Microsoft 365 Copilot’s declarative agents give your team a friendly, tailored assistant inside the apps you already use—no coding required. Think of an agent as your department’s own branded helper in Teams, Word, or Excel that knows your company’s info, can run tasks for you, and guides people through common workflows. Here’s why you’ll love them, how they work in everyday language, and what you can do to get one up and running for your business.

What you actually get with a declarative agent

1. A built-in helper in Copilot
   • Your users don’t install a new app or learn a new screen—they just open Copilot and click your agent’s name. It shows up right alongside the regular Copilot suggestions, so it feels like part of the Microsoft 365 family they already know .
2. Your company’s knowledge at their fingertips
   • Instead of generic AI answers, the agent pulls from your own files and systems—SharePoint sites, internal wikis, or data you expose via Microsoft Graph connectors. That means policies, best practices, or product specs are always up-to-date and on-brand .
3. Hands-off automation
   • Want to let people reset passwords, check order statuses, or launch an HR onboarding workflow with a simple chat? Just wire in a plugin for your service desk, ERP, or HR system—then the agent can do it for them without a phone call or ticket .
4. Security and IT control
   • Everything an agent does lives inside your Microsoft 365 security perimeter. IT controls who can see or use each agent. You don’t open new holes in your firewall or give third parties broad data access .

Why your team will actually use it

• No extra login-pain: They’re already in Teams, Word, or Outlook. One click opens your agent—no new passwords, no extra browser tabs.
• Clear, guided conversations: You define the “starter questions” so people know exactly what to ask (“How do I onboard a new hire?” or “Show me last month’s sales report.”). That removes the guesswork and frustration.
• Faster answers, fewer interruptions: Instead of emailing IT or hunting through SharePoint, users get answers in seconds—right where they work.
• Consistent guidance: Every employee sees the same approved instructions. No more “Bob in IT said do it this way” vs. “Jane in HR said something else.”

Everyday scenarios

How it works

1. Package up your “who, what, and how”
   • You create a small zip file (an “app package”) that says:
     • Who your agent is (name, icon, permissions)
     • What it knows (connectors to SharePoint, Graph, or other data sources)
     • What actions it can take (prebuilt tasks or plugins for your systems)
     • How to start conversations (suggested prompts and tuning hints) .
2. Upload and assign
   • In the Microsoft 365 admin center, you upload that package and decide who gets to use it—your whole company, just Finance, or a pilot group of five people .
3. Responsible AI checks
   • Microsoft automatically vets your prompts and data usage against Responsible AI standards to make sure you’re handling privacy, bias, and security correctly. You get feedback if something needs tweaking .
4. People start chatting
   • End users open Copilot in Teams or Office, pick your agent, and type or click a starter question. Behind the scenes, Copilot routes the request through your agent’s instructions, data connectors, and any plugins you hooked up—and returns an answer or runs the action.

Getting started: your first agent in four steps

1. Pick your tool
   • Use Copilot Studio’s visual agent builder, the Teams Toolkit, or even hand-craft the files in your favorite IDE .
2. Define your knowledge
   • Point to the SharePoint sites, document libraries, or Graph-connected data you want the agent to draw on.
3. Add one action
   • Start simple: maybe a “Check order status” plugin call, or a “Create ticket” action in your ITSM tool.
4. Write clear prompts
   • Give people 3–5 example questions in the manifest so they know exactly how to interact.
5. Upload & test
   • Publish to a small test group, gather feedback, tweak your prompts or data sources, then roll out more broadly.

• 80% faster resolution on common help-desk requests when agents handle first-level support .
• 50% reduction in context switching for customer-service teams pulling data from multiple systems .
• 100% consistent answers on policy questions—no more outdated PDFs floating around .

Declarative agents let you give every team a smart assistant that feels built-in, knows your internal world, and frees people from repetitive tasks. You don’t write AI code—you just describe who the agent is, what it can do, and where to find the data. Then Copilot handles the rest: UI, security, model plumbing, and compliance checks .Ready to see what an agent can do for your organization? Here are some official resources for deep-dives and step-by-step guidance on Microsoft 365 Copilot declarative agents:

Graph Connectors for Enterprise Knowledge
https://learn.microsoft.com/microsoft-graph/connectors-overview

Overview & Concepts
https://learn.microsoft.com/microsoft-365-copilot/extensibility/overview-declarative-agent

Declarative Agents Manifest Reference
https://learn.microsoft.com/microsoft-365-copilot/extensibility/declarative-agent-manifest

Copilot Studio (Agent Builder) Quickstart
https://learn.microsoft.com/microsoft-365-copilot/extensibility/copilot-studio-agent-builder

Teams Toolkit for Declarative Agents
https://learn.microsoft.com/microsoft-365-copilot/extensibility/teams-toolkit-declarative-agents

Responsible AI Validation for Copilot Extensibility
https://learn.microsoft.com/microsoft-365-copilot/extensibility/responsible-ai

Plugin Integration Guide
https://learn.microsoft.com/microsoft-365-copilot/extensibility/plugin-manifest-reference