This event, designed to foster learning and innovation among teams, provides an opportunity to build custom agents using Copilot Studio for your business use cases, with direct guidance from Copilot Studio experts. 

The key aspects your team should keep in mind when building your custom agent include the following:

• Innovative use of generative AI
• Business value
• Technical feasibility
• Enterprise readiness (including data security and compliance)

Participants will receive support through various learning sessions to envision their use cases, hands-on labs, inspirational demos, and office hours with tech experts. Additionally, your team will have access to a Teams channel with daily tips and async Q&A support from experts throughout the length of the hackathon.

All hackathon submissions will have an opportunity to win Credly digital badges and certificates, branded swag, and a chance to be featured in a customer story for their team and project.  

Recent success story

The Power CAT team recently wrapped the pilot Enterprise Hackathon, which invited teams from 13 enterprise organizations to build impactful and compliant enterprise agents using Copilot Studio to address real business use cases. We received incredible submissions from the enterprise teams throughout the hackathon. Here’s what some of the participants had to say:  

“Hearing all the new ideas from everyone is really inspiring and helped us think about our next step and implementing those ideas into our ideas as well.”

“One part of the hackathon was asking questions in the community chat. The interaction & the experience of the other people was amazing to understand how we can fix it and continue building.”

“Copilot Studio is very user-friendly. It helps innovate and streamline development process, so it can make time to delivery very fast and efficient.”

“Go for Copilot Studio so you can have a more robust solution with your use case.”

Registration is now open

Ready to innovate and transform your business? Gather your team and your proposed use cases and sign up today! We look forward to your team’s participation, engagement, and innovation at the hackathon. 

Questions? Email Enterprise Agent Challenge Support at EACSupport@microsoft.com.

The post Register now for the upcoming Copilot Studio Enterprise Agent Challenge! appeared first on Microsoft Copilot Blog.

Today, we are excited to announce that computer use is coming to Copilot Studio through an early access research preview. This new capability allows your Copilot Studio agents to treat websites and desktop applications as tools. With computer use, agents can now interact with any system that has a graphical user interface!

Achieve new efficiencies with computer use 

Computer use enables agents to interact with websites and desktop apps by clicking buttons, selecting menus, and typing into fields on the screen. This allows agents to handle tasks even when there is no API available to connect to the system directly. If a person can use the app, the agent can too.

Computer use adapts to changes in apps and websites automatically. It adjusts in real time using built-in reasoning to fix issues on its own, so work continues without interruption. It is also built on Copilot Studio’s robust security measures and governance frameworks, to help ensure compliance with organizational and industry standards.

With computer use in Copilot Studio, makers can build agents that automate tasks on user interfaces across both desktop and browser applications, including Edge, Chrome, and Firefox. Additionally, computer use runs on Microsoft-hosted infrastructure, meaning organizations don’t need to manage their own servers. Enterprise data stays within Microsoft Cloud boundaries and is not used to train the Frontier model. This helps your organization accelerate deployment, reduce maintenance, and lower infrastructure costs.

Unlock new value with agentic and automation scenarios

To bring this technology to life, consider the following high value use cases:

• Automated data entry: Imagine a scenario where an enterprise needs to input large volumes of data from various sources into a centralized system. Computer use can automate this process, reducing manual effort and minimizing errors.
• Market research: Marketing teams can leverage the tool to automate the collection of market data from various online sources for analysis, providing valuable insights without the need for manual intervention.
• Invoice processing: For finance departments, the tool can automate the extraction of data from invoices and input it into accounting systems, streamlining the entire invoicing process and reducing manual errors.

Reimagining robotic process automation (RPA)

Computer use agents are transforming robotic process automation (RPA). They overcome traditional limitations, like the fragility of UI elements, and can handle complex dynamic interfaces. This makes automation accessible to people beyond professional RPA developers.

In Copilot Studio, computer use addresses common RPA challenges by making automation smarter and more intuitive:

• It responds to changes in real time: When buttons or screens change, the tool keeps working without breaking your flow.
• It is easy to use: You can describe what you want in natural language, no coding needed, and test and refine the prompt with real-time side-by-side video of the computer use reasoning chain and the planned UI automation.
• It is built with intelligence: The agent sees what is on the screen and makes smart decisions in real time, even in complex or constantly changing environments.
• It comes with full visibility: Makers can view a history of computer use activity at will, including captured screenshots and reasoning steps.

The future of innovation with Copilot Studio

Copilot Studio is the end-to-end agent platform designed to help organizations achieve their AI and operational goals. We want to empower you to streamline processes, enhance productivity, and drive innovation.

If you are interested in exploring the new computer use capability, we would love to hear from you! Please fill out this form to let us know you would like to participate.

We will also share more about this new announcement at Microsoft Build in May 2025—register here to join us! 

The post Announcing new computer use in Microsoft Copilot Studio for UI automation appeared first on Microsoft Copilot Blog.

Expanding Copilot capabilities with UiPath Agent Builder

Last year, we showcased how UiPath Autopilot could be seamlessly integrated into Teams and Copilot conversations (here). Now, this bi-directional integration between Microsoft Copilot Studio and UiPath Studio unlocks new use cases and scenarios and helps our mutual customers solve unique business challenges.

Graham Sheldon, UiPath Chief Product Officer, highlighted the significance of this partnership: “We believe in the power of people, agents, and robots working together to solve tough, ambiguous business problems,” he said. “This collaboration with Microsoft brings the promise of agentic automation and orchestration to life for our customers, offering unparalleled capabilities and flexibility.”

UiPath has enhanced their Power Platform connector to enable the integration of UiPath agents and automations directly into Copilot Studio. Developers can build actions, topics, or agent flows that include UiPath agents and automations with ease.

Once the delegated task is complete, the UiPath agent closes the loop using an API trigger to hand back to Copilot and continue the process. This enables Copilot Studio agents to work across Microsoft 365 and activity in UiPath Studio, including applications such as Outlook and SharePoint.

Real-world business processes are complex, spanning IT systems, offline documents and user actions. This integration, built with close collaboration between the engineering teams, facilitates seamless interaction between UiPath and Microsoft agents and supports orchestration across both platforms.

Joint customers can now automate and orchestrate tasks that span multiple applications, semi-structured and unstructured documents, and User actions. The bi-directional connectivity unlocks several scenarios:

• Invoke UiPath agents & automations from Copilot Studio agents: Dynamic input parameters in the connector enable Copilot Studio agents to hand off processes to UiPath agents with the necessary data required to complete the task.

• Invoke Microsoft agents from UiPath Studio: UiPath Studio activities enable the ability to use Copilot Studio agents as tools/functions in UiPath Studio, enabling UiPath agents in turn, to handoff process to Copilot Studio agents. 

• Automate long-running and asynchronous processes: UiPath agents can now hand off a process back to the Copilot Studio agent to resume processing and trigger flows. This allows a Copilot Studio agent to delegate a task to a UiPath agent and wait for task completion before resuming the next steps.

Real-world applications

The UiPath Connector helps customers tap into UiPath agentic capabilities in the Microsoft platform, third-party systems, documents, and desktop applications. For example, a Copilot Studio agent monitoring a shared Outlook inbox for customers can perform email generation, intelligent case assignments, and record updates based on emails.

