RSMUS.comMost organizations running Dynamics 365 Finance and Operations have already asked where AI could take real work off their plate. The obstacle is rarely the model; it’s the plumbing. Every agent that needs to read an order, update a record or trigger a process tends to arrive with its own custom API integration behind it, and that cost gets paid again with every new use case. Dynamics 365 Model Context Protocol (MCP) is valuable because it changes the pattern rather than adding another feature.
What the connection layer actually does
MCP acts as a common protocol between AI agents and F&O applications, spanning enterprise data, application functions, and business logic. In practice, it exposes three kinds of tools:
- Data tools let an agent create, read, update and delete records through data entities
- Form tools let an agent work with F&O forms through server APIs, including opening a form, filtering it, setting control values, saving and selecting from the available actions
- Action tools let developers expose custom business logic that agents can discover and invoke
The distinction that matters for a business audience is straightforward. This is not another question-and-answer layer sitting on top of ERP data. An agent working through these tools can operate inside a process rather than just describe it.
Why standardization is the real story
The value is less about any one tool and more about reducing what must be rebuilt for each use case. MCP does not eliminate integration, security design, or testing. It provides a reusable pattern for discovering and invoking ERP data and operations, allowing teams to spend more of the project on decisions, exceptions, approvals, and user experience.
The strongest first candidates are high-volume workflows with clear rules, reliable data, and exceptions that can be routed to a person. Sales-order exception handling, vendor-maintenance review, and approval preparation are better starting points than broad, open-ended automation.
Choose the data path deliberately
Data access is one of the first architecture choices to test, not an implementation detail to revisit after the pilot. The ERP MCP server supports SQL-backed querying alongside OData-based data access. Microsoft positions the SQL tool as a way to add operators, improve aggregation, and improve query quality and performance.
That improvement does not make performance automatic. Test the actual entities, filters, result sizes, security roles, and transaction volumes the agent will encounter in production. Set acceptance thresholds for response time, successful completion, and exception handling before expanding the workflow.
Treat identity and permissions as part of the workflow
An MCP connection does not create a new security boundary around Finance and Operations. Requests are authenticated, authorization is enforced by the application, and existing business logic, validations, and workflows continue to apply. The relevant question is therefore not simply what the server can expose. It is what the identity behind a particular agent is allowed to discover and execute.
For a first deployment, give the agent only the permissions required for the selected workflow. Separate read and write scenarios where possible, require human approval for material changes, and make each resulting transaction traceable to the initiating person or autonomous-agent identity.
How to approach adoption
Start with one workflow where action creates measurable value. Confirm that the required data, form, and action tools are supported, then define the identity the agent will use and the minimum permissions it needs. Test the workflow with realistic entities, filters, result sizes, transaction volumes, and exception conditions rather than a clean demonstration dataset.
Establish acceptance thresholds for response time, completion rate, incorrect actions, and human overrides. Before production approval, document what the agent can change, when a person must approve the action, how the resulting transaction will be traced, and how an incorrect action will be stopped or reversed.
Connect with us to learn more about building an ERP agent strategy that holds up in production.
Explore Microsoft Fabric Solutions
Mitchell Rose
Mitchell Rose is an Associate in AI Advisory & Solutions at RSM, helping organizations transform data into actionable insights and AI-powered business solutions. Mitchell focuses on designing and delivering solutions across artificial intelligence, Microsoft Fabric, Power BI, data engineering, and data modernization initiatives. He partners with clients to solve complex business challenges by combining modern analytics platforms, intelligent automation, and scalable cloud technologies that drive operational efficiency and decision making.