Covered
9.1 Foundations of Instructional Prompting, Contextual Grounding, Chain-of-Thought Prompting, Linked Prompts & Comparative Prompts
Learn how to provide clear instructions and relevant context so AI agents can produce more useful and consistent responses.
Topics include:
Understanding instructional prompting and how to structure instructions for AI systems.
Defining an agent's role, purpose, responsibilities and expected output.
Using contextual grounding to provide relevant information for AI responses.
Understanding step-by-step reasoning prompts and how to guide AI through complex tasks.
Creating linked prompts to complete tasks through a sequence of related instructions.
Using comparative prompts to evaluate alternatives, identify differences and support decisions.
Improving prompt quality through clear objectives, constraints and expected output formats.
Practical example: Create a sequence of prompts for an HR assistant that reviews employee policy information, compares leave conditions and prepares a response based on the organisation's guidelines.
9.2 AI Agent Types Overview – Retrieval Agents vs Augmented Agents vs Automated Agents vs AI Agentic Framework
Explore different types of AI agents and understand how they support information retrieval, task execution and business process automation.
Topics include:
Introduction to AI agents and their role in business applications.
Understanding retrieval-based agents that find and provide information from connected knowledge sources.
Exploring augmented agents that use additional context, tools and capabilities to improve responses.
Understanding automated agents that perform defined tasks based on instructions and triggers.
Introduction to agentic AI frameworks and how agents can plan and coordinate multiple steps.
Understanding the differences between conversational assistants, retrieval agents and task-oriented agents.
Identifying suitable agent types for different business requirements.
Practical example: Compare an employee policy assistant that retrieves information, a reporting assistant that analyses supplied data and an automated agent that initiates an approval workflow.
9.3 Copilot Studio Overview and Capabilities Deep Dive – Instructions, Topics, Tools, MCPs, Knowledge, Channels & Evaluations
Understand how Microsoft Copilot Studio can be used to create and manage AI agents for business applications.
Topics include:
Introduction to Copilot Studio and its agent-building environment.
Configuring agent instructions, roles and response behaviour.
Creating topics to manage conversations and specific user requirements.
Connecting tools and actions to extend agent capabilities.
Understanding Model Context Protocol (MCP) and how it supports connections between AI agents and external tools or services.
Adding knowledge sources to provide relevant business information.
Understanding deployment channels and how agents can be made available to users.
Exploring evaluation capabilities to test agent responses and performance.
Reviewing agent behaviour and refining instructions based on test results.
Practical example: Create an internal employee support agent with defined instructions, knowledge sources and conversation topics to answer common HR and administrative questions.
9.4 Build, Test & Deploy a Custom RAG Agent with Copilot, SharePoint & OneDrive
Learn how to build a Retrieval-Augmented Generation (RAG) agent that uses organisational documents to provide context-based responses.
Topics include:
Understanding the RAG approach and how retrieval works with generative AI.
Preparing business documents and knowledge sources for an agent.
Connecting SharePoint and OneDrive as information sources.
Configuring an agent to retrieve relevant information before generating responses.
Creating instructions to guide how the agent uses retrieved content.
Testing the agent with different questions and business scenarios.
Reviewing response accuracy and checking whether answers are supported by source documents.
Preparing the agent for deployment and user access.
Practical example: Build an internal knowledge assistant that answers employee questions using company policies, training documents, process guidelines and other approved resources stored in SharePoint and OneDrive.
9.5 Build a RAG Custom GPT with OpenAI, Claude & Gemini Pro – Compare and Demonstrate Capabilities
Explore how custom AI assistants can be created using different AI platforms and how their capabilities can be compared for business applications.
Topics include:
Understanding custom GPTs and their role in creating task-specific AI assistants.
Preparing instructions and knowledge files for a custom assistant.
Configuring a RAG-based assistant using available knowledge and retrieval capabilities.