For specialized tasks like claims processing, it delegates to a UiPath claims agent, which leverages their Intelligent Document Processing to extract information, validate against existing data in systems of record, and input into the claims administration platform for handoff to the claims adjuster.

If an escalation is required, UiPath agents can identify the right user using Microsoft Graph and raise the exception in UiPath Action Center. Once the claim is processed, confirmation is passed back to the Copilot Studio agent for post-processing steps.

Get started today

Sangya Singh, Microsoft VP of Power Platform Intelligent Automations, noted that this integration will help makers solve complex workflows and optimize business processes. “As we expand the surface area of what agents can do in the enterprise,” she said, “we are making it easier for customers to achieve their agentic goals.”

Customers have also expressed excitement about this partnership, especially as it helps with business process automation across an organization:

You cannot automate a process in isolation; integrating across technology boundaries is necessary for real business impact. This bi-directional integration harnesses the combined strengths of Microsoft Copilot and UiPath agents to fully automate complex workflows across documents, emails, PowerApps, and enterprise systems.” 

– Ramnath Natarajan, Director of Global Intelligent Automation & Integration
Johnson Controls

We look forward to seeing the many ways Microsoft and UiPath customers will leverage these new capabilities to drive their business forward. Try UiPath Connector actions in Copilot Studio today.

The post Announcing Agentic Automation with Bidirectional Integration between Microsoft Copilot Studio & UiPath appeared first on Microsoft Copilot Blog.

Agents and agent flows: Better together

Agents are intelligent entities designed to independently perform tasks and make decisions based on predefined goals, working alongside or on behalf of an individual, team or organization. Agent flows are built to automate deterministic workflows and are enhanced with powerful AI actions. By coupling agents with agent flows in Copilot Studio, customers not only unlock unparalleled efficiency and agility but also experience these benefits within a single, unified platform, simplifying automation from design to deployment.  

Agents can dynamically handle complex scenarios—like tax audits—by using generative AI to make real-time decisions. For example, an agent determines when an audit should be triggered based on anomalies in financial data. However, every organization has unique and indispensable processes with some fixed components. Now with agent flows, those fixed components are easily automated. Once triggered, an agent passes the context to its pre-built agent flow, which executes a structured, predefined audit procedure. This procedure systematically collects required documents, performs intelligent summarization, and secures the necessary, organization-specific, human-in-the-loop approvals.   

Copilot Studio’s intuitive integrated maker interface supports quick iteration, allowing agents and intelligent workflows to evolve seamlessly alongside business needs. Agent flows can also be reused across multiple autonomous agents, ensuring consistency, optimizing tasks universally, and simplifying process management. 

Agent flows: Intelligent, deterministic workflow automation

Unlike purely dynamic autonomous orchestration in agents today, deterministic workflows leverage predefined action sequences. This ensures that tasks always run consistently and rapidly. Key benefits include control and monitoring, providing transparency and accountability as well as enabling swift identification and resolution of workflow issues. Ideal for highly repetitive steps, agent flows can significantly accelerate workflow execution. 

For example, standard finance operations like invoice processing, which are critical yet often costly to operate, can greatly benefit from agent flows. When an invoice arrives, the agent flow can automatically perform intelligent document processing to extract relevant data, validate entries against existing records, and route invoices for approval with impressive accuracy and speed.  

The intelligence behind agent flows

Copilot-assisted natural language authoring experience

Just like building agents in Copilot Studio, makers can effortlessly create agent flows by simply describing their intent in natural language. Powered by the latest language models, Copilot intelligently interprets this input—understanding user context, intent, and logical structure—and generates the appropriate workflow steps automatically. This dramatically simplifies the authoring experience, enabling users to go from idea to automation faster, without needing deep technical expertise.

Enterprise-grade workflow building blocks

Agent flows offer the following rich set of workflow action types that empower makers to build intelligent, scalable, and flexible automations tailored for enterprise needs: 

1. AI actions: Use intelligent document and image processing to handle complex documents and visual data efficiently, intelligent summarization to quickly synthesize information from structured and unstructured data, and intelligent reasoning to enhance workflows by enabling AI-driven recommendations and decisions. 

2. Generative actions: Dynamically plan and execute automation steps, ensuring real-time adaptability and ongoing improvement through generative AI. 
3. Integration actions: Connect seamlessly with enterprise business applications and external services, enhancing interaction with systems-of-record internally and externally. 
4. Human-in-the-loop actions: Embed human judgment through advanced approval mechanisms and collaboration tools. In addition to notifications via Outlook, Gmail, Teams, and Slack, standard and advanced multi-stage approvals incorporate human decisions effortlessly into automation processes. Advanced approvals include customizable reminders, delegation features for unavailable approvers, and advanced, multi-stage approval mechanisms. This ensures critical business decisions are timely, efficient, and adaptable to complex scenarios.

Copilot Studio: Your agent platform for enterprise-grade agentic automation 

By combining outcome-driven autonomous agents with deterministic agent flows, Copilot Studio stands as the premier AI-first agent platform for transforming business processes at scale. This unified approach empowers organizations to achieve unprecedented efficiency, agility, and innovation, setting a new standard for intelligent automation in the enterprise. 

To get started, simply click the Flows tab in the left-hand pane of Copilot Studio. For more details, check out our comprehensive documentation and licensing guide so you can get started on this exciting journey.  We look forward to seeing how you use agent flows to power the automation behind your business process transformation.  

The post Introducing agent flows: Transforming automation with AI-first workflows appeared first on Microsoft Copilot Blog.

1. Autonomous agents: now generally available
2. Deep reasoning in agents
3. Agent flows in Copilot Studio
4. Generative orchestration: now generally available
5. Model Context Protocol
6. Tools and prompts inside Copilot Studio
7. Open web search
8. Find and install Microsoft-built agents inside Copilot Studio
9. Private preview opportunity: try GPT-4.5 in Copilot Studio

1. Autonomous agents: now generally available

Responding instantly to critical events is more vital than ever, especially for organizations operating on a global scale. Whether it’s a finance team managing budget thresholds, a sales team seizing renewal opportunities early, or a frontline team addressing urgent low-supply alerts, timely action drives efficiency and impact.

That’s why we’re thrilled to announce the general availability of autonomous agents, designed to keep you proactive by automating responses to critical events. The need for manual checks, scheduled workflows, or complex custom API solutions is gone. Using triggers, you can configure your agents to wait for specific events and execute a set of actions when that event happens. The instructions you provide—written in natural language, not code—add autonomous capabilities to your agent which can then monitor, detect, and react to pertinent events to automatically carry out the workflow you defined.

These workflows can handle a wide variety of tasks, from simple actions, like notifying users of updates, to more complex automation, such as automatically ordering new equipment for your department when a new employee joins. While these agents run in the background, you’ll maintain full visibility into every action the agent takes through the Activity tab, ensuring transparency and control.

With triggers, agents evolve from working with you, to also working for you. Autonomous agents dramatically save time for individuals, teams, and entire organizations by automating business processes and getting tasks done across every department.

