MCP Enablement takes AI automation in Dynamics 365 to a whole new level of scale
The foundation of a modern ERP is, today, one of the best places to see the real impact of Artificial Intelligence on businesses. And the combination of Dynamics 365 and Model Context Protocol (MCP) is one of the clearest examples of that.
During the AI Agent & Copilot Summit, Microsoft shared a figure that really shows the scale of this transformation: there are already 650,000 ERP actions available to Dynamics 365 customers, directly accessible by AI agents. In practice, this means those agents can execute nearly everything a human user would do inside the ERP — only in an automated way, following the same security and governance rules.
As Sachin Gandhi, Principal Solution R&D Architect at Microsoft, put it during his session at the event, everything you can do manually when you log into the system — create a sales order, register a purchase order, post a journal entry, work in the general ledger — can also be done by an AI agent, because MCP is enabled in the ERP and respects the same access profiles and controls as Dynamics 365.
How MCP changes the game inside the ERP
The Model Context Protocol (MCP) has become the key piece for connecting AI agents to corporate data in a broad, secure, and standardized way. Instead of relying on specific integrations for each application or database, MCP offers a consistent way to expose functionalities and information from systems like Dynamics 365 to language models and agents.
In a traditional scenario, if someone wanted to answer a question like:
What were the sales for a specific item category, for a certain customer group, over a given time period?
that would probably require:
- involving a developer or BI analyst,
- defining a custom data model,
- setting up reports or dashboards,
- waiting through the development, testing, and publishing cycle.
With an ERP MCP server, the workflow changes completely. An AI agent can receive that same request in natural language, interpret the ask, apply the correct business logic (including customer-specific customizations), and access the application the same way a human would — only automated and much faster.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
MCP, in this context, acts as a layer that:
- exposes ERP actions in a structured way for agents;
- respects credentials and permissions already in place within Dynamics 365;
- allows the agent to navigate through forms, entities, and workflows in the ERP as if it were using the human interface;
- maintains governance, audit, and security controls.
The result is a level of contextual automation that doesn’t require rewriting the ERP or building massive integrations. What’s already configured becomes direct input for the agents.
Practical example: from purchase requisition to purchase order
To show how this works in practice, Microsoft walked through a complete flow during Gandhi’s session, using a classic procurement example: selecting a vendor for a requisition and following through to creating the purchase order.
In this scenario, an MCP agent was able to carry out a complex sequence of actions across different modules and applications, including:
- identifying the best negotiated price for the requested item;
- collecting vendor performance data, such as delivery history and quality;
- reasoning over the available data to choose the most suitable vendor;
- locating the requisition record and accepting the associated workflow task;
- updating the requisition with the selected vendor and completing the workflow step;
- documenting the decision logic, enabling auditing of why that vendor was chosen;
- sending an email to the human approver with a summary of the requisition and the rationale for the selection;
- releasing the requisition and automatically generating the purchase order in the system.
That’s a set of steps that would normally require human interaction across multiple ERP modules, manual verification of information, and coordination between departments. With MCP, all of this happens in an autonomous, traceable way that’s aligned with company rules.
AI agents operating as power users of the ERP
An important point in Microsoft’s approach is that agents aren’t bypassing the ERP. They don’t operate in parallel — they use the same mechanisms a power user would when interacting with the system.
This means that:
- the business rules configured in Dynamics 365 remain in effect;
- role-based security continues to be enforced;
- approval flows and workflows run as they were designed;
- any customization of screens, fields, or processes is respected by the agent.
In practice, MCP doesn’t ignore what’s been built over the years in the ERP. It transforms that body of rules, configurations, and data into a more accessible environment for agentic AI.
Real-world cases: MCP in daily operations
It’s not just Microsoft talking about concepts. Other companies are already using MCP integrated with their ERP to solve very concrete problems.