Exploring custom assistant capabilities across OpenAI, Claude and Gemini Pro.
Understanding how each platform handles instructions, uploaded documents and contextual information.
Comparing responses based on relevance, accuracy, consistency and usefulness.
Identifying suitable platforms based on business requirements and use cases.
Practical example: Create a custom assistant using business documentation and test the same set of questions across different platforms to compare how effectively each one retrieves and explains the information.
9.6 Evaluations & Grounding for RAG Agents
Understand how to assess the quality of RAG agents and ensure their responses are supported by relevant information.
Topics include:
Understanding the importance of evaluation in AI agent development.
Checking whether an agent retrieves relevant information from its knowledge sources.
Evaluating response accuracy, relevance and completeness.
Understanding grounding and how responses should be supported by source content.
Identifying unsupported answers, incorrect information and potential hallucinations.
Testing agents with different questions, scenarios and variations in wording.
Refining instructions, knowledge sources and retrieval settings based on evaluation results.
Establishing a review process to maintain agent quality over time.
Practical example: Test a company policy assistant using questions with clear answers, incomplete information and questions outside its knowledge base. Review whether the agent provides supported answers or appropriately indicates when the information is unavailable.
Skills You Will Gain
By completing this module, learners will develop the ability to:
Write structured instructions for task-specific AI agents.
Apply contextual grounding and linked prompting techniques.
Understand the differences between retrieval, augmented and automated agents.
Identify suitable AI agent approaches for business requirements.
Configure agents using Microsoft Copilot Studio.
Work with agent instructions, topics, tools, knowledge sources and channels.
Understand how MCP can connect agents with external tools and services.
Build RAG agents using SharePoint and OneDrive.
Create and configure custom GPTs and explore comparable assistants in Claude and Gemini.
Test agent responses against business knowledge sources.
Evaluate response accuracy, relevance and grounding.
Improve agent performance through testing and refinement.
Learning Outcomes
After completing this module, learners will be able to design and evaluate AI agents that support everyday business activities and improve access to organisational information.
Faster access to business information: Build knowledge assistants that help employees find answers from company policies, process documents, training materials and other approved resources without manually searching through multiple files.
Improved employee support: Develop agents for HR, administration, IT support and other internal functions to handle common questions and provide consistent information.
Better use of organisational knowledge: Connect AI agents with SharePoint and OneDrive content so employees can interact with existing business documents through natural language.
Reduced repetitive work: Identify suitable tasks for AI agents, such as responding to routine enquiries, retrieving information and supporting defined business processes.
More consistent business communication: Use structured instructions and contextual grounding to guide agents in producing responses that follow business requirements and approved information.
Improved decision support: Develop assistants that can retrieve, organise and explain relevant business information to help employees and managers make informed decisions.
Better control over AI responses: Evaluate agent outputs, identify unsupported answers and refine knowledge sources and instructions to improve reliability.
Practical AI adoption: Understand how to select suitable platforms and agent-building approaches based on business needs, available resources and intended outcomes.
AB-731 Exam Alignment
Learning area | Relevance |
|---|---|
Understanding AI capabilities | Introduces AI agents, generative AI and different approaches to building task-specific assistants. |
Identifying business use cases | Demonstrates how agents can support employee services, information retrieval and business processes. |
Microsoft AI solutions | Provides practical exposure to Copilot Studio and its capabilities for building business agents. |
AI integration and adoption | Explores how agents can connect with organisational knowledge sources and support business workflows. |
Evaluating AI outputs | Develops an understanding of response quality, grounding, accuracy and the need for human review. |
Continue to the Next Module
You have now explored how to design, build, test and deploy AI agents using different platforms and organisational knowledge sources.
You will be able to apply these concepts to create task-specific assistants, improve access to business information and identify opportunities to use AI agents across different business functions.
Explore how these agent capabilities can be applied to more advanced business workflows and AI-powered solutions.
Still need help?
Contact us