Learn more about getting started with triggers here.

2. Deep reasoning in agents

On March 25^th, CVP Charles Lamanna announced the preview release of deep reasoning for agents, which enables agents to handle complex business processes more effectively. With this new release, agents can leverage advanced reasoning models like OpenAI o1, along with your enterprise data, to process information, analyze data, and help you take action in subjective, complex situations.

Deep reasoning can be applied in both declarative and autonomous agents—and, in fact, you can build your own deep reasoning agents to solve complex problems specific to your organization and data. There are many ways to use deep reasoning agents’ critical thinking skills, including analyzing intricate patterns in demand forecasting, qualifying and categorizing sales leads, identifying opportunities and providing smarter recommendations, and more.

Check out this video for an example demonstration of how deep reasoning can help a team create RFPs and respond to customers more quickly:

Copilot Studio agents with deep reasoning capabilities process information more thoroughly, analyze complex business data sets to uncover deeper insights, and help enterprises make better decisions. Employing these skills in autonomous agents, especially, can create a whole new world of opportunity to offload, streamline, or augment business processes throughout your organization.

Learn more about deep reasoning in autonomous agents here.

3. Agent flows in Copilot Studio

At the other end of the reasoning spectrum, agent flows in Copilot Studio enable you to execute repeatable business processes by automatically following a predefined sequence of actions. Generally available on March 31^st, agent flows are structured, rules-based workflows that incorporate AI actions to quickly produce predictable results.

When do agent flows make sense over the use of a self-assembling agent? They’re ideal for business processes that require consistency and control, such as routing tasks or applying rules, and can be very helpful in maintaining compliance. Because they are designed to replicate a process every time—even hundreds or thousands of times a day—agent flows don’t require the reasoning or generative effort of a self-assembling agent, which makes them faster.

Note that agent flows can operate independently or integrate as actions within existing agents. Built using prebuilt or custom connectors, flows can also be reused across agents to reduce setup time and standardize efficiencies across your organization.

Watch this demo and see how you can build an agent flow inside Copilot Studio to manage customer feedback:

With options to add both deep reasoning and agent flows to your automated business processes, you can choose the right AI tool for the job at hand and enjoy a truly well-rounded agentic experience.

4. Generative orchestration: now generally available

Anticipating every possible question an end-user might ask agents is nearly impossible. With generative orchestration now available in general availability for English (en-US) agents, you no longer have to try. This powerful feature eliminates the need to manually define countless trigger phrases or delineate every conversation flow through topics.

Unlike classic orchestration, which relies on predefined, manually authored topics, generative orchestration enables agents to generate contextual intelligent responses on the fly based on your organizational knowledge sources, actions, and topics. Agents can even generate their own questions to collect information they need to solve the end-user’s problem, or ask for confirmation when needed before executing an action. This creates more natural conversations, decreases escalations and abandonment rates, and broadens the range of scenarios agents can successfully support.

Generative orchestration further enables your agents to handle multiple requests that come in a single query, such as an end-user asking to find their nearest store and get that store’s opening hours. By being able to comprehensively answer the end-user’s questions, the agents create a better experience and higher satisfaction during agent interactions.

To enable generative orchestration, simply toggle it on from your agent’s overview page. You can also find it in “Agent Settings” under the Generative AI tab. Try it out today and unlock more intuitive, more engaging, and more natural conversational agent experiences.

Learn more about generative orchestration here.

5. Model Context Protocol

In March, we announced the public preview of Model Context Protocol (MCP) support in Copilot Studio — a powerful new way to streamline how you connect AI apps, APIs, and data sources into your agents. MCP makes it easy for makers to integrate knowledge servers and external tools directly into Copilot Studio, helping you build and scale intelligent agents faster than ever.

MCP connectors automatically bring actions and knowledge from your server into your agent, including names, inputs, outputs, and descriptions, and keeps them in sync as updates happen on the backend. That means no more manually updating outdated tools. Even better, you now have access to a growing library of pre-built, MCP-enabled connectors in the marketplace, making it faster and easier to tap into existing systems while maintaining enterprise-grade security and governance.

For makers, MCP unlocks a new level of flexibility and efficiency. With Software Development Kit (SDK) support, you can build your own MCP server, publish it through a connector, and instantly use those tools in your agents. Whether you’re working with internal APIs or third-party services, MCP empowers you to create more responsive intelligent agents with less complexity, and more time to focus on innovation.

Learn more about how to get started with MCP here.

6. Tools and prompts inside Copilot Studio

Copilot Studio has a new centralized place where makers can create, manage, and assign reusable functionality across agents: Tools. At the Copilot Studio application level, you’ll find Tools in the left-hand navigation. Click into it to try out the first tool available, Prompts.

In Copilot Studio, prompts instruct the GPT model to behave in a certain way or perform a specific task. A custom prompt can guide the model to answer a question, complete text, translate languages, summarize a document, and identify tasks, to-dos, and action items in text. The complexity of a custom prompt can range from a single sentence to something more intricate, depending on the task. When this prompt is assigned to an agent, the agent follows the prompt before responding, giving the maker more control over the responses the agent will generate and resulting in higher-quality answers or actions from the agent.

The Prompts tool allows makers to easily create custom prompts inside Copilot Studio, either at the app level—then assigning the prompts to agents—or within an agent or topic. Inside the tool, you’ll find out-of-the-box templates to help you instruct the model effectively, or you can write the prompt from scratch. Then you can test the prompt with sample data or inputs before assigning it to an agent.

The Tools menu and in-app prompt builder are powerful new additions for two reasons. First, they streamline the experience so that it’s easier to get started on prompts, easier to test, and easier to work on prompts at the level you need them (app, agent, or topic), all inside one tool. These prompts also use the same consumption-based meter as those created in the external tool, no extra license or cost involved.

Second, they make these functions endlessly reusable. Makers can build a prompt once and assign it to multiple agents, and any maker can use prompts created by other makers in the same Copilot Studio environment. More consistent, more collaborative, and more scalable. All this results in more effective agents and more efficient business processes.

Future updates will expand available tool types, but you can start using the prompt builder inside Copilot Studio right away. Check out Tools in the left navigation of Copilot Studio, and keep an eye out for new tools coming soon!

7. Open web search

Open web search has arrived in Copilot Studio, giving your agents the ability to surface high-quality, real-time information from the web when configured knowledge sources fall short. This new capability helps your agents deliver relevant responses in broader, more open-ended scenarios like research, content creation, and competitive analysis—without requiring makers to anticipate every possible data source in advance.

Connecting agents to publicly available information through open web search significantly expands their ability to provide accurate and up-to-date answers. Whether enhancing internal knowledge with the latest technological advancements or empowering end-users with richer, more flexible responses, this integration ensures your agents stay relevant and valuable. This can help to increase usage, drive higher satisfaction rates, and create greater impact.

Build smarter, more adaptable agents that seamlessly meet a broader range of business needs with this capability, now available in public preview.

8. Find and install Microsoft-built agents inside Copilot Studio

The Microsoft-built and Microsoft-certified agent catalog is now available in preview for admins to discover and install within Copilot Studio! On the “Create” page in Copilot Studio, admins will now see available agents under the “Built by Microsoft” heading and can click on an agent to install it.

