Home » Copilot Studio – managing with Git, keeping costs under control and Copilot Studio Kit

Copilot Studio – managing with Git, keeping costs under control and Copilot Studio Kit

In a world where efficiency, speed, and process automation are becoming key success factors, tools like Microsoft Copilot Studio are gaining importance. But to fully use their power, you need a deep understanding of available features and how to skillfully implement them within your organization. In this article, I will explore the most important aspects of working with Copilot Studio, showing you how a well-planned implementation can bring tangible business benefits.

Copilot Studio is a platform that lets you create intelligent agents. These agents can support human work, reduce customer service times, automate repetitive tasks, and open up new ways to interact with your company’s systems. We’ve previously covered the specifics of creating agents in a dedicated article. However, as part of the Power Platform, Copilot offers many additional features and capabilities that allow you to perfectly tailor the solution to your organization’s unique needs. Let’s take a closer look at what it can do.

How to estimate the costs of Copilot agents?

Copilot Studio opens up huge opportunities for companies in the field of process automation and customer service. However, every implementation involves an investment. That’s why it’s important to estimate your costs up-front, to understand how the new tool will affect your budget.

It is worth knowing that as part of the Dynamics 365 Sales and Dynamics 365 Customer Service licenses, we get access to the Copilot agents built into these services. Each of these services has its own Copilot agent that you can use and modify at no additional cost.

For organizations without the M365 license, Microsoft has developed a specialized calculator of tokens and credits, which allows you to estimate the monthly operating costs of an agent. All you need to do is specify the number of users, the type of queries and the AI functions you’ll use, and the tool will provide an approximate cost. For example, a single message typically costs about $0.01. Although the tool provides an estimate rather than final cost, it gives a good insight into potential expenses, providing invaluable guidance to key business decisions.

Additional features mean additional costs

One of Copilot’s greatest strengths is its ability to perform tasks directly on a computer – e.g. clicking buttons, filling out forms, or searching for files. This functionality is known as a Computer Using Agent (CUA). This feature resembles Power Automate desktop workflows. However, GenAI significantly expands the capabilities of CUA Copilots, making them more intelligent and flexible in operation.

It’s important to know that these operations come with additional costs, e.g. for data storage or system queries. Microsoft provides specific rates: for instance, $0.10 per day for each GB of vector memory (responsible for storage of previous answers/questions and matching the meaning of queries), and $2.50 for 1,000 queries. To keep your budget under control, it’s important to design and optimize consciously. You’ll want to carefully analyze what you really need and what can be simplified.

Standardizing Power Platform app development with Git

While low-code and no-code solutions are intuitive, they still require occasional changes and continuous development. What’s more, because they’re simpler to create, updates can come not just from programmers but also from business users or so-called citizen developers. This makes it essential to manage the development process for low-code and no-code apps with the same care as fully coded applications.

Connecting your Power Platform solutions to a Git repository (such as GitHub or Azure DevOps) is a current DevOps best practice for supporting development and change management. Microsoft emphasized that native Git integration provides a standardized CI/CD process, faster environment launches, and better version control. With GitHub Actions and Power Platform Build Tools, you can automatically build, test, package, and deploy apps, maintaining complete version control.

Connecting Copilot Studio to GIT

To make standardization even easier, the Power Platform Preview environment now includes Git Integration. This tool lets you directly connect Power Apps, Automate, and Pages to a repository. This allows you to isolate developer work, audit changes, easily revert to older versions, and track history using Git. At Netwise, we consider this the standard – every major change should be versioned in Git. This practice improves the quality of deployments and speeds up team collaboration.

Sample agent repository in Azure DevOps

The Copilot Studio Implementation Guide

To make the implementation of Copilot Studio even easier, Microsoft has created the Implementation Guide. This practical document guides your team through every step of building an agent – from defining business goals and analyzing data to maintaining security and effectiveness. It includes control questions, design patterns, and a list of best practices. With this guide, you can avoid common mistakes and move from concept to real-world implementation much faster. It provides excellent technical support and helps you plan well for an effective return on your investment.

Adding data sources with MCP

Model Context Protocol (MCP) is a technology that allows agents to use enterprise data and tools without tedious, manual configuration. Simply connect a data source (e.g. CRM or ERP system) and the agent will automatically gain access to it. MCP simplifies integration by offering a library of verified MCP servers, allowing businesses to quickly select an existing solution or build their own using the Microsoft SDK.

For example, connecting an MCP server that handles bank data immediately gives the agent the ability to perform financial operations without manual configuration. This generative context expansion makes agents faster and requires much less maintenance. Microsoft emphasizes that MCP drastically reduces the need for manual intervention and speeds up the development of new features.

The simplified configuration of external connections or third-party data sources is a huge time and money saver. Instead of building everything from scratch, we can use ready-made components. For businesses, this means faster agent deployment and less risk of project delays.

Example configuration of an agent using an external data source

Get started faster with Copilot Studio Kit

A great help when building agents is the Copilot Studio Kit. It is a set of tools, pre-built components and test environments that make it easier to start a project. Sample capabilities include:

Automated testing framework – allows you to define test suites that verify the agent's answers and compare them with the expected results (e.g. using additional validation instructions),
Conversation KPI views – save and aggregate statistics (e.g. number of sessions, resolution/deflection indicators) in Dataverse, making it easier to analyze agent effectiveness than through analyzing raw logs,
Adaptive Cards – a gallery of notification templates for typical UI scenarios,
Agent Inventory – a panel displaying all Copilot agents in the organization and their configuration (knowledge sources, orchestration types etc.),
Agent Review Tool – automated deployment analysis tool (detects anti-patterns and helps fix them).

In addition, the Kit enables SharePoint content synchronization (import of knowledge files to the agent database, including PDF and aspx pages) and a webchat appearance editor (allowing you to easily adjust colors and fonts or generate HTML code). With this Kit, teams don’t have to build everything from scratch but can focus on tailoring the agent to real business needs.

Copilots pay off

From a business perspective, the key question is: is it worth the effort? My answer is yes. Although the implementation of Copilot requires initial investment (licenses, time, configuration), the savings are significant. Each properly functioning Copilot means hundreds of hours saved on human work. The cost of one agent session is about $0.50, while the same service performed by an employee can cost up to 10 times more.

At Netwise, we have been helping companies implement Power Platform and Dynamics 365 for years. We see that generative automation tailored to specific needs is the future of work. That is why we know that it is worth betting on Copilot Studio: it means well-thought-out, quick implementations and measurable results. If you want to learn more about how Copilot Studio can speed up operations at your company, feel free to contact me.

Unlock the power of agents

Discover the possibilities with Copilot Studio and empower your teams with smart, responsive AI solutions.

Copilot Studio – managing with Git, keeping costs under control and Copilot Studio Kit

How to estimate the costs of Copilot agents?

Additional features mean additional costs

Standardizing Power Platform app development with Git

The Copilot Studio Implementation Guide

Adding data sources with MCP

Get started faster with Copilot Studio Kit

Copilots pay off

Unlock the power of agents

See the latest insights from Netwise

CRM in the age of Autonomous Agents

Copilot Studio – managing with Git, keeping costs under control and Copilot Studio Kit

Enabling the future of work: A deep dive into Microsoft Copilot Studio