Build Copilot extensions with Copilot Studio
summary
Demonstrates building a working Copilot extension end-to-end in roughly 90 seconds using Copilot Studio, pulling fields from Dataverse tables (account name, revenue, credit limit), turning the prompt into SQL, executing it, then feeding results into an LLM (3.5 or 4) to generate the reply — the demo works and exposes the source CRM record. Publishes to Teams app catalog, supports per-user auth, has developer debug mode and cache invalidation, and connects to 1,000+ connectors.
AI-written summary · about
chapters (16)
- 0:00Introduction and agenda
- 2:01Extending Microsoft Copilot
- 3:44Knowledge and action vectors
- 6:32Extensibility capabilities
- 7:31Inside a Copilot extension
- 10:37Copilot Studio and Azure AI
- 12:14The evolution of extensions
- 14:48Partner ecosystem examples
- 16:20New connector capabilities
- 17:15Customer implementation
- 18:38Live demo: building a prompt
- 23:47Live demo: Copilot testing
- 28:55Publishing architecture
- 31:52Security and governance
- 35:45Partner certification
- 37:47REST API plugins and wrap-up
related talks
49mBuild your custom Copilot with Microsoft Copilot Studio
19mMicrosoft Copilot Demystified: A 20-Minute Introduction
43mHarness the power of AI: Extend Copilot and beyond
21mExpanding Your Airspace | Extend your Copilot with Plugins, Work IQ Dev Tools and more!
49mThe future of app development with the Microsoft Power Platform
31mFrom Build to Operate: ALM Patterns for Production‑Ready Copilot Studio Agents
47mGitHub Copilot for SQL Development: Integrating the Power of AI into Database Development
53mKeynote - Business Apps and Agents