Once installed, the admin can go to the new agent from the installation dialog box or the banner at the top of the page. From there, they can customize the agent, install first-party add-ons, and configure connections as needed.

By accessing the whole agent installation process inside Copilot Studio, it’s faster and easier to create the agents you need, empowering your team to more efficiently deploy, manage, and scale properly configured agents across your organization and processes.

Learn more about how to install agents from the catalog here.

9. Private preview opportunity: try GPT-4.5 in Copilot Studio

Copilot Studio is actively testing GPT-4.5, the latest AI model from OpenAI, to power generative answer experiences. Today, we are making GPT-4.5 available to customers through a limited preview!

Users will have the opportunity to explore the capabilities of Copilot Studio agents powered by GPT-4.5 on an agent-by-agent basis. If you would like to explore GPT-4.5 in Copilot Studio and register your interest in the preview, click here.

More ways to stay up to date on all things Copilot Studio 

Check out all the updates live as we ship them, as well as new features releasing in the next few months here: What’s new in Microsoft Copilot Studio – Microsoft Copilot Studio | Microsoft Learn 

Get started today 

To learn more about Microsoft Copilot Studio and how it can transform your organization’s productivity, visit the Copilot Studio website or sign up for our free trial today. 

The post What’s new in Copilot Studio: March 2025 appeared first on Microsoft Copilot Blog.

What’s new: Model Context Protocol integration

Model Context Protocol (MCP) enables makers to connect to existing knowledge servers and APIs directly from Copilot Studio. When connecting to an MCP server, actions and knowledge are automatically added to the agent and updated as functionality evolves. This simplifies the process of building agents and reduces time spent maintaining the agents.

MCP servers are made available to Copilot Studio using connector infrastructure. This means they can employ enterprise security and governance controls such as Virtual Network integration, Data Loss Prevention controls, multiple authentication methods—all of which are available in this release—while supporting real-time data access for AI-powered agents.

MCP enables our customers to:

• Easily connect to data sources: Whether you have a custom internal API or external data providers, the MCP protocol enables smooth and reliable integration into Copilot Studio.
• Access the marketplace of existing servers: In addition to custom connectors and integrations, users can now tap into a growing library of pre-built, MCP-enabled connectors available in the marketplace. This capability gives you more ways to connect with other tools and makes using them faster and easier.
• Flexible action capabilities: MCP servers can dynamically provide tools and data to agents. This enables greater flexibility while reducing maintenance and integration costs.

To get started, access your agent in Copilot Studio, select ‘Add an action,’ and search for your MCP server! (Note: generative orchestration must be enabled to use MCP.)

Each tool published by the MCP server is automatically added as an action in Copilot Studio and inherits the name, description, inputs, and outputs. As tools are updated or removed on the MCP server, Copilot Studio automatically reflects these changes, ensuring users always have the latest versions and that obsolete tools are removed. A single MCP server can integrate and manage multiple tools, each accessible as an action within Copilot Studio. This streamlined process not only reduces manual effort, it means less risk of errors from outdated tools.

This offering additionally includes Software Development Kit (SDK) support, enabling further customization and flexibility for your integrations. To create your own Model Context Protocol server, the process can be broken down into three key steps:

1. Create the server: The first step in integrating Copilot Studio with the MCP is to create a server via one of the SDKs that will serve as the foundation for handling your data, models, and interactions. You can tailor the server to your specific needs, such as enabling it to handle custom model types and data formats or to support specific workflows.   
2. Publish through a connector: Once the server is in place, the next step involves creating a custom connector that links your Copilot Studio environment to the model or data source.
3. Consume the data via Copilot Studio: Finally, once the server is set up and the connector is defined, you can begin consuming the data and interacting with the models via Copilot Studio.

By following these three steps, you create a streamlined, adaptable integration with Copilot Studio that not only connects systems but also enhances your ability to maintain and scale that integration according to your needs.

We support Server-Sent Events (SSE) as the transport mechanism; this feature is currently in environments in preview regions and will be available across all environments shortly.

Learn more about these new capabilities here: Extend your agent with Model Context Protocol (preview) – Microsoft Copilot Studio | Microsoft Learn.

What’s next?

We’re excited about the potential of Model Context Protocol and its ability to transform the way users interact with Copilot Studio. But this is just the beginning. Our team is actively working on additional features and improvements to further enhance the integration experience. Stay tuned for more updates, as we plan to introduce even more ways to connect your data and tools effortlessly into Copilot Studio.

We look forward to your feedback and learning more on how this new capability enhances your experience and helps you unlock the full power of Copilot Studio.

The post Introducing Model Context Protocol (MCP) in Copilot Studio: Simplified Integration with AI Apps and Agents appeared first on Microsoft Copilot Blog.

To that end, we are always experimenting with new technologies to improve agent experiences, including the latest model innovations. We’re excited to share that Copilot Studio is actively testing GPT-4.5, the latest AI model from OpenAI, to power generative answer experiences. 

Today, we are making GPT-4.5 available to customers through a limited preview. Users will have the opportunity to explore the capabilities of Copilot Studio agents powered by GPT-4.5 on an agent-by-agent basis. If you would like to explore GPT-4.5 in Copilot Studio and register your interest in the preview, click here.

“Early testing shows that interacting with GPT4.5 feels more natural. Its broader knowledge base, improved ability to follow user intent, and greater ‘EQ’ make it useful for tasks like improving writing, programming, and solving practical problems. We also expect it to hallucinate less.” – OpenAI 

We recognize that GPT-4.5 is one of many models, and we remain committed to enabling makers to use your model of choice in your agents – whether that’s one of the more than 1,800 models in Azure AI Model Catalog, a custom model built in Azure AI Foundry, or a model fine-tuned on your business data.

We look forward to sharing more about model choice and innovation in Copilot Studio at Microsoft Build in May 2025. Of course, models are only one part of an agent, so keep an eye out—we have much more to share about new Copilot Studio capabilities in the coming weeks. 

More ways to stay up to date on all things Copilot Studio

Bookmark this page to learn about features as we ship them, as well as upcoming releases: What’s new in Microsoft Copilot Studio – Microsoft Copilot Studio | Microsoft Learn.

The post Announcing a Limited Preview of GPT-4.5 in Copilot Studio appeared first on Microsoft Copilot Blog.

Microsoft is no exception and as we sought to build an employee self-service agent, we learned a great deal about how to securely build, deploy, and measure the success of an enterprise-wide agent.

In this blog post, you’ll understand our best practices for deploying enterprise-wide agents with a focus on our own development and deployment of the Employee Self-Service (ESS) agent in Microsoft Copilot Studio to Microsoft 365 Copilot respectively.

Our journey to a simple enterprise-wide employee self-service agent began back in 2013 with the launch of our HRWeb SharePoint site as a centralized repository of critical HR policy and benefits information. That led us to eventually build a chatbot in 2021 to reduce HR inquiry escalation to our HR policy teams.

Now with generative AI, we’ve revolutionized employee support by building the ESS agent to support HR inquiries, dramatically enhancing efficiency and the employee experience.

