We spend a lot of our time in Dynamics 365 CRM watching the same problem repeat itself in different forms: an agent, a bot, or a Copilot integration has data access, but no idea how our organization actually wants a task done. Give an AI agent the right permissions and it can read and write records all day. But ask it to “log this call the way we normally log it” and it has no reference point. Every agent ends up guessing, and every guess looks a little different.

This is exactly the gap Business Skills in Dataverse are meant to close. Instead of teaching an agent our process every single time through a long prompt, we write the process once, save it inside Dataverse, and any connected agent can pick it up and follow it consistently. We picked one everyday scenario, a support call getting logged against a Case, and tested the whole thing end to end using an AI agent connected through the Dataverse MCP server. This blog walks through exactly what we did.

What Is a Business Skill?

A Business Skill is a plain-language set of instructions stored inside Dataverse. It describes how a specific piece of work should get done: the steps involved, the information needed, and any rules that apply. Think of it as a written SOP, except this one is readable by an AI agent, not just a human.

The skill has three parts:

• Description – a short line telling the agent when this skill applies. This is what the agent scans through to decide whether a skill is relevant to what the user just asked.
• Instructions – the actual step-by-step process, written in markdown.
• Resources (optional) – a supporting file, like a policy document, that the agent can refer to.

Once saved, any agent connected to that Dataverse environment can discover the skill and follow it, whether that agent is sitting in Copilot Studio, Microsoft 365 Copilot, or, in our case, a coding agent connected through the Dataverse MCP server. The MCP server is simply one of the doors an agent can walk through to reach these skills; it is not something the feature depends on.

Prerequisites

Before starting, make sure these are in place:

• Managed Environment – Business Skills only work inside a Managed Environment. Power Platform Admin Center → Environments → select your environment → enable Managed Environment.
• Dataverse intelligence turned on – Power Platform Admin Center → your environment → Settings → Product → Features → Dataverse intelligence → turn on “Turn on Dataverse intelligence for agents and AI experiences.”
• Dataverse MCP server enabled – same Features page → Dataverse Model Context Protocol → allow MCP clients to interact with the Dataverse MCP server.
• A connected agent client – we used VS Code with the GitHub Copilot extension in Agent mode. If you plan to use a client other than Copilot Studio, remember to also allow that specific client under Advanced Settings → Allowed MCP Clients. As shown in the screenshot below

• Node.js installed, since VS Code’s MCP connection runs through it.

The Scenario We Tested

A support rep gets off a call with a customer. Instead of opening the Case record, scrolling to the timeline, and typing out a note by hand, we wanted an AI agent to do this for them: take the case number and a quick summary of the call, find the Case, and log it as a note automatically, in a consistent format every time.

Step 1: Create the Business Skill

Go to make.powerapps.com → your environment → More → Business skills (preview).

Click Create skill. Give it a name, for example LogCallTranscript. Power Apps generates a unique name with an auto-assigned prefix. If that prefix does not read well, you can edit the unique name field directly before saving

In the Description field, write a short line describing when the agent should use this skill and add a few trigger phrases:

Use this skill when the user wants to log a call transcript or conversation summary against a Case (Incident) record in Dataverse. Trigger phrases: “log this call”, “log this transcript”, “save this conversation to the case”.

In Instructions, write out the actual process. This is the part the agent will follow step by step, so it needs to be specific about which tool to call and what data to use, not just a general description:

## Purpose
Log a customer call transcript as a note on the related Case record.

## Steps
1. Ask the user for the Case number (ticketnumber) if not provided.
2. Call the Dataverse MCP `read_query` tool to run:
SELECT incidentid
FROM incident WHERE ticketnumber = ‘<case number>’.
3. If no case is found, ask the user to confirm the case number and retry.
4. Call the Dataverse MCP `create_record` tool to create a new “annotation”
record with:
– objectid: the incidentid GUID from step 2
– objecttypecode: “incident”
– subject: “Call Transcript – <today’s date>”
– notetext: the full transcript/summary provided by the user
5. After the create_record call succeeds, confirm to the user that the transcript was logged, and share the Case number.

