In our context, an AI agent is a software-based assistant that uses AI to automate and execute business processes, either alongside or on behalf of a person, team, or entire organization. Instead of being a simple chatbot, an agent can: - **Retrieve and reason over information** from your business data - **Summarize and answer questions** for employees or customers - **Take actions in your systems** (like updating records or creating files) - **Automate workflows** that would otherwise be repetitive manual tasks Agents can be tailored to different roles and scenarios, such as: - **IT Helpdesk agent** – answers questions like “How do I connect to the corporate network?” using your IT knowledge base. - **Budget Manager agent** – reviews outstanding open purchase orders (POs) and helps with financial planning. - **Customer Support agent** – identifies new support issues from CRM data and routes or triages them to the right agents. - **HR agent** – supports HR processes when a new hire starts, using HR systems and data. - **Returns or Refunds agent** – handles refund and replacement requests, checks internal systems, and can escalate approvals (for example, when a refund is over a certain amount). These agents are built and managed on a secure, governed platform (Microsoft Copilot Studio) that connects to your business data (Microsoft Dataverse, Microsoft Fabric, SharePoint, CRM, etc.) and productivity tools (Microsoft 365, Microsoft Graph). This lets you reimagine how routine work gets done without adding infrastructure overhead.

You can design agents with different levels of sophistication depending on your business need. Broadly, there are four capability levels: 1. **Simple Retrieval agents** These agents focus on answering questions by pulling from your grounding data. - Retrieve information from internal sources - Reason over that information - Summarize and answer user questions **Example:** An IT Helpdesk agent that answers “How do I connect to the corporate network?” using your IT documentation. 2. **Enhance Your Current Experience** These agents go beyond Q&A and can take actions when asked. - Automate workflows - Replace repetitive tasks for users **Examples:** - A **Budget Manager agent** that reviews open POs and helps kick off financial planning. - An **Operations agent** that updates CRM records or retrieves order history when requested. 3. **Trigger-Based Autonomous agents** These agents react automatically to specific events or triggers. - Operate when a defined event occurs (e.g., “when a record is modified”) - Dynamically plan and orchestrate other agents - Learn from interactions and escalate when needed **Examples:** - A **Customer Support agent** that detects new support issues when a CRM record is modified and triages them to other agents. - An agent that acts **when an item is created in SharePoint** or **when a new hire starts** in HR systems. 4. **Instruction-Following agents for new experiences** These agents follow detailed instructions you define to handle more complex scenarios. - You specify when they should act and what steps to follow - They can combine search, reasoning, and actions across systems **Example: Customer service / returns agent** Instructions might include: - Help customers with product questions, refunds, and replacements. - If a customer is struggling with a product and asks for a refund, proactively provide relevant product guides and information first. - Search for customer information in internal systems. - For returns and replacements, perform checks on eligibility and approvals (e.g., “I have received a refund request that is over 200 and requires your approval.”). Across all levels, you can: - **Tell agents when to act** (e.g., new email in Outlook, new item in SharePoint, new hire in HR, record change in CRM). - **Define their behavior** (instructions, escalation rules, orchestration with other agents). - **Use pre-built templates** (e.g., get CRM record, update CRM record, get order history, create file, send a message in Teams) and a visual canvas to design flows. This range of capabilities lets you gradually reshape existing processes or build entirely new experiences, without having to start from scratch each time.

Agents are designed to work directly with your existing data and applications, and to show up in the channels your employees already use. **1. Connecting to your data and systems** Agents use data connectors and platform integrations to access: - **Business data** via Microsoft Dataverse - **Analytics data** via Microsoft Fabric - **Productivity data** via Microsoft Graph and Microsoft 365 (M365) - **Content and records** in systems like SharePoint, CRM, and HR tools - **External systems** through available connectors and ISV integrations This allows an agent to, for example: - Get or update a CRM record - Retrieve order history - Create a file - Look up customer information in internal systems **2. Building and orchestrating agents** You use **Microsoft Copilot Studio** as the main platform to: - Design agent behavior using a **visual canvas** - Use **pre-built templates** for common actions - Orchestrate multiple agents (e.g., a Customer Support agent handing off to a Returns agent) - Define instructions and triggers (e.g., when a new email arrives in Outlook, when an item is created in SharePoint, or when a new hire starts in HR) Developers can also integrate agents with **Azure AI** to extend capabilities while still using the same platform. **3. Deploying agents to user channels** Once built, agents can be deployed to: - **Microsoft 365 Copilot in Business Chat**, so end users can interact with them in the flow of work - **Microsoft Teams**, where agents can send and receive messages - Other channels accessible via ISV integrations with Copilot Studio **4. Governance, security, and lifecycle** The platform is designed as a secure, trusted environment: - No infrastructure management required - Granular governance controls over who can build and deploy agents - Enterprise data protection for your business data - Streamlined lifecycle management so you can version, update, and retire agents as needed In practice, this means your teams can start small—such as an IT Helpdesk or simple Returns agent—and then gradually build a network of agents that work across departments like Finance, HR, Operations, and Customer Support, all within your existing Microsoft ecosystem.