Employees receive accurate, context-driven responses and automated actions—from policy lookups to managing time-off requests—significantly reducing HR costs and enabling greater employee productivity and satisfaction.

Following a successful pilot, Microsoft Digital expanded the ESS agent globally across key markets, including the United States, the United Kingdom, and India.

The journey of creating, securing, and rolling out the ESS agent enterprise-wide revealed five key considerations when building and deploying an enterprise-wide agent:

1. Plan with purpose: Define why, success, and the challenges you’re solving
2. Select and secure optimal knowledge sources for your AI agent
3. Ensure security, compliance, and responsible AI in enterprise-wide agent deployments
4. Build and test the pilot agent with target audiences
5. Scale enterprise-wide adoption and measure impact

Plan with purpose: Define why, success, and the challenges you’re solving

Before you write a single line of code or plan the agent design, it’s critical to define your agent’s purpose. Identify the specific challenges your organization is trying to address. Ask questions such as:

• What pain points exist in the current support experience?
• Who are the end users, and what does their ideal interaction look like?
• How can generative AI improve the resolution process compared to legacy chatbots?

To learn how Microsoft built the agent prompt and responses, check out our  Copilot Studio Implementation Guide blog post.

At Microsoft, we recognized that our existing chatbot solution—while effective—wasn’t optimal, as it required developers to anticipate every possible user question and compelled employees to use exact verbiage to find HR and benefits information, as well as IT-related queries.

To significantly improve this experience, we envisioned an enterprise-wide AI agent using generative answers, orchestration, and actionable capabilities, enabling higher resolution rates and delivering a more intuitive, efficient support experience for our employees.

Below is an example of goals to outline when designing your agent:

Select and secure optimal knowledge sources for your AI agent

Identify core data that optimally solves user challenges

Once you’ve determined the agent’s purpose, the next critical step is identifying the knowledge sources and data on which you’ll build your agent. For most organizations, especially in your initial enterprise-wide agent development and deployment, you’ll want to restrict yourself to the most essential knowledge sources for a few reasons:

• These knowledge sources should be among the most secure in your organization.
• You’ll typically employ some form of role-based access controls (RBACs) to prevent data proliferation.
• They should provide the breadth of knowledge necessary to effectively serve your initial target audience(s) and should be accurate and up to date.

At Microsoft, our journey in HR self-service began in 2013 with HRWeb, our SharePoint intranet, which provided extensive coverage and robust security through built-in role-based access controls. Using HRWeb as the foundation for our generative AI agent allowed us to efficiently deliver comprehensive, secure answers with minimal configuration.

However, recognizing that even high-quality data can become outdated, we proactively analyzed common user searches and frequent support tickets to curate a targeted set of essential questions and answers. This practice highlights the importance of regular data reviews and cleanup to continually optimize your agent’s effectiveness.

Implement data loss prevention policies for knowledge and data connectors

At Microsoft, we recommend establishing three separate environments—development, test, and production—in Copilot Studio, aligned with software development lifecycle best practices.

Using Copilot Studio alongside Microsoft Power Platform, organizations can enforce consistent routing rules and data loss prevention (DLP) policies across all environments, specifically targeting knowledge sources, connectors, and API endpoints.

This approach ensures secure, compliant data interactions, which at Microsoft is mandatory whenever an agent performs actions beyond simple personal retrieval. Learn more about implementing DLP policies in Copilot Studio.

Ensure security, compliance, and responsible AI in enterprise-wide agent deployments

After building your pilot agent in the development environment, it’s crucial to perform thorough security, compliance, privacy, responsible AI, and accessibility assessments before moving into testing and production environments.

At Microsoft, this begins with a comprehensive software development lifecycle (SDL) assessment, where development teams submit detailed documentation to our internal security council, closely monitoring data usage and enforcing stringent security requirements. Key components of the SDL evaluation include:

• Service Tree Metadata
• Confirmation that no production data resides in dev/test environments
• Verification of data encryption at rest and in transit
• Completed threat modeling analysis
• Documentation of secure development standards
• Verified auditing and logging for agent interactions and chat transcripts

Following the SDL assessment, we perform rigorous accessibility tests aligned to Microsoft’s standards and comprehensive responsible AI evaluations by subject matter experts to ensure responses are accurate, consistent, and inclusive. Additionally, proactive “red team” security testing is recommended to identify and mitigate potential vulnerabilities.

Agents then transition into testing through Microsoft Power Platform pipelines in Copilot Studio, making compliance management especially critical. At Microsoft, we perform additional Tenant Trust Evaluations, including detailed documentation, comprehensive security questionnaires, and formal IT council reviews, to ensure readiness for broader deployment. For sensitive scenarios involving personal data or employee monitoring, we also conduct Works Council evaluations to maintain strict adherence to privacy standards.

We repeat these rigorous assessments whenever introducing new knowledge sources, connectors, or APIs. For example, during the integration of our HRWeb-based agent with our IT Helpdesk agent—creating the unified Employee Self-Service (ESS) agent—these evaluations helped manage data proliferation risks, enabling the ESS solution to remain secure and compliant throughout its lifecycle.

Regardless of your organization’s specific processes, documenting each step thoroughly, maintaining transparency, and actively involving your IT and Center of Excellence teams is essential for effective governance and robust security oversight.

Build and test the pilot agent with target audiences

Once appropriate data-loss prevention measures are approved, your organization can begin building the AI agent. For scenarios involving straightforward generative orchestration—primarily reading and writing—development is streamlined.

However, when handling sensitive data requiring consistent, precise responses, we recommend using the full Copilot Studio instead of simpler tools like Agent Builder in Microsoft 365 Copilot or Microsoft 365 Copilot Chat.

This approach allows the creation of customized response topics, triggered by specific phrases, enabling compliance and accuracy.

At Microsoft, we transitioned to Copilot Studio to precisely handle sensitive HR inquiries, avoiding non-compliant or inconsistent responses for complex questions such as those related to employee compensation comparisons.

Select pilot audience: Select your ideal early adopters for initial feedback

Choosing the right pilot audience and timing is as critical as selecting high-quality data and securing your agent. We recommend initially deploying your AI agent to a focused group—around 100 users—and conducting A/B testing against existing solutions to measure its impact.

Gathering targeted feedback from these early adopters allows for rapid iteration, refinement, and informed expansion of your data sources and agent capabilities.

At Microsoft, we piloted our ESS agent with 100 United Kingdom employees, using the maturity of our HRWeb data in that region to maximize initial impact and insights.

Scale enterprise-wide adoption and measure impact

Rollout at scale: Leverage pilot feedback to refine agent and scale adoption

The initial ESS pilot demonstrated strong results, but scaling the agent enterprise-wide required expanding beyond our primary HRWeb SharePoint site, as critical HR data was distributed across more than 100 different line-of-business (LOB) systems.

Recognizing it wasn’t feasible to integrate all sources simultaneously, we analyzed two years of HR interactions—including tickets, searches, and chatbot conversations—to map data locations, prioritize essential sources, and determine rapid integration strategies.