## Example
Input: “Log this call for case CAS-01023-X4V2: Customer asked about refund status, informed 5-7 business days.”
Output: Note created on CAS-01023-X4V2 confirming refund status was discussed.

## Troubleshooting
– If case not found: verify ticket number format (CAS-XXXXX-XXXX), retry.
– If multiple matches: ask user to clarify.

Save the skill. We left the Resources section empty since our process did not need a supporting document. For visibility, we chose Individual, meaning only we could see and use this skill. Choosing Organization instead would make the skill visible to every maker and agent in that environment.

Step 2: Connect an Agent to the Dataverse MCP Server

We used VS Code with GitHub Copilot’s Agent mode for this. From your environment, grab the instance URL (Settings → Session details → Instance URL).

In VS Code, open the Command Palette and run MCP: Add Server, choose HTTP/SSE, and paste your instance URL with /api/mcp appended:

https://<yourorg>.crm.dynamics.com/api/mcp 

Open Copilot Chat, switch to Agent mode, and sign in when prompted.

Step 3: Ask the Agent to Find the Skill

Before testing the actual scenario, we confirmed the agent could see the skill at all:

Show me all business skills in this environment

Step 4: Run the Scenario

With a real case number pulled from the environment, we asked:

Find and follow the LogCallTranscript business skill to log this call for case <case number>: Customer asked about refund status, informed 5-7 business days.

The agent located the skill, ran a query to find the Case, created a note against it, and confirmed back with the case number.

The Result:

Opening the Case record and checking the Timeline, the note showed up exactly as the skill described, with the subject formatted as “Call Transcript – <date>” and the summary in the note body.

No manual note-taking, and every future call gets logged the same way, regardless of which agent or which person triggers it, because the process itself lives in Dataverse and not in someone’s head.

Conclusion:

What stood out to us through this exercise is that Business Skills are not tied to any one type of agent. We tested it through the Dataverse MCP server because that gave us the most direct visibility into which tool the agent called and when, but the same skill would work just as well if it were picked up by a Copilot Studio agent or Microsoft 365 Copilot, since the skill lives inside Dataverse itself. Once the process is written down properly, any agent connected to that environment inherits the same behavior. For any team looking to bring AI agents into their day-to-day Dynamics 365 workflows without every agent behaving differently, this is worth setting up early.

FAQs

What are Business Skills in Dynamics 365 Dataverse?

Business Skills are reusable, natural language instructions stored in Microsoft Dataverse that define how AI agents should perform specific business tasks. Instead of relying on lengthy prompts, organizations can create standardized workflows that agents such as Microsoft 365 Copilot, Copilot Studio, or other MCP-compatible AI clients can discover and execute consistently.

Why should organizations use Business Skills in Dynamics 365?

Business Skills help organizations standardize how AI agents perform repetitive tasks. Instead of every agent interpreting instructions differently, they all follow the same documented workflow, improving consistency, reducing manual effort, and ensuring business processes remain aligned across Dynamics 365 applications.

How can businesses implement Business Skills in Dynamics 365?
Organizations can create Business Skills directly in Power Apps and make them available to connected AI agents. If you need guidance on implementing Business Skills, Dataverse MCP Server, or AI-powered Dynamics 365 solutions, Inogic helps organizations design and deploy intelligent business automation tailored to their workflows.

How do Business Skills work with the Dataverse MCP Server?

The Dataverse MCP Server enables AI agents to discover Business Skills, retrieve Dataverse data, and perform actions such as creating or updating records by following predefined business instructions instead of relying only on prompts.

Can organizations customize Business Skills for their own processes?
Yes. Every Business Skill can be tailored to match an organization’s internal SOPs, approval rules, and business requirements. Inogic helps businesses design customized Dynamics 365 and AI solutions that align with their unique operational workflows.

Can a Microsoft Dynamics 365 partner help implement Business Skills?
Yes. An experienced Dynamics 365 partner can help identify suitable use cases, configure Dataverse, and create reusable Business Skills. Inogic assists organizations in implementing AI-powered Dynamics 365 solutions that improve productivity and governance.

