Microsoft 365 Copilot: Create Chat Buddy

Key insights

• Demo & presenter: The video demo shows a "Communication Buddy" built with Microsoft 365 Copilot, a .NET API, PnP Core SDK and SharePoint.
  Paul Bullock (Avanade) demonstrates drafting, saving and managing communications with delegated authentication and API plugins.
• Declarative Agents: These are customizable Copilot assistants defined with declarative JSON instead of imperative code.
  They let organizations add instructions, knowledge and workflows so the agent follows business rules and style guides.
• API Plugins: Plugins connect the agent to enterprise systems so it can read and write data, not just generate text.
  The demo shows secure writes to SharePoint lists using delegated permissions and single sign-on so actions respect user access.
• How it works: Define the agent with JSON, add API plugins to link systems, and deploy the agent into Copilot-enabled apps like Teams or Word.
  Users chat to draft messages while the agent can call APIs to save drafts or update records in real time.
• Key benefits: The approach boosts productivity with tailored writing help, reduces manual steps, and integrates actions into workflows.
  It runs within Microsoft 365 security models so data and operations remain permission-aware.
• What’s new and practical: Combining declarative agents with API plugins moves Copilot from a text tool to an interactive assistant that performs business actions.
  The demo provides a secure, reusable template organizations can adapt for internal communication workflows.

The Microsoft-authored YouTube demo presented by Paul Bullock on May 8, 2025, showcases a working assistant called the Communication Buddy built with Microsoft 365 Copilot, declarative configuration, and API plugins. In this newsroom summary, we explain how the demo ties conversational AI to enterprise systems and why this matters for organizations adopting productivity AI. Moreover, the video highlights a sample architecture that uses a secure .NET API, the PnP Core SDK, and SharePoint to record drafts and manage communication workflows within user permission boundaries. Consequently, the demo offers a concrete example of moving Copilot from passive suggestion to active process participant.

Overview of the Demo

The recorded community call walks viewers through a fully functioning scenario where staff use an AI assistant to draft, store, and manage communications. Paul Bullock demonstrates how the declarative agent is configured with JSON to define knowledge, actions, and responses, while API plugins extend its ability to act on behalf of users. Furthermore, the demo emphasizes delegated authentication and single sign-on so that actions respect each user’s permissions and audit trails. Thus, the session serves as both a technical walkthrough and a practical template for teams.

How the Communication Buddy Works

First, the system defines the assistant’s behavior through declarative instructions rather than custom imperative code, which simplifies changes and governance. Then, the agent uses API plugins to connect to enterprise endpoints so it can read from, and write to, data sources such as SharePoint lists and related services. In addition, the video shows a .NET Web API layer built with the PnP Core SDK that acts as a controlled bridge between Copilot and enterprise data, enabling actions while enforcing security constraints.

Moreover, the demo highlights the flow of delegated permissions: the user authenticates via SSO, the agent operates within that user context, and the API enforces policy checks before committing changes. As a result, the Communication Buddy can save drafts, update records, or trigger workflows without creating broad privileged service accounts. This design balances automation and control, allowing the assistant to be useful while limiting unintended access to sensitive systems.

Benefits and Tradeoffs

Using a declarative agent with API plugins offers clear benefits: faster deployment, easier maintenance, and assistants that understand organizational context and style. For example, staff can produce higher-quality messages and reduce back-and-forth reviews, which increases productivity and helps standardize communications. However, tradeoffs arise because integrating live actions into business systems increases surface area for errors and latency, and demands stronger operational monitoring and logging. Therefore, teams must weigh speed and convenience against the need for robust testing and safeguards.

Additionally, the reduced need for deep coding through declarative definitions speeds adoption but can obscure complex logic that would otherwise be explicit in code. Consequently, debugging and auditing agent behavior sometimes become more difficult unless organizations adopt disciplined change controls and observability tools. In practice, the balance between simplicity and transparency will depend on how critical the automated actions are to business outcomes and compliance requirements.

Security, Permissions and Compliance

Security is central to the demo’s design, and the presenter stresses that the system relies on Microsoft identity models, delegated permissions, and SSO to preserve user-context and accountability. Moreover, the .NET API layer enforces policy checks and ensures that only authorized operations proceed, which helps align the assistant with organizational governance and auditing needs. At the same time, enterprise teams must implement monitoring and role-based access controls to detect anomalous behavior and prevent privilege escalation. Consequently, this architecture reduces risk but does not eliminate the need for rigorous compliance practices.

Challenges and Practical Considerations

Implementing a Communication Buddy involves technical and human challenges: developers must handle error handling, rate limits, and data schema evolution, while business teams must train users to trust and correctly use the assistant. Furthermore, organizations must manage the lifecycle of declarative instructions, plugin configurations, and API endpoints to avoid drift between the assistant’s behavior and company policies. In addition, performance tradeoffs may appear when agents call external systems in real time, so teams should consider caching, batching, or asynchronous workflows to preserve responsiveness.

Implications for Organizations

Overall, the YouTube demo provides a practical roadmap for organizations that want Copilot to do more than generate text: it can act, store, and orchestrate work while respecting permissions and governance. Moving forward, leaders must balance the operational overhead of secure integrations with the clear productivity gains that such assistants deliver, and they should plan for testing, observability, and user adoption workstreams. Finally, the sample shown by Microsoft and Paul Bullock offers a replicable pattern that teams can adapt, and it signals a broader shift toward AI assistants that are both conversational and operational.

Keywords

Microsoft 365 Copilot Declarative Agent, Copilot API plugins, Communication buddy with Copilot, Build Copilot declarative agent, Copilot plugin integration tutorial, Enterprise communication bot Microsoft 365, Low-code Copilot agent, Integrate APIs with Copilot plugins