We focused first on systems easily integrated through existing Microsoft Power Platform connectors and rapidly established corresponding DLP policies. Additionally, we assessed whether certain data sources could be consolidated under the primary HR SharePoint site or other high-value repositories to streamline management.

For maximum organizational impact, Microsoft strategically prioritized deployment in key regions—initially the United States, the United Kingdom, and India—and focused on high-impact teams, such as our sales organization (MCAPS), to showcase the agent externally during customer engagements.

Large enterprises considering similar deployments should anticipate and carefully evaluate these trade-offs to align their AI agent rollout effectively with planned timelines and strategic goals.

Measure and report: Use insights to further refine agent and track ROI

The last stage is where you measure and report on the success of the agent to your team and leadership. Thus, generating the insights needed to refine your agent and demonstrating the ROI to secure continued leadership investment.

At Microsoft, we’re still refining our measurement and reporting strategy for the ESS agent, but some of the key metrics we recommend tracking for any agent are the number of sessions, engagement rates, customer satisfaction (CSAT) scores, resolution rates, abandonment rates, and knowledge source accuracy rates—all of which are available to development teams out of the box in Copilot Studio Analytics.

Final takeaways about enterprise-wide agents

Enterprise-wide deployment can be highly intensive on the front-end but it’s critical that your organization invests most heavily in the planning, data source selection, and security phases.

These are critical to ensure that your organization has the appropriate agent design in mind from the outset that can optimally impact your target key performance indicators (KPIs), data sources that are most effective to solve end user challenges and drive the greatest ROI, and security practices that thoroughly account for any potential gaps and lock down any possibilities for data proliferation.

This blog post is just step one in helping your organization roll out an enterprise-wide agent. In spring 2025 we’re making our ESS agent generally available as a product to all Microsoft customers so you can use not only our learnings in this blog post, but the efforts of our development teams to help you stand up your own ESS agent.

To learn more about the Employee Self-Service Agent read our Inside Track blog from Microsoft Ignite 2024.

The post How to deploy transformational enterprise-wide agents: Microsoft as Customer Zero appeared first on Microsoft Copilot Blog.

• Publish custom agents to Microsoft 365 Copilot Chat
• Add enterprise data with new Microsoft Graph connectors 
• Import and export agent components in Microsoft Copilot Studio 
• Analytics updates, including actions analysis and transcript enhancements
• Security: Cross-prompt injection attack (XPIA) mitigation improvements
• Microsoft 365 Agents Software Development Kit for JavaScript
• Embedded agent builder updates, including public web scoping and user feedback
• Twelve new languages supported in conversational agent creation experience

Publish custom agents to Microsoft 365 Copilot Chat

We’re excited to announce that the ability to publish your custom agents created in Copilot Studio directly to Microsoft 365 Copilot Chat is now in public preview, unlocking new capabilities to customize and extend Microsoft 365 Copilot. This latest update enables your agents to take advantage of the full functionalities of Copilot Studio within Copilot Chat.

Accordingly, your agents published to Microsoft 365 can now utilize features like agent handoff, Azure AI Foundry integrations, orchestration selection, and—where more control of the conversation is needed—topics. They can also automate and execute business processes using autonomous triggers. With this greater functionality comes the need for greater oversight, so makers can also monitor their custom agents’ trends and performance with analytics (see more updates on analytics below).

Customers can also access the full security and governance features in the Microsoft Power Platform admin center, along with messages, quick replies, and adaptive cards enabled for Copilot Chat. Whether you’re launching a new agent or publishing an existing one, you can now benefit from additional channels and enable employees to access these agents through Microsoft 365 Copilot. To get started, head to Copilot Studio, enable the Microsoft 365 Copilot Channel, and select “Submit to Teams or Microsoft 365” when publishing.

Add enterprise data with new Microsoft Graph connections 

Enterprise data is the cornerstone of businesses: shaping strategies, driving decisions, and making it possible for your agents to provide accurate and context-rich responses. This month, we released new out-of-the-box Microsoft Graph connectors in Copilot Studio and Copilot Chat out to public preview. Now you can connect your agents to data in Stack Overflow, Salesforce Knowledge, GitHub, and Unily, which makes it easy for your organizations’ agents to access and use those content sources.

Development teams can now retrieve documentation, code snippets, and track issues directly from GitHub repositories using simple queries from within Copilot Studio. Teams gain immediate access to both community and internal Q&A data, enabling instant solutions to coding and IT challenges. Additionally, sales, frontline, and customer support teams can pull customer insights, FAQs, and case resolutions from Salesforce Knowledge.

To get started, configure the new connectors in the Microsoft 365 admin center. For more details, check out our guide: Set up Microsoft Graph connectors in the Microsoft 365 admin center | Microsoft Learn. Once the applications are connected, add them as Knowledge Sources in Copilot Studio. These connectors will also show up as additional knowledge in the Microsoft 365 agent builder.

Import and export agent components in Copilot Studio 

In many cases, an organization’s agents will share common messaging and content sources, actions, triggers, and entities. Rather than repeatedly copying and pasting these foundational elements for every new agent, makers and admins can now use Component Collection to import, manage, and export core agent components—including topics, knowledge, actions, and entities—across various environments. This process not only accelerates agent creation, but it also allows you to designate a preferred solution that automatically integrates changes through Component Collection and Solution Explorer.

Customers can also configure pipelines and guide makers, which empower admins to centrally govern citizen-led and pro-dev-led projects at scale with less effort. Admins set up the appropriate safeguards that govern and facilitate solution development, testing, and delivery across the organization. Then, makers can submit deployment requests to admins from within Copilot Studio. 

With the ability to set, template, reuse, and govern components, the process of building and scaling agents becomes easier, faster, and more consistent. To learn more about these capabilities, check out the following resources: 

• Create and manage solutions in Microsoft Copilot Studio 
• Export and import Copilot using solutions 
• Create reusable component collections (preview) 
• Set your preferred solution 
• Overview of pipelines in Microsoft Power Platform 

Analytics updates: Actions analysis, transcript enhancements, and Government Community Cloud availability

When it comes to understanding agent experience, optimization, and ROI, analytics are critical. This month, Copilot Studio rolled out several new analytics features. The built-in analytics page will now feature detailed insights and telemetry for the actions configured within the agent. This new view allows makers to identify the most-used actions, evaluate execution success, and optimize workflows with real-time data—critical for continuing to optimize and improve an agent’s performance and value.

Additionally, makers and admins with the “Transcript Viewer” role and permissions can now access a comprehensive view of conversations and responses for trend analysis and diagnostics. This feature has been eagerly anticipated; it provides deeper insights into the conversations your agents have with end users and how they track to the various key performance indicators (KPIs) within the analytics dashboard. These transcripts include detailed topic initiation, node triggers, and outcomes to help diagnose and refine interactions.

Lastly, we’ve expanded the regions in which customers can take advantage of these robust analytics insights by expanding to the Microsoft 365 Government Community Cloud (GCC) in general availability. Review and benefit from these advanced analytics to gauge user engagement, performance, and satisfaction so you can further refine your agents. All these updates empower makers and admins to spot trends, resolve issues faster, and optimize performance, helping agents work more effectively for your end users.

