Back in February and March 2026, we published blogs about the Copilot AP Agent and Copilot Studio, respectively. This month we’ll discuss Agentic AI in Business Central. More precisely, we’ll provide insight on various strategies that can be leveraged to produce AI agents in Business Central.

Native Business Central Copilot Agent

The Business Central custom Agent Designer runs inside Business Central. It leverages Business Central’s security model, can execute multi-step workflows autonomously and can read or write Business Central data with full validation rules. If you need to keep track of things, it can log every action or trigger human approvals.

Such an agent can be used if you need to replicate the behaviour of a BC user, if you need strict compliance with BC logic, if you want autonomous processing or need the agent to run continuously in the background. In this scenario, you may need the Job Queue or specific event triggers.

Copilot Studio Agents connected to Business Central

Copilot Studio lets you build conversational agents that connect to BC using Business Central Connector available on the Power Platform or the Business Central Model Context Protocol Server.

The Power Platform Connector works well for low-code solutions and simple workflows where the agent needs to act as a user. This type of agent can create, read, update or delete data on various tables.

The MCP Server supports multi-step reasoning and can support complex workflows or custom .AL logic. It provides access to standard or custom API pages with more flexibility and data model awareness.

These agents can be published to Teams, websites, M365 Copilot, etc. They are good for scenarios that require flexibility and cross‑system workflows with limited programming. They can also be used as a conversational agent that interacts with BC. Although these agents do not have the depth or autonomy of native Business Central agents, they are good for customer or employee-facing chatbots.

Custom .AL Extensions with APIs & MCP Tools

If you need maximum control, custom business logic, and advanced AI orchestration, connect your agent to your own extension using the BC MCP Server or your own APIs.

This ensures full control over the agent’s behavior and the most flexibility and functional depth. This approach is needed if you need custom endpoints that are not available in the Connector.

A custom approach can also be used to provide a more user friendly or controlled output. As an example, an extension can present the results of a workflow or process in a page and thus make it easier to consume.

Moreover, using an extension called by Copilot instead of having the agent perform the task can provide faster results and be more cost effective over time as it doesn’t consume tokens. This applies when the required process is deterministic; in other words when an algorithm is sufficient to perform the task.

Please note that the MCP Server and Custom Copilot Studio for BC are in preview mode, only available in English in a sandbox environment at the moment.

A Custom Agent is developed using the AI development toolkit as per this article:

Designing and coding agents (preview) – Business Central | Microsoft Learn

With the BC Connector and the MCP Server, you can now interlace Business Central, power Automate and other applications using Copilot and thus improve your operations. All of this on the Microsoft Cloud where your data is secured.

Please don’t hesitate to write if you have questions or comments.

Cordially,

Patric and Sylvain.

 

Last updated: July, 2026