Customer data, orders, documents, inventory levels, and financial reports are already stored in the ERP system. The challenge is that finding the right information often still requires navigating between records, filters, and multiple screens. Business Central MCP Server makes it possible to connect the system with AI models such as Claude, ChatGPT, GitHub Copilot, and solutions built in Copilot Studio, enabling users to work with data and processes through natural-language commands.
The Data Is in the System. Why Does Finding It Still Take So Much Time?
An ERP system organizes the most important information and processes within a company. Microsoft Dynamics 365 Business Central users can manage finances, sales, purchasing, inventory, warehousing, and order fulfilment within a single solution.
However, simply collecting data does not solve every problem. Employees still need to know:
- where to find a specific piece of information,
- which filters to apply,
- which records and documents need to be connected,
- how to move from finding the relevant data to performing the appropriate action.
In practice, even a simple question may require opening several screens. Do we have enough of a particular product in stock? When can we fulfil an order? Which invoices are still awaiting processing? Have all purchase orders been registered correctly?
An AI agent can shorten the path between the user’s question and the answer stored in the system.
Can Claude, ChatGPT, and Copilot Work with Business Central?
Yes. Microsoft’s official documentation lists the following supported MCP hosts:
- Visual Studio Code with GitHub Copilot,
- Microsoft Copilot Studio,
- MCP-compatible clients such as Claude and ChatGPT,
- custom solutions developed by an organization.
For Claude, ChatGPT, and other clients that do not provide built-in Business Central support, the connection must be configured manually. This includes registering an application in Microsoft Entra ID, authenticating the user, and specifying the relevant environment, company, and MCP Server configuration.
This means that simply opening any version of an AI application will not provide immediate access to the company’s ERP system. The connection must be properly implemented and secured.
It is also worth clarifying that the term “Copilot” can refer to several different Microsoft products. In the context of Business Central MCP Server, the documentation primarily refers to GitHub Copilot and agents created in Copilot Studio.
What Can an AI Agent Do in Business Central?
The scope of its capabilities depends on the MCP Server configuration, the API pages made available, and the permissions granted. An agent can not only answer questions but, when properly configured, also create, update, and delete records or perform business actions such as changing a status or posting a document.
Finance and Reporting
An agent can retrieve financial data, combine information from different records, and prepare summaries. For example, it can assist with analysing balances, identifying missing data, or compiling a report for a specific period.
However, it does not replace the accounting logic built into Business Central. Calculations, validations, and business rules remain within the ERP system.
Sales and Customer Service
An employee can ask the agent to check product availability, the expected fulfilment date, or the status of a customer order.
With the appropriate permissions, the agent can also support the creation and updating of records, such as customer cards or sales orders. Microsoft lists item creation and sales order processing as examples of operations that can be carried out using GitHub Copilot connected to Business Central MCP Server.
Purchasing
An agent can help find purchase orders, check supplier information, analyse delivery dates, or identify documents that require further action.
Instead of manually opening multiple records, the user can give the agent a single command covering several related questions.
Warehousing and Inventory Management
Access to current inventory data allows the agent to answer questions about product quantities, availability, and location. It can also support data comparisons and help identify potential discrepancies.
The range of available operations will depend on the APIs and processes configured in the specific Business Central environment.
Document Workflows
An agent can search for documents based on specific criteria, analyse their status, and identify items that require attention. When the solution is appropriately extended, it can also support processes related to document approval, registration, and further processing.
Data Analysis
One of the greatest advantages of AI agents is the ability to ask questions in natural language. Users do not have to begin by building a report. They can first request an answer, a summary, or the identification of unusual values, and then explore the results further by asking follow-up questions.
Is This Type of Connection Secure?
An agent does not automatically receive unrestricted access to the entire system.
By default, Business Central MCP Server only allows agents to read data made available through API pages. Operations that enable the creation, modification, or deletion of records, as well as the execution of specific actions, must be configured by an administrator.
This means that the agent’s capabilities depend on two key elements:
- the data and APIs made available in Business Central,
- the permissions assigned to the user and to individual operations.
If the agent does not have permission to create a customer or perform another action, the operation will not be completed. Microsoft emphasises that an agent can only use the data and functions permitted within the configuration of a specific environment.
The extended version of MCP Server also uses Microsoft Entra ID authentication, as well as telemetry and auditing mechanisms. These allow administrators to monitor configuration changes and track whether MCP calls succeed or fail.
Therefore, instead of saying that “the agent cannot do anything without the user’s approval each time,” it is more accurate to say:
The agent operates exclusively within the scope of the data made available, the configured operations, and the permissions granted.
Will an AI Agent Replace the ERP System?
No. An agent does not replace Business Central or become the company’s new source of data.
Business Central still:
- stores information,
- applies business rules,
- validates operations,
- controls permissions,
- records documents and transactions.
The agent acts as an additional communication layer. It helps users find data more quickly, understand it, and initiate an authorised process.
In other words, an agent does not replace the ERP system — it helps users take advantage of its capabilities more efficiently.
What Can a Company Gain from AI Agents?
The greatest value does not come from the ability to communicate with the system itself. It comes from shortening the path between asking a question and completing a task.
A well-designed agent can reduce the time spent on:
- searching for data across different areas of the system,
- repeatedly performing the same operations,
- manually preparing simple summaries,
- checking the status of documents and orders,
- navigating between multiple records.
However, not every process should be delegated to an agent immediately. Implementation should begin with repetitive scenarios in which users frequently search for the same information or perform a similar sequence of actions.
How Can Business Central Be Prepared to Work with AI Agents?
The first step is to define a specific use case. Simply connecting another AI tool will not deliver value unless it is clear which problem the tool is intended to solve.
The next steps include:
- identifying the data and processes required by the agent,
- verifying the availability of the relevant APIs,
- configuring MCP Server,
- defining the permitted operations,
- assigning the appropriate permissions,
- connecting the selected AI client,
- testing the solution in a controlled scenario.
Business Central MCP Server is designed for Business Central online environments. Its configuration and implementation scope should be adapted to the system architecture, extensions, and security policies of the specific organization.
Less Clicking, More Action
MCP Server introduces a new way of working with Business Central. Users no longer need to start every task by searching for the right screen in the system. Instead, they can begin with a question or command, and the agent will help them access the relevant data and — provided it has the necessary permissions — carry out a specific action.
This does not mean full autonomy or unrestricted AI access to company data. The organisation decides which capabilities are made available and to what extent the agent can use them.
Would you like to find out which Business Central processes could be streamlined with AI agents? Contact us. We will help you assess your system’s capabilities, define the right use case, and configure the integration securely.
📞 Tel. (+48) 12 378 78 00
✉️ office@itvision.pl
CEO of IT Vision, with over 25 years of experience in the IT and business consulting sector; an expert in ERP systems, process automation, and AI-driven solutions, delivering implementation and advisory projects in Poland and international markets.