Security: Cross-prompt injection attack (XPIA) mitigation improvements

Every year, cyberattacks get more sophisticated. In this age of generative AI, cross-domain or cross-prompt injection attacks (XPIAs) can occur when there are malicious instructions passed to the agent from an external system. These attacks can result in the agent taking an action or sending a message based on the malicious instructions.

At Microsoft, we’re committed to continuing cutting-edge research and proactive development to keep Copilot Studio language models and customers secure. It’s imperative that customers are protected from attacks that target agents to perform malicious actions that the’re not configured to automate. Based on state-of-the-art research, we recently rolled out a feature that adds more protection against—and better detection of—XPIAs, further reducing the risk of a successful attack. This feature is enabled by default everywhere and cannot be disabled.

What does this mean for Copilot Studio customers? It means you’ll be able to meet more stringent security requirements related to security protections, and that your data and operations are safeguarded against evolving threats. This proactive approach helps ensure that your organization remains compliant with industry standards and best practices for cybersecurity.

Microsoft 365 Agents Software Development Kit for JavaScript

The Microsoft 365 Agents Software Development Kit (SDK) is a powerful tool that provides developers with building blocks to create agents that handle user interactions, orchestrate requests, generate responses based on enterprise knowledge, and collaborate with other agents. In November 2024, we announced that the Microsoft 365 Agents SDK would support C# development. Now, developers can also use JavaScript to create agents.

With the rollout of this feature, developers can take advantage of their existing skills and knowledge in JavaScript to build sophisticated agents that can enhance productivity and streamline business processes. The Microsoft 365 Agents SDK provides interoperability with Copilot Studio, allowing developers to extend existing agents built using Copilot Studio with additional functionality.

For all the information and documentation, check out the Agents for JavaScript repository on GitHub. You’ll find samples, templates, test bots, and plenty of other resources to get started.

Embedded Microsoft 365 agent builder updates

We’re excited to announce two powerful improvements to the embedded agent builder in Microsoft 365 Copilot, powered by Copilot Studio: Public web scoping and end user feedback.

Makers can now ground their Copilot agents on scoped publicly available websites, and leverage these websites as knowledge. Rather than letting agents perform a full web search, makers can constrain their agents’ external knowledge sources to specific websites for the most relevant answers. This capability already existed in custom agents built from Copilot Studio, but it now extends to agents built in Microsoft 365. This feature is enabled automatically under the “Configure” tab. In the “Knowledge” section, simply enter the URLs of approved public websites and your agents will use those as their knowledge base.

This new feature gives makers greater control over the information that Copilot agents access, and helps ensure responses are more accurate and scoped to the right context. By specifying trusted websites, makers can enhance the reliability of the agents’ answers and provide end-users with more consistent responses. Additionally, this capability helps in maintaining the integrity of the information provided by the agents, by limiting the sources to those that have been vetted and approved by the makers.

The public web scoping feature is particularly useful for organizations that have specific knowledge bases or industry-specific websites they rely on. By integrating these sources, agents can deliver more specialized and contextually appropriate responses, which can be crucial for customer support, technical assistance, and other use cases where more precise information is necessary. This increased function can even help reduce the frequency of escalations to human support.

Makers can also now submit feedback—compliments, problems, or suggestions—directly to the product team from inside the Microsoft 365 agent builder. Just as end users can rate an AI-generated agent response with a thumbs-up or thumbs-down, makers can now do so in preview chat. You can also use the “Send Feedback” option to send feedback at any time while using agent builder, and there’s a specific feedback option if a responsible AI (RAI) error pops up. When submitting feedback, you can choose to write a comment and, optionally, allow the product team to collect supporting metadata like AI Prompt, AI Response, and Conversation ID. (You can review this data before clicking send.)  

This feature is exciting because makers now have a way to share feedback without breaking their flow during the agent authoring process, and with relevant surrounding context. This helps the product team react quickly to feedback, without needing to go back and forth to get additional troubleshooting information, and gets you back on track as soon as possible.

Twelve new languages supported in conversational agent creation experience 

Meillä on ilo ilmoittaa…

Miło nam ogłosić…

เรามีความยินดีที่จะประกาศ…

We’re delighted to announce the addition of twelve new languages to the conversational agent creation experience in Copilot Studio and the Microsoft 365 agent builder. The new supported languages include Chinese (traditional), Czech, Finnish, Hindi, Indonesian, Korean, Norwegian Bokmål, Polish, Russian, Swedish, Thai, and Turkish.

This feature enables expanded language support in areas where the Microsoft 365 agent builder was recently launched. It’s auto-enabled for makers whose browser language is one of the supported languages and whose environment is in a Microsoft Azure OpenAI Service supported region. The language of the conversational agent creation experience will default to the browser language if the specific language is supported in this new update. This means that thousands of makers can create new agents in their native language, without further configuration, and serve their local customers natively.

Read more about language support and use in Copilot Studio here.

More ways to stay up to date on all things Copilot Studio

Bookmark this page to learn about features as we ship them, as well as upcoming releases: What’s new in Microsoft Copilot Studio – Microsoft Copilot Studio | Microsoft Learn.

Get started today

To learn more about Copilot Studio and how it can transform your organization’s productivity, visit the Copilot Studio website or sign up for a free trial today.

The post What’s new in Copilot Studio: February 2025 appeared first on Microsoft Copilot Blog.

For this edition, we’ll talk about:

• Copilot Chat recap
• PAYGO meter for Microsoft Copilot Studio
• Enhanced Answers with Semantic Search
• Knowledge tuning public preview
• Image upload
• Security data loss prevention (DLP) updates
• Arabic (ar-SA) language support

Copilot Studio in Microsoft 365 Copilot Chat

As a part of our continued mission to empower every person with a Copilot and transform their business with agents, we announced Microsoft 365 Copilot Chat, a free and secure AI chat experience powered by GPT-4. With Copilot Chat, your entire organization can start using Copilot and agents today. These agents are easy to create directly in Copilot, highly discoverable, and can be published to Copilot Chat. This enables org-wide agents to be used in a common UI for all employees, regardless of Microsoft 365 Copilot licensing, making agents more accessible for all.

For example, a team could build a Budget Agent that is equipped with knowledge about the team and helps submit expense reports, provide financial insights, and check policy compliance—and everyone on the team could use this agent through Copilot Chat, even if they don’t have an individual Copilot license.

This experience is designed to be enterprise-grade, providing security, privacy, and compliance features for all users. Additionally, IT controls are available to manage and monitor the usage within the organization.

Read the full announcement of agents experienced in Copilot Chat.

Shared consumption meter for Copilot Studio

In December 2024, we introduced consumption-based pricing for agents built in Copilot Studio. This new billing model offers greater flexibility and scalability for organizations looking to easily use natural language to build agents, without committing to upfront license purchases.

The PAYGO meter allows customers to pay only for the messages they consume, providing a cost-effective alternative to the traditional prepaid message packs. This model is designed to accommodate varying usage patterns, making it ideal for businesses with fluctuating demands.

