Course
digicode: MCPAI
M365 Copilot Adoption with Business Objective Focus – Advanced
Course facts
Download as PDF- Creating custom agents in Agent Builder to automate repeatable business tasks and streamline workflows
- Designing effective agents by incorporating clear instructions, scoped knowledge sources, and practical usage patterns
- Implementing governance and security measures to manage data access, prevent prompt injection, and reduce organizational risks
- Leveraging data analysis tools like Excel Copilot and Analyst Agent for tasks such as data reconciliation, enrichment, and forecasting
- Optimizing agent performance through rigorous testing, iterative improvement, and prompt engineering best practices
- Hardening agents for production environments by adding guardrails, self-check blocks, and adversarial defenses
- Managing agent lifecycles and knowledge hygiene to ensure long-term reliability, scalability, and maintainability
- Deciding when to move beyond basic Agent Builder scenarios toward advanced solutions like Copilot Studio, custom tools, and AI workflows
Recommended course version: This self-developed course covers the topics of the official Microsoft course «MS-4019: Transform your everyday business processes with agents».
In this course, you will learn how to simplify complex use cases with individually configured Copilot agents, develop tailor-made scenarios, and optimally integrate AI into demanding processes. After completing the course, you will be able to use Copilot specifically with your own agents for challenging tasks and design innovative, cross-team solutions.
This course covers only «Copilot Studio light» agents. These can be easily created by all business users without the need for IT or development skills. Complex «Copilot Studio Full Experience» agents are covered in course PL-7008: Create Custom Agents in Microsoft Copilot Studio.
1 Creating M365 Copilot Agents
Learn how to create, configure, test, publish, share, and use Microsoft 365 Copilot agents for repeatable business tasks. The module focuses on declarative Agent Builder agents: saved Copilot instructions connected to explicitly selected knowledge sources and used from Copilot Chat.
Exercise: Participants create and improve practical agents, starting with a Meeting Summary Generator and optional examples such as a Personal Translator, Social Media Buddy, Presentation Buddy, Data Insight agent, and Call Prep Buddy. They also test scheduled prompts with agents, edit existing agents, and combine two agents in a single chat.
2 Governance & Security
Understand how agents access Microsoft 365 data, why oversharing and prompt injection matter, and how builders and IT can reduce risks through scoped knowledge, permissions, sensitivity labels, DLP, and lifecycle ownership.
Exercise: Hands-on security and governance experiments:
- Run an oversharing test with a broadly connected SharePoint or OneDrive source.
- Rebuild the same agent with a specific folder or file and compare the results.
- Test how prompt content can influence internet search and potentially expose sensitive information.
- Use a document containing prompt-injection instructions to observe how agents can be manipulated and why source trust matters.
3 Data Agents
Learn how to use Excel Copilot, Analyst Agent, and custom agents with Code Interpreter for data preparation, reconciliation, enrichment, data-quality checks, forecasting, and report-ready executive summaries.
Exercise: Hands-on data exercises using Excel Copilot and Analyst Agent:
- Transform human-friendly Excel layouts into machine-ready tables for Power BI and downstream analysis.
- Reconcile risk-team and core-banking extracts and explain discrepancies in business language.
- Enrich project trackers from PDF memos and identify missing evidence without guessing.
- Use Analyst Agent to join datasets, run quality checks, detect duplicates and orphan records, create forecasts, and draft a one-page portfolio summary.
- Create a custom Code Interpreter agent for repeatable reconciliation or project-tracker update workflows.
4 Surviving Production & Beyond Agent Builder
Learn how to keep agents useful after launch, harden them against misuse, evaluate their outputs, manage knowledge hygiene, and decide when Agent Builder is no longer enough. The module also introduces Copilot CoWork, Copilot Studio, skills, tools, orchestration, and AI workflows as the next step beyond basic Agent Builder scenarios.
Exercise: Advanced, scenario-driven exercises:
- Build a multi-source production agent that uses multiple knowledge files and five suggested prompts as workflow paths.
- Apply few-shot examples, structured outputs, self-check instructions, adversarial framing, and prompt-injection defenses.
- Run adversarial prompts and score how safely the agent responds.
- Review where Agent Builder hits limits and map scenarios to Copilot Studio, CoWork, tools, skills, connected agents, and workflows.
- Create a practical next-step plan: audit one agent, define golden questions, review knowledge sources, and clarify ownership.
This training is structured in four business objective areas and contains theory and hands-on lab exercises focusing on processes with M365 Copilot.
This training is intended for users who already have extensive experience with M365 Copilot and prompting and now want to create their own agents for specific tasks.
- The course primarily covers creating your own agents. This feature must be enabled in your own environment so that the course content can be implemented in everyday use (functionality is included in the M365 Copilot license, Copilot Studio light only).
- Advanced user knowledge analogous to the following course (or equivalent knowledge) and at least 4 months of practical experience with M365 Copilot:
The configuration of M365 Copilot in your company may differ from the standard configuration. All standard functions are available in the training environment.