TL;DR

Microsoft’s February 2026 update to Microsoft Dynamics 365 Business Central introduces custom AI agents that can be created and run directly inside the ERP system. Users can configure built-in templates, build their own agents with structured instructions, manage permissions, and execute tasks such as inventory replenishment. Copilot billing and Azure setup are required, but no Copilot Studio or MCP server is needed. This release provides a practical introduction to agentic AI within Business Central.

Custom AI Agents in Microsoft Dynamics 365 Business Central

Microsoft recently introduced new functionality in Microsoft Dynamics 365 Business Central (February 2026 update) that allows users to design and run their own AI agents directly inside the ERP system. This marks a major step forward in bringing agentic AI into day-to-day business processes without requiring external tools or complex integrations.

This walkthrough demonstrates how to review Microsoft’s built-in agent template, create a custom agent from scratch, and configure the necessary Copilot billing to make everything work.

Reviewing the Built-In Sales Validation Agent

Before creating a new agent, it helps to understand how Microsoft’s provided template works.

Accessing the Custom Agent Designer

Within Business Central, you now have the ability to create your own agents. Selecting Create opens available templates, including the Sales Validation template. This template performs structured validation tasks against sales data.

Configuring the Agent

The setup process is straightforward:

• Review the agent summary describing what it will do
• Assign a profile already available in the system
• Review and confirm permissions (standard Business Central permissions apply)
• Edit the instructions the agent will follow

The instructions clearly define the agent’s role and expected outcomes. When activated, the agent logs all of its activity within Business Central, providing full traceability.

Running the Sales Validation Agent

After activation, the agent can be run immediately. In this example, it was executed in a sandbox environment using standard demo data.

The agent followed its instructions and generated a report as the final step. After reviewing the output, the task was completed. This demonstrates how built-in custom agents operate and how simple it is to test them.

Creating a Custom Inventory Replenishment Agent

To better understand the functionality, a new custom agent was created from scratch. The purpose was to explore how agentic AI can solve practical ERP challenges.

Defining the Business Problem

This custom agent reviews inventory item by item. It checks the Projected Available Balance for each item. If the quantity is negative, the agent creates transactions to bring the balance back to zero.

This acts as a basic replenishment function.

For example, if a conference table item shows a negative projected available balance, the agent identifies the shortage and generates transaction lines in the requisition worksheet to correct it.

Setting Up the Agent Profile and Permissions

The setup includes:

• Assigning a user profile
• Configuring permissions
• Defining the agent’s description

In a sandbox environment, permissions may not be critical. However, in production, permissions must be reviewed carefully to ensure proper governance and control.

Writing the Agent Instructions

The instruction section is where the intelligence is defined.

The structure includes:

1. An introduction that defines the agent’s role
2. Step-by-step instructions
3. The expected output

In this case, the agent:

• Reviews projected available balances
• Identifies negative quantities
• Calculates required adjustments
• Creates requisition worksheet lines

Ideally, this agent would run automatically on a recurring schedule, such as hourly. However, scheduled execution is not yet available for custom agents in Business Central.

Running the Custom Agent

Executing the agent is simple:

1. Review or edit instructions
2. Start a new task
3. Optionally include additional instructions
4. Run the task

The system displays the task in progress. Once completed, the requisition worksheet is updated and ready for processing.

This confirms how easily Business Central can support practical agent-based workflows.

Setting Up Copilot Billing

To run custom agents, Copilot billing must be configured.

Microsoft provides a limited number of Copilot credits to get started. In this example, the agent ran several times before the credits were depleted.

Step 1: Create an Azure Environment

An Azure environment must be set up if one does not already exist.

Step 2: Configure Billing in the Power Platform Admin Center

In the Power Platform Admin Center:

• Navigate to Licensing
• Create a billing plan
• Connect it to your Microsoft subscription

Step 3: Link Billing to Business Central

Within the Business Central Admin Center:

• Link the Azure environment
• Assign the billing policy to your Business Central environment

Once connected, agents can run using available Copilot credits.

Why This Matters for Agentic AI in ERP

This update, introduced in version 27.4 of Microsoft Dynamics 365 Business Central, provides hands-on exposure to agentic AI directly within the ERP system.

Notably:

• You do not need Copilot Studio
• You do not need to configure an MCP server
• The functionality is built directly into Business Central

This lowers the barrier to experimenting with AI-driven automation inside ERP.

As Microsoft continues developing this capability, features such as event-based triggers and scheduled execution are likely to expand the power of custom agents even further.

If you are exploring Copilot deployment within Microsoft 365, reviewing this functionality in a sandbox environment is an excellent starting point.