topic hubCopilot & Agentsofficial - Microsoftreviewed - checked by Opus

Extend Copilot in Business Central

Extending Copilot in Business Central with AL: building a Copilot capability with the System.AI module and Azure OpenAI, designing the user experience with the PromptDialog page type, and testing capabilities and agents. It answers how-to questions for AL developers.

Path: Integration with AI > Integrate AI using developer tools for Copilot > Extend Copilot in Business Central · tier official · system copilot · narrative reviewed (checked by Opus)

Overview

This section is for AL developers who add generative AI to Business Central extensions. It starts with an introduction to the building blocks: a prompt dialog page, the AI module of the system application, and an Azure OpenAI resource.

The page “Build the Copilot capability in AL” covers the code side: registering the capability, storing authorization, configuring parameters, defining a metaprompt and generating responses. Two subtopics cover the rest. “Build Copilot user experience” explains the PromptDialog page type and its prompt, generate and content modes. “Test Copilot capability in AL” covers the Evaluation framework and how to write AI tests and agent tests.

Start with the introduction, then build the capability, then design the dialog, and finally set up tests.

Key points

  • A Copilot experience needs three parts: a prompt dialog page, the AI module of the system application, and an Azure OpenAI resource.
  • Building the capability covers capability registration, authorization storage in IsolatedStorage, parameter configuration, metaprompt definition and response generation.
  • The capability page covers chat completion, text completion and embeddings through the System.AI module, plus the token limits to keep in mind. It references 2025 release wave 2.
  • The PromptDialog page type has prompt, generate and content modes, and the pages cover error handling and launching Copilot features from prompt actions.
  • Testing uses the Evaluation framework (formerly AI Test Toolkit) with JSONL and YAML datasets.
  • Test pages cover AI tests and agent tests, plus best practices for non-determinism, safety, languages and model changes.

Subtopics

More Learn pages

Videos and posts

Linked by a Haiku matcher with a grounding quote from the item’s summary (link/topics.ts); machine-generated.

Source: Microsoft Learn (CC BY 4.0). Descriptions are Learn’s own.

What Learn documents here 8 of 15 Learn pages

Show 7 more