Marc Kase, CIO of Altman Plants, shared at the event how the company has been using an MCP server connected to its ERP to:
- resolve transaction errors more quickly;
- identify recurring errors and understand their root causes;
- analyze transaction volume by store and by customer in retail;
- answer recurring business questions directly from the ERP.
In this type of scenario, MCP acts as a kind of universal interface between business questions, transactional data, and ERP processes. Instead of always relying on pre-built reports or technical queries, the company can use agents to navigate what already exists and compose answers in near real time.
The MCP ecosystem in Dynamics 365
During the presentation, Gandhi also highlighted the expansion of the MCP server ecosystem within the Dynamics 365 universe. Among the examples mentioned were:
- Dataverse MCP — providing access to data stored in Dataverse with entity and relationship context;
- ERP Client MCP — focused on operational actions directly on the ERP;
- ERP Analytics MCP — geared toward data analysis and exploration for insights;
- Commerce MCP — built for commerce and customer experience scenarios.
Each of these servers extends the reach of agents across different types of data and processes. Together, they create a comprehensive MCP layer for the Dynamics ecosystem, paving the way for automations that cross boundaries between applications that were previously treated as silos.
Commerce MCP and agentic experiences in retail and beyond
One of the standout highlights was Commerce MCP, focused on agentic commerce — meaning experiences where AI agents work directly within shopping journeys, customer service, and commerce operations.
With this approach, Microsoft has been enabling scenarios such as:
- conversational shopping — where the customer interacts in natural language and an agent handles the entire flow of recommendations, cart, conditions, and checkout;
- intelligent customer service — agents that access order history, policy rules, and inventory data to resolve issues without escalating everything to humans;
- automated post-sale workflows — returns, exchanges, refunds, and status updates handled in near real time.
A core benefit of this Commerce MCP + agents combination is the reduced time to value. Customers can build agents specific to their processes with far less integration effort, because the data building blocks and actions are already exposed via MCP.
MCP as the standard for AI access to corporate data
The advancement of MCP in Dynamics 365 is yet another sign of something bigger: the protocol is solidifying its position as a standard for AI agent access to corporate data and actions. By functioning as a kind of universal interface between models and systems, it helps to:
- reduce the cost and time of integration;
- standardize how applications expose capabilities to agents;
- make AI adoption easier in complex, legacy-heavy environments;
- accelerate the automation of complex workflows that span multiple systems.
In the case of Dynamics 365, the message is clear: the more MCP spreads across the ecosystem, the more agents can operate end to end within business processes, without requiring huge middleware projects for every new automation initiative.
Impacts on operations and the ERP user experience
From the perspective of those who live in the ERP day to day, this shift changes the game quite a bit. Instead of interfaces packed with screens and fields that demand intensive training, the trend is toward more:
- natural language interactions with the ERP, mediated by agents;
- automated operational tasks, like postings, reconciliations, and status updates;
- more consistent processes, with less variation between users and fewer human errors;
- greater focus on oversight and exceptions, rather than the mechanical execution of steps.
At the same time, AI gains historical and contextual visibility into Dynamics 365 data, enabling agents to:
- detect risk patterns before they become problems;
- assist with operational forecasting, like demand spikes or supplier bottlenecks;
- suggest process adjustments based on recurring errors or rework.
It’s a gradual but deep shift in how teams relate to their ERP: less field-filling, more decision-making on top of information already digested by agents.
One more step in the AI automation journey
Gandhi’s update on Dynamics 365 and the growing support for MCP reinforces an important point: AI automation in enterprise environments is moving past the talking stage and into the heavy operational layer — where the processes that actually move the business live.
With 650,000 ERP actions exposed to agents and an MCP server ecosystem covering everything from transactional data to commerce, Dynamics 365 is positioning itself as a platform where an agent can do nearly everything a person does in the ERP — only at scale, with traceability, and within the rules of the business.
For anyone following the rise of AI agents in the enterprise world, this is yet another signal that MCP is establishing itself as the standard interface between models and systems, opening space for increasingly complex automations that remain grounded in the reality and governance of business operations.
