- Copilot Studio adds autonomous agents with deep reasoning and intelligent flows
- Improvements to SharePoint URL and data support, and agent editor in Copilot Chat
- New message-based usage billing model with detailed metrics
- Expanded features for customization, accessibility, and user management
Microsoft has launched multiple Major updates for Copilot Studio in March 2025, consolidating its proposal as an environment for the development of intelligent agents within the Power Platform ecosystem. This batch of changes not only introduces functional improvements for developers, but also redefines key aspects such as the consumption model, intelligent orchestration architecture, and integration with enterprise data sourcesHere we break down the main improvements and adjustments announced.
These new features reinforce the role of Copilot Studio in the development of business assistants who can act autonomously, adapt to multiple contexts, and facilitate rapid AI-powered decision-making. Below, we detail the most notable features added this month.
Autonomous agents and deep reasoning: the next level of conversational AI
One of the star additions is the general availability of self-employed agents.These agents combine advanced AI and automation to act without direct user intervention, responding to events and situations in real time.
Thanks to its integration with the deep reasoning -also know as deep reasoning–, officers can now address more complex situations. This model allows the agent to analyze multiple variables, infer relationships and make more sophisticated decisions. based on the context and configured instructions. Deep reasoning is dynamically activated depending on the needs of the moment, orchestrated by an intelligent internal system.
Some practical examples of these capabilities include:
- Generate complete business proposals (such as requests for proposals or RFPs)
- Detect risks and fraud in complex financial operations
- Optimize inventories or supply chains in the face of external changes
Autonomous agents can be configured with specific triggers (such as the arrival of a document or a data change), and from there, execute decisions without the need for direct interaction. Furthermore, it's possible to view their activity in detail, which offers traceability for IT and business teams.
Intelligent Agent Flows: Predefined and repeatable automation
Along with self-employed agents, Copilot Studio introduces so-called 'agent flows', available from March 31, 2025. These action sequences allow you to automate common processes that require a logical sequence of steps, but previously required manual intervention.
Agent flows can be designed using a graphical interface or natural language, similar to what Power Automate already offers. These flows are useful for repetitive tasks such as:
- Validate orders and send confirmations automatically
- Perform periodic regulatory checks
- Extract data from documents and feed into other systems
Combined with AI actions, flows extend Copilot Studio's reach beyond simple conversations, enabling complex tasks to be executed in a structured way.
Expanding data sources: Greater integration with SharePoint and the web
This month, Microsoft has significantly expanded support for knowledge sources to feed agentsFrom now on it is possible to:
- Add URLs to SharePoint sites, folders, or individual files as a knowledge base
- Limit an agent's web search to specific domains, improving the accuracy of responses
- Using external data connectors such as Salesforce Knowledge or Stack Overflow, now available through Graph Connectors
These options allow for more controlled and relevant experiences, especially in environments where knowledge must be based on well-defined corporate content.
Agent editor from Copilot Chat: direct and collaborative creation
Another relevant development is that now Agents can be created directly from the Copilot Chat interface, without needing to access the full editing environment. This makes agent creation more agile and accessible, even from mobile devices.
Thanks to this functionality, users can:
- Start an agent directly in Copilot Chat, defining basic instructions and data sources
- Reuse previous agents across multiple sessions
- Share and edit agents in collaboration with other colleagues
In addition, the editor compiles suggestions and comments which can be sent in real time to the product team, improving the feedback loop.
Message-based consumption model: new billing system
One of the highlights of the month has been the introduction of the billing model based on consumption, based on messagesInteractions are now counted in blocks called “messages,” with different rates depending on the event type.
For example, the following costs apply:
- 1 message per classic reply (defined manually)
- 2 messages per generative response produced by AI
- 5 messages per autonomous agent action how to run a flow
- Up to 100 messages for the use of premium generative AI tools
This allows organizations to adjust their expenditure based on actual usage. The option to link Copilot Studio environments to an Azure subscription has also been enabled to enable pay-as-you-go consumption without requiring upfront licensing.
To make management easier, the Power Platform admin center offers daily reports of message consumption by environment, and excess consumption rules are applied with a grace period before suspending functions.
Other improvements and additional features
In addition to the main new features mentioned above, the development team has included other useful features to improve the user experience:
- Support for shared agent editing through comments and metadata
- Improved mobile accessibility to use Copilot Chat on iOS and Android
- Support for new languages in the agent interfaces and responses
- Administration tools in the Microsoft 365 portal to control access to shared agents and review declarative metadata
March features strengthen Copilot Studio as a complete platform for developing advanced conversational agents. Its key strengths include autonomous agents with contextual reasoning, structured workflows integrated with AI, and the flexibility to connect to internal data sources. With a transparent consumption model and tools designed for enterprise environments, Microsoft provides one more step in its commitment to automate tasks through applied artificial intelligence.
I am a technology enthusiast who has turned his "geek" interests into a profession. I have spent more than 10 years of my life using cutting-edge technology and tinkering with all kinds of programs out of pure curiosity. Now I have specialized in computer technology and video games. This is because for more than 5 years I have been writing for various websites on technology and video games, creating articles that seek to give you the information you need in a language that is understandable to everyone.
If you have any questions, my knowledge ranges from everything related to the Windows operating system as well as Android for mobile phones. And my commitment is to you, I am always willing to spend a few minutes and help you resolve any questions you may have in this internet world.