topic hubCopilot & Agentsofficial - Microsoftreviewed - checked by Opus

Integrate AI using developer tools for Copilot

Developer tools for Copilot in Business Central: what counts as a Copilot extension, the System.AI module and PromptDialog page type, Azure OpenAI setup and Business Central AI resources, Help preparation, and the transparency note. It answers planning, setup and design questions for partners building generative AI features in AL.

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

Overview

This section is for partners and AL developers who want to add generative AI to Business Central extensions. It starts with concepts: what the developer tools are (the System.AI namespace, the PromptDialog page type, partner key management, Power BI monitoring, responsible AI) and what is considered an extension to Copilot. An extension must be LLM-based, respond to natural language, and manage risk with human-in-the-loop patterns such as previews, confidence indication, and undo or discard.

The setup pages cover two routes to Azure OpenAI. You can get your own Azure subscription, access, permissions, resource, and deployment. Or you can use Business Central AI resources, which are Microsoft-managed, use GPT-4.1 and GPT-4.1-mini, and bill customers by consumption. Bring-your-own subscription is optional for custom scenarios. The pages also cover how to influence Copilot and agents without extending them, how to prepare app Help for Copilot (preview), and the transparency note.

Start with “What are the developer tools for Copilot?” and “What is considered an extension to Copilot?”. Then follow the setup pages for your authorization model. The subtopic “Extend Copilot in Business Central” (15 pages) has the AL how-to detail: building a capability, designing the PromptDialog experience, and testing capabilities and agents.

Key points

  • A Copilot extension must be LLM-based, respond to natural language to help users finish tasks, and manage generative AI risk with human-in-the-loop patterns (preview, confidence level, undo and discard).
  • The developer tools include the System.AI namespace, the PromptDialog page type, partner key management, Power BI monitoring, and responsible AI guidance.
  • Azure OpenAI prerequisites: an Azure subscription, verified service access, role-based access control permissions, a resource, a model deployment, and API key management.
  • Business Central AI resources offer Microsoft-managed Azure OpenAI with GPT-4.1 and GPT-4.1-mini, billed to customers by consumption.
  • Use the SetManagedResourceAuthorization method for managed resources and SetAuthorization for your own subscription. Handle errors with AOAI Operation Response.
  • Copilot and agents can use content from AL extensions without direct extensibility, for example in Analyze list, Autofill, Chat, Summarize, Sales Order Agent, and bank reconciliation. The page references 2025 wave 1.
  • App Help for Copilot (preview) needs public HTTPS English documentation, Bing indexing, the app.json Help property, and a favicon for brand visibility.
  • The transparency note covers text and chat completion APIs, prompt dialog UI, embeddings, feature governance, the feedback loop, and the admin screen for managing AI features.

Subtopics

More Learn pages

Videos, posts and code changes

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 24 Learn pages

Show 16 more