Microsoft 365 Agents SDK and Copilot Extensions (SPO_COPILOT)
This five‑day course focuses on the design, development, validation, and operation of AI agents built using the Microsoft 365 Agents SDK and Copilot Extensions. The main emphasis is on building enterprise‑ready, multi‑channel agents that can be deployed across Microsoft 365 Copilot, Microsoft Teams, and web channels.
A 5‑day hands‑on developer course delivered exclusively for GOPAS
Course Overview
Participants learn the full engineering lifecycle—from architecture and tooling through action handlers and secure data access, to Responsible AI, evaluation telemetry, and deployment lifecycle management using Microsoft 365 Agents Toolkit and Agents Playground.
Who Should Attend
This course is ideal for:
- Technology consultants designing enterprise AI integrations
- Developers and architects building Copilot/Teams/web conversational experiences with code-first approach
- Teams responsible for operationalizing agents (validation, telemetry, rollout, versioning)
Prerequisites
Recommended prior knowledge:
- Optional/advantageous: prior work with Microsoft Graph
- Optional/advantageous: experience with prompt engineering
Training Format
- 5 days instructor‑led, hands‑on labs
- Tooling workflow using Microsoft 365 Agents Toolkit and Agents Playground
- Multi‑channel deployment patterns (Copilot, Teams, web)
- Focus on secure data anchoring (SharePoint/Graph) and operational readiness
What You Will Learn
1. Copilot Architecture & Agents SDK Fundamentals
- High‑level architecture of Copilot extensibility and where Agents SDK fits
- Extension options: Agents SDK, plugins, Graph connectors, actions
- Understanding “activities” and “turns” and how agent logic processes them
2. Multi‑Channel Deployment Model (Copilot / Teams / Web)
- Channel concepts and adapters (Azure Bot Service bridge)
- Deploying one agent to multiple user surfaces while keeping consistent behavior
- Patterns for channel-specific UX constraints and validation
3. Tooling: Microsoft 365 Agents Toolkit & Agents Playground
- Project templates (Visual Studio / VS Code), scaffolding and local testing
- Local development using Agents Playground
- Supported languages and prerequisites (C#, JavaScript/Node, Python)
4. Application Agent, Manifest & API Surface
- Agent application structure (AgentApplication / AgentApplicationOptions)
- Endpoint mapping and agent API surface fundamentals
5. Action Handlers & Microsoft Graph Integration
- Designing action handlers and orchestrations
- Integration patterns with Microsoft Graph and SharePoint as a secure data anchor
- Choosing runtime orchestration: Azure Functions vs Logic Apps vs Durable orchestration styles
6. Responsible AI, Evaluation & Telemetry
- Practical approach to Responsible AI in enterprise agents
- Qualitative vs quantitative evaluation, and “human‑in‑the‑loop” review patterns
- Telemetry foundations for validating agent quality and safety
7. Performance & Cost Control
- Token economy basics and prompt efficiency
- Cache layers and retrieval optimization for RAG scenarios
- Output sanitization and safe output handling (where applicable)
8. Lifecycle Management (Environments, Versioning, Rollback)
- Propagating agents across environments (DEV/TEST/PROD approach)
- Versioning strategies and rollback readiness
9. End‑to‑End Capstone
- Presentation of a complete end‑to‑end solution that combines Agents SDK with a RAG‑style data grounding approach
Outcomes
After the course, participants will be able to:
- Design and build code-first Microsoft 365 agents using the Agents SDK
- Deploy agents across Copilot, Teams, and web with the right channel adapters
- Implement secure Graph/SharePoint integrations and action handlers
- Establish evaluation telemetry and operational rollout practices for enterprise usage