Under the PAYGO model, customers are billed $0.01 per message. This includes both classic (non-generative AI) and generative answers. Agent responses generated from grounded knowledge, outside of the Microsoft Graph, will cost two messages. The consumption rate for autonomous agents and agents with generative orchestration will scale based on the complexity of the task completed. The PAYGO meter uses your Azure subscription as the payment instrument, ensuring seamless integration with existing billing processes. This meter is available in regions where Copilot Studio’s generative AI capabilities are also available.

Following is a list of Copilot Studio events and their respective billing rates:

• Classic answer: 1 message
• Generative answer: 2 messages
• Autonomous action: 25 messages
• Tenant Microsoft Graph grounding for messages: 30 messages

By paying only for the messages consumed, organizations can optimize their spending and avoid overpaying for unused capacity. There are no in-product feature differences between the PAYGO meter and the prepaid message packs. Customers can have the same capabilities and benefits, regardless of the billing model they choose. However, if the purchased capacity of message packs is reached with the pre-paid packs, makers will be restricted from creating, editing, or publishing agents for the duration of the month in which capacity was reached.

We recommend signing up for the meter even if you purchase message packs to handle these scenarios without disruption. The meter won’t count utilization in any month until purchased message packs have been fully exhausted. If capacity is reached, agents that are already published will remain functional without interruption, but updates to the agent cannot be published during that time. Learn more about message allocation.

Setting up the PAYGO meter for Copilot Studio is a straightforward process. Administrators can manage their message capacity and monitor consumption through the Microsoft Power Platform admin center (PPAC). The capacity management features provide detailed, daily usage data, helping organizations plan and budget effectively.

See more information on setting up consumption billing.

Improved SharePoint responses with semantic search

We’re excited to announce enhanced search capabilities in Copilot Studio, which enables semantic search over your SharePoint knowledge.

By enabling the new “Enhanced search” results, your agent can leverage the semantic index for Copilot feature to semantically index the data you have stored in SharePoint and yield more relevant search results. You can easily control the “Enhanced search” results setting from the “Generative AI” settings page.

This feature is turned on by default, so you can get immediate benefits without additional setup. Tenants with Microsoft 365 Copilot can create custom agents and leverage the same index used by Microsoft 365 Copilot. These same tenants can also use SharePoint and Microsoft Graph connectors containing files up to 200 MB, resulting in answers that are more relevant for you.

To benefit from this more robust indexing capability, agents must share a tenant with a Microsoft 365 Copilot license. Learn more about billing for tenant Graph grounding for messages.

Learn more about enhanced search.

Knowledge tuning in public preview

The ability to make data-driven decisions and fine-tune your knowledge accordingly is now available to all in public preview. Adding knowledge to agents equips them with your enterprise-specific data and enables agents to generate responses grounded on that information. However, agents can still face challenges in identifying the correct data. By leveraging these new advancements for knowledge tuning and knowledge analytics, makers can get recommendations and insights into how their knowledge sources are utilized by their agents and ensure their knowledge is more robust and effective.

The “Knowledge” page now includes relevant analytics such as knowledge usage, errors, and answer rates. Makers will also have access to suggestions for how to improve their knowledge, including Microsoft Dataverse tables, SharePoint libraries, or public websites that may be relevant to their agents.

Additionally, makers can designate verified and trusted knowledge sources as official sources so that these specific sources are prioritized and highlighted in responses provided by the agent.

We hope that these improvements will help makers derive even greater value from their data sources. We’re not stopping here though, additional capabilities coming soon include further recommendations on relevant Microsoft Graph files or SharePoint libraries, and the ability to define custom instructions for connectors or SharePoint data sources.

Learn more about the advanced knowledge tuning capabilities.

Let agents analyze images for you

Today, employees and customers who chat with agents for support are limited to sending and receiving messages only. However, many scenarios require sending an image for the agent to analyze and respond appropriately. This capability is now available in public preview.

When end-users upload images to the chat, agents can seamlessly analyze both text and visuals, ensuring a smooth conversation flow and expanding the range of supported scenarios. Whether you’re a field technician attaching a photo of a faulty part for troubleshooting or an analyst sharing a chart for insights on emerging trends, this feature enhances how agents can support your needs. Users can upload a wide range of images, including products, blueprints, maps, and charts. 

To try out this feature, head to Copilot Studio and enable “Image input” under the “Generative AI” section of the agent’s settings. Keep in mind that enabling end-user image uploads requires “Generative mode” to be turned on. Once enabled, see how your agent can analyze and respond to the images you send.

Learn more about image input from users.

Security and data loss prevention (DLP) updates

We believe it’s critical to secure agents built in Copilot Studio and are always investing in ways to improve security and governance features in this pursuit.  Currently, DLP enforcement is disabled by default. The process to enable and enforce these DLP policies is a two-step process, resulting in many agents published without sufficient DLP configured.  

Starting January 2025, the default mode for all tenants has been changed to be Soft-Enabled. The Soft-Enabled DLP state allows published agents to continue to run without enforcing any DLP policies. However, it blocks any further updates to agents that are affected by DLP blocks. In addition, the option to toggle to Disabled mode in the PowerShell module will be disabled. 

Furthermore, in February 2025 the default mode for all tenants will be changed to Enabled. Enabled mode enforces DLP checks completely for running agents and updates to existing agents. All published agents and updates will be subject to the DLP policies that apply as defined by the tenant.

These changes will help ensure all agents built and published through Copilot Studio are safe and secure, and follow the DLP policies defined by the tenant.  

Learn more about DLP and other best practices.

Arabic support is now available

We now have 27 languages in Copilot Studio, with the addition of Arabic. This exciting update brings enhanced accessibility and usability for our Arabic-speaking users.

Users can now chat in Arabic to ask questions or seek support, with responses also generated in Arabic. Agents grounded in knowledge can provide accurate and relevant answers in Arabic, ensuring seamless communication. Additionally, Arabic is available for interactive voice response agents, making voice interactions more inclusive and effective. 

However, there are some important limitations. Arabic is not currently available in the UI of the authoring canvas or maker experiences within Copilot Studio. Additionally, Arabic is not yet supported for “Generative mode,” which enables agents to leverage large language models to dynamically select the best combination of topics, actions, and knowledge sources during conversations.

We’re committed to continuously improving our platform and expanding language support to better serve our user base. Stay tuned for more updates and enhancements.

Learn more about all the supported languages.

More ways to stay up to date on all things Copilot Studio

Bookmark this page to learn about features as we ship them, as well as upcoming releases: What’s new in Microsoft Copilot Studio – Microsoft Copilot Studio | Microsoft Learn.

Watch monthly episodes highlighting Copilot Studio news and feature deep dives in “This Month in Copilot Studio” with Dona Sarkar, Microsoft Principal Cloud Advocate, and Kendra Springer, Principal Group Product Manager with Copilot Studio: This Month in Copilot Studio.

Get started today

To learn more about Microsoft Copilot Studio and how it can transform your organization’s productivity, visit the Copilot Studio website or sign up for our free trial today.

New Copilot Studio documentation updates

• Language support
• Manage message capacity
• Disable or limit sharing of agents
• Allow image input from users, and image analysis (preview)

The post What’s new in Copilot Studio: January 2025 appeared first on Microsoft Copilot Blog.