The post Create and Use Business Skills in Dynamics 365 Using the Dataverse MCP Server first appeared on Microsoft Dynamics 365 CRM Tips and Tricks.

Today’s businesses rarely operate within a single application. Applications such as Sales Hub, which manage leads, opportunities, accounts and revenue to help sales teams close deals faster. On the other hand, Customer Service helps handle cases, tickets and customer support interactions to deliver faster issue resolution. Field Service is used to schedule technicians, manage work orders and track on-site services efficiently. Marketing Automates campaigns, emails, and customer journeys to generate and nurture leads. Project Operations plans, tracks, and bills projects by combining project management, resources, and financials in one system.

Here, each application works well individually, but together they often feel disconnected. Users need to switch between platforms, copy information manually, and spend more time navigating systems than actually solving problems. Even with AI copilots available, the experience doesn’t always improve because most copilots are still limited to one application at a time.

Key Takeaways

• Traditional copilots are limited to single applications
• MCP Server enables structured, cross-system interactions
• Copilot agents can execute workflows across Business Central, Dataverse, and SharePoint
• Reduces manual effort and system switching significantly

Limitations of Traditional Copilots:

Traditional copilots are helpful assistants, but their scope is short. They can answer questions, summarize information or perform small actions within the app they belong to. However, they struggle when there are tasks that require answers from multiple systems.

As a result, users still do most of the work manually. The copilot informs but it doesn’t truly execute.

What is MCP Server in Copilot? (Simple Explanation)

MCP (Model Context Protocol) Server provides a structured way for copilots to interact with multiple systems using predefined tools instead of direct API calls.

Copilot with MCP Server:

To make Copilot genuinely useful, it needs accessibility across multiple systems. This is where Microsoft Copilot, combined with the Model Context Protocol (MCP) comes into the picture.

MCP Server acts as a bridge between Copilot agents and enterprise tools. Instead of directly accessing databases or API’s in an unstructured way, the agent communicates through standardized tools exposed by the MCP server. These tools define exactly what the copilot needs to do. Whether it’s retrieving data or updating records.

MCP Server with different Systems:

With MCP Server, the copilot can interact with different platforms such as Microsoft Dynamics 365 Business Central, Microsoft Dataverse and Microsoft SharePoint.

Instead of treating each system separately, the copilot works across them as if they were one connected environment. For the user, it feels like interacting with a single intelligent assistant.

Key Criteria to Consider While Setting Up MCP:

Before building an MCP-powered agent, start with –

• Creating a Copilot agent from scratch by attaching the right tools. Each tool represents a specific capability, such as reading customer data, updating records or accessing documents.
• Clear descriptions and proper instructions ensure the agent understands how and when to use each tool. Once activated, the copilot can securely communicate with multiple systems.

Steps to Implement Copilot with MCP Server

Let us understand this by 3 real-life scenarios –

Scenario 1: Sales Deal Closure with Dynamics 365 Business Central

In this scenario, the copilot agent will check customer financial information from Business Central before closing a sales deal. It helps verify details like credit limits or financial exposure, so the user can decide whether the deal can be approved or not.

Step 1: Let us start by creating a blank Copilot agent named ‘Enterprise Unified Agent’. At this stage, the agent has no knowledge or access to any systems.

Step 2: A tool needs to be added that allows the agent to connect with ‘Microsoft Dynamics 365 Business Central’ through the MCP server.

Step 3: Inside Business Central, the MCP Server Configuration needs to be opened in order to connect Copilot with it.

Standard API’s are then exposed as tools so the copilot can access them.

After saving the configuration and refreshing the tools, they become available for the agent to use.

Step 4: Each tool needs to be provided with a clear ‘Description’ explaining its purpose. These descriptions help the copilot understand when and how each tool should be used during interactions with the user.

Step 5: Add ‘Instructions’ in the agent, which provides guidelines to the copilot on how to respond, what tasks to perform, and how to use the available tools during interactions. Once the configuration is complete, the agent can start interacting directly with Business Central data.

