Since Microsoft first introduced the Model Context Protocol (MCP) at Microsoft Build in May 2025, the vision for an “Autonomous Enterprise” has moved at a lightning pace. What started as a static set of tools has evolved into a dynamic gateway, and with the General Availability of the ERP MCP server in February 2026, the era of Agentic AI, artificial intelligence that doesn’t just suggest steps but performs them, is officially here.
As we look toward the upcoming March 2026 release of version 10.0.47, our technical Dynamics 365 expert, Daniel Kirk, explored how this protocol is transforming Dynamics 365 Finance and Supply Chain Management from a traditional system into a proactive partner. You can watch the video or read the updates below.
Dynamics 365 ERP MCP Server enables move from AI Chatbot to “Agentic” AI
The jump from a standard chatbot to an “Agentic” AI lies in the architecture.
The Model Context Protocol (MCP) acts as a standard language between the AI (the “Model”) and your business applications, like Dynamics 365 Finance or Supply Chain Management.
While standard AI is limited to its training data, an AI connected via an MCP server has secure, real time access to your specific Dynamics 365 environment. This allows the AI to read your data, understand your unique business logic, and interact with the system on your behalf. It serves as a centralised integration hub that is far more efficient than building individual, process specific tools.
Three Ways AI Can Now Interact with Your Dynamics 365 Data
The Dyanmics 365 ERP MCP server organises the AI’s capabilities into three distinct toolkits, allowing it to navigate the system much like a human user would.
- Form Tools
These allow the AI to “drive” the Dynamics 365 user interface. It can open specific forms, add filters to grids, sort columns, and even click buttons or save records. - Data Tools
These provide direct access to your Dynamics 365 database records through data entities. The AI can retrieve available entities (including custom extensions), read specific records, and perform standard create, update, or delete operations. - Action Tools
These allow the AI to execute complex business logic within Dynamics 365. For example, if you ask it to “post a journal,” the AI analyses the available actions, identifies the correct parameters, and executes the process in Dynamics 365.
What could Dynamics 365 AI Agents in Action look like?
We are already seeing specialised agents deployed within Dynamics 365 to handle time consuming financial and operational tasks. Here are some Real-World Use Cases.
- Supplier Communication Agent in Dynamics 365
This agent can autonomously receive emails from vendors, process the inquiry, and update the corresponding purchase order inside Dynamics 365. - Reconciliation Agents in Dynamics 365
Dedicated agents now exist for identifying differences between sub-ledger and general ledger transactions, as well as an agent that can compare and reconcile two datasets within Microsoft Excel. - Data Quality Agent in Dynamics 365
This tool reviews your transactions and settings to suggest improvements, for instance, noticing when a product’s configured lead time doesn’t match the reality of your data.
The journey from Manual Entry to Strategic Oversight in Dynamics 365 thanks to the new ERP MCP server
The move toward an MCP based architecture marks the transition of Dynamics 365 from a passive database into an active participant. For decades, using an ERP meant a high volume of “manual clicks”. With the new ERP MCP server, the AI handles the heavy lifting of navigation and data manipulation.
This has the potential to create massive productivity gains; instead of spending hours on data entry, your team moves into a role of strategic oversight. You become the “agent manager,” reviewing high level outcomes and anomalies, allowing you to focus on growth and innovation rather than clerical maintenance.
The wild thing about the Dynamics 365 ERP MCP architecture is that we are only at the beginning
As Microsoft continues to expand the capabilities announced last year and finalised in the February 2026 general release, the opportunity for businesses to innovate is truly limitless.
The MCP architecture finally bridges the gap between human language and technical ERP logic. By embracing these Dynamics 365 “Agentic” tools today, your organisation isn’t just keeping up with technology, you are building a faster, smarter, and more resilient future.
Don’t know where to start or where “Agentic” AI fits into your Dynamics 365 workflows or processes?
While the potential of “Agentic” AI is vast, it can feel overwhelming. At Walkerscott, we specialise in helping organisations navigate these new technical frontiers, from configuring your first AI agents to fine tuning the data entities they interact with. Whether you are looking to automate vendor communications or streamline complex financial reconciliations, we are here to help you turn this new technology into a structured, high performing reality. Get in touch with us.
Dynamics 365 ERP MCP Server & AI Agents FAQs
Can AI "hallucinate" or make mistakes when it creates data in Dynamics 365?
AI can sometimes make mistakes, and its decisions are not “deterministic”, meaning it might follow a slightly different path or provide a different result each time. This is why it is essential to test these features in a sandbox Dynamics 365 environment and have human users check the results before finalising records.
How do I make sure Dynamics 365 AI agents don't see sensitive data?
Your organisation has full control over whether your internal Copilot has access to Finance and Operations data. Access is typically governed by security settings and roles to ensure that data remains private and is only accessible to authorised users.
Do I have to use the Copilot "sidecar", or can I use Dynamics 365 AI agents elsewhere?
While the sidecar is the most common interface, you can also publish Copilot agents to other applications like Microsoft Teams. This is particularly convenient for team members who don’t regularly log into the full Dynamics 365 Finance and Operations application but still need to query data or trigger processes.
Can the Dynamics 365 AI agents handle my specific custom fields and entities?
Yes. One of the primary functions of the Data Tools in the Dynamics 365 ERP MCP server is to retrieve the metadata of your system, which includes extensions and any custom data entities you have added. This allows the agentic AI to “read and write” to your unique customisations just as easily as it does to standard fields.
If my Dynamics 365 AI agent doesn't have enough information to complete an action, what happens?
If you give an instruction that is missing a required parameter (like a warehouse location or a specific vendor ID), the Dynamics 365 AI won’t just fail; it will ask you for that missing information so it can complete the task correctly.