The development of AI agents has reached a mature stage, enabling businesses to automate more daily tasks, enhancing operational efficiency and customer experience.
In 2025 Release Wave 1, Microsoft introduced voice conversation capabilities, enhanced automation actions, and a more comprehensive analytics dashboard. These upgrades empower businesses to scale automation through AI while fine-tuning AI models based on data insights to better align with their operational needs.
Let’s dive into these key Copilot Studio upgrades and explore how they can drive digital transformation for enterprises!
1. Interactive Voice Response
📽 Demo Video:
Businesses can now enable AI Agents with voice capabilities, customizing multiple voice tones and styles to deliver more natural and personalized conversations. This feature also supports:
- Automatic Language Detection: If users speak in a language different from the default setting, the AI agent can instantly detect and switch to the user’s preferred language.
- Response Time Adjustment: Businesses can modify AI response wait times based on operational needs, ensuring an optimal customer experience across various scenarios.
2. Automated Actions: AI Agents Move from Assistants to Active Executors
Seamless Integration with Thousands of Applications—No API Required
Microsoft has introduced a new “Action Configuration Panel”, preloaded with thousands of connectors that allow AI Agents to seamlessly integrate with both internal and external enterprise systems, including:
✅ Microsoft Ecosystem: Dynamics 365, Power Automate, SharePoint, Excel, and more.
✅ Third-Party Software: SAP, Oracle, Salesforce, and even custom external data sources.
✅ Common Automation Actions: Data retrieval, Excel updates, data queries, workflow triggers, and more.
With these enhanced automation actions, businesses can unlock broader application scenarios when configuring AI Agents in Copilot Studio.
Updated Action Quota Calculation
As AI Agents take on increasingly diverse tasks, Microsoft has also updated the message quota calculation model, where different types of AI interactions will now consume quotas at varying rates.
| Copilot Studio feature | Billing rate | Explanation |
| Classic answer | 1 message | Manually written, fixed responses. |
| Generative answer | 2 messages | AI-generated answers that integrate conversation context and knowledge base content. |
| Autonomous action | 25 messages | Actions triggered by workflows (excluding knowledge searches, retrieval, or AI Builder prompts). |
| Tenant Microsoft Graph grounding for messages | 30 messages | High-quality retrieval-augmented generated responses based on Microsoft Graph. |
| Text and generative AI tools (basic) | 0.1 messages | Uses the 4o-mini model. Consumes 0.1 message quota per 1,000 tokens (Effective from April 7, 2025). |
| Text and generative AI tools (standard) | 1.5 messages | Uses the 4o model. Consumes 0.1 message quota per 1,000 tokens (Effective from April 7, 2025). |
| Text and generative AI tools (premium) | 10 messages | Uses the o1 model. Consumes 0.1 message quota per 1,000 tokens (Effective from April 7, 2025). |
Here, let’s illustrate it with a senario:
If your conversation flow consists of multiple steps, each step will consume the corresponding message quota. Here’s a detailed breakdown:
1️⃣ User asks for the weather on a specific date → The AI agent retrieves weather information from an online source using a connector. Consumes 25 message quotas.
2️⃣ User inquires about internal knowledge base information → The AI agent fetches data from knowledge sources & the organization’s Microsoft Graph, then compiles a response. Consumes 2 + 30 = 32 message quotas.
3️⃣ User asks about flight reservations → This triggers a “topic”, which the system counts as an automated action. Consumes 25 message quotas.
⏳ Quota Calculation Method: The quota resets on the 1st of each month, rather than based on the subscription activation date. Unused quotas do not carry over to the next month.
💡 Overage Usage Options: When the quota is exhausted, businesses can either redirect AI interactions to human agents or opt for a Pay-as-you-go model at $0.01 per message quota.
3. Upgraded Analytics Dashboard: Making AI Smarter with Data
The latest Copilot Studio now offers more comprehensive data analytics, enabling businesses to track AI Agent performance and optimize AI models based on insights.
Key Metrics: Aligning AI with User Needs
Copilot Studio now provides integrated analytical reports, giving visibility into various conversation metrics, including:
- Outcome & Engagement Analysis: A pie chart visualization categorizes conversation results into three key groups: Resolved, Escalated, and Abandoned.
Session Outcome Determination:
- Resolved: The session ends under one of these conditions:
- A success-confirmed topic is triggered.
- The conversation ends, and the user confirms a successful interaction or lets the session time out.
- Escalated to Human: The session ends after an AI-handled transfer to a live agent (even if the human agent does not respond).
- Abandoned: The conversation times out without being resolved or escalated.
- Knowledge Source Utilization: A line chart visualization shows the frequency and proportion of different knowledge sources being used, helping businesses understand which information is most relevant to users.
- User Feedback: Aggregated user satisfaction scores provide a high-level view of AI performance, ensuring continuous learning and optimization.
Conclusion: How Can Businesses Leverage Copilot Studio for Competitive Advantage?
In 2025, AI Agents are no longer just assistants—they are the driving force behind enterprise automation. With voice interactions, advanced automation, and enhanced data analytics, Copilot Studio empowers businesses to create smarter, more efficient AI experiences.
🚀 Get Started with Copilot Studio Today!
🔹 Try it now: Microsoft Copilot Studio
🔹 Contact us for a free consultation on custom AI solutions!
Reference:
Use actions with custom agents (preview) – Microsoft Copilot Studio | Microsoft Learn
Manage message capacity – Microsoft Copilot Studio | Microsoft Learn
Review and improve agent effectiveness – Microsoft Copilot Studio | Microsoft Learn