Step 6: Now the user needs to simply ask a question regarding sales deal closure, and the copilot retrieves the required information, such as customer details, along with the final decision as ‘Approved’ or ‘Not Approved’, without the need to manually navigate through the system.

Scenario 2: Technician Scheduling and Service Booking with Business Central and Microsoft Dataverse Server

In this scenario, the copilot helps manage service requests by checking asset details, verifying spare parts availability, and scheduling a technician. It collects issue details, date, and time, then books the service by using information from both Business Central and Dataverse.

Now we have to enhance the copilot by connecting it to multiple systems instead of just one.

Step 1: The second tool for ‘Microsoft Dynamics 365 Business Central’ needs to be configured using the same MCP setup process. This allows the agent to access additional data and perform more actions within the system.

Add ‘Description’ for the tool as per the requirement:

Step 2: An additional tool is needed to be added for ‘Microsoft Dataverse’ so the agent can retrieve records that are stored in it.

Add ‘Description’ for the tool as per the requirement:

Step 3: Add ‘Instructions’ to the agent as per the requirement:

Step 4: Now we need to make a connection between ‘Microsoft Business Central’ and ‘Microsoft Dataverse’ by entering the environment URL and enabling the data synchronization so the copilot can access updated information from both systems.

Step 5: Now we need to add Dataverse tables as knowledge sources, which will allow the Copilot to retrieve data from them whenever needed.

Necessary Tables are added as below:

Step 6: Now the user asks a question regarding ‘Technician booking’, so the copilot retrieves the required information by using both Microsoft Dataverse and Microsoft Dynamics 365 Business Central to process the request and provide the response.

Step 7: The agent interacts with the user to collect required details such as:

• Issue description
• Service date and time

As you can see above, the copilot asks for the required details like the ‘Issue & service time’, then checks the system and schedules the technician accordingly

Scenario 3: Retrieve Instructions from Documents with SharePoint

In this scenario, the copilot accesses documents stored in SharePoint to answer user questions. It reads policies, guidelines, or reference files and provides the required instructions so users can quickly find the information they need.

Step 1: To retrieve instructions, documents need to be added to Microsoft SharePoint.

Step 2: To make these documents accessible, Microsoft SharePoint needs to be added as a knowledge source for the Copilot, as shown below

We need to upload the necessary ‘Documents’ to Microsoft SharePoint as below

Step 3: ‘Descriptions’ and ‘Instructions’ need to be added to guide the agent on how and when it should use the document-based information during conversations.

Step 4: Now the user needs to ask the agent questions that require information from documents in SharePoint.

As you can see above, the copilot searches the documents and shows the required instructions step by step based on the question.

Conclusion:

By the end of these 3 scenarios, the copilot grows from a simple blank agent into a smart assistant that can connect to different systems and help complete tasks. Instead of opening multiple apps like Business Central, Dataverse, or SharePoint, users can just ask the copilot and get the information they need in one conversation. This shows how a basic copilot can gradually become a powerful tool that not only answers questions but also helps get real work done faster.

FAQs

1. What is MCP Server in Microsoft Copilot?

MCP (Model Context Protocol) Server enables Copilot to interact with multiple systems using predefined tools instead of direct API calls. It standardizes how Copilot accesses and executes actions across platforms.

2. Can Copilot work across multiple Dynamics 365 applications?

Yes. With MCP Server, Copilot can interact with Business Central, Dataverse, and SharePoint within a single workflow, enabling cross-system execution.

3. How is MCP different from traditional API integrations?

APIs require custom integration logic, while MCP exposes standardized tools with predefined schemas, making it easier for Copilot to discover and use capabilities dynamically.

4. What are MCP tools in Copilot?

MCP tools are predefined actions (like retrieving data or updating records) that Copilot can invoke. Each tool represents a specific business capability exposed by a system.

5. Is MCP Server required for building advanced Copilot agents?

For cross-system and action-oriented Copilot scenarios, MCP Server is essential as it enables structured, secure, and scalable integrations.

The post How to Build a Cross-System Copilot Agent in Dynamics 365 Using MCP Server first appeared on Microsoft Dynamics 365 CRM Tips and Tricks.