Skip to main content
This lesson explains one of the core concepts in the Microsoft 365 Copilot ecosystem: agents. Microsoft 365 Copilot is a powerful AI assistant that helps users complete tasks and make decisions. Agents extend Copilot by becoming specialized AI applications that can automate work on behalf of users. Rather than only answering questions, agents can retrieve and surface information, execute business processes, and — at higher levels — make decisions and coordinate actions across systems. There are three broad categories of agents. Below we define each type, show practical examples, and highlight the primary considerations for adoption.

​
1) Retrieval agents (simple)

Retrieval agents focus on locating, organizing, and presenting information. Their main role is to gather grounding data that downstream reasoning, summarization, or person-facing answers can rely on. Examples:
  • An HR user asks, “What is our organization’s remote work policy?”
    The retrieval agent searches approved sources such as SharePoint, policy handbooks, and knowledge bases, then returns the most relevant policy excerpts and citations.
  • A sales rep asks, “Show me all interactions with customer XYZ over the last six months.”
    The agent aggregates emails, meeting notes, CRM records, and related documents, then summarizes the interactions.
Key point: retrieval agents primarily access and summarize information; they do not change systems or perform actions.

​
2) Task agents (intermediate)

Task agents carry out specific business processes when directed by a user. Think of them as digital workers that automate repeatable administrative tasks and workflows. Examples:
  • A manager says, “Create a new project workspace, schedule a kickoff meeting, and notify the project team.”
    A task agent can provision the workspace, add relevant members, create the calendar event, and send the invitation automatically.
  • An employee onboarding agent: when a new hire joins, the agent can create user accounts, assign Microsoft 365 licenses, add Teams memberships, schedule orientation sessions, and send welcome communications.
Key point: task agents act on user instructions within defined boundaries and deliver measurable automation and productivity improvements.

​
3) Autonomous agents (advanced)

Autonomous agents manage complex, multi-step processes with minimal human intervention. They plan activities, coordinate with other agents, monitor conditions, make recommendations, and escalate issues when necessary. Examples:
  • Supply chain: an autonomous agent continuously monitors inventory levels, supplier performance, shipping delays, and demand forecasts. If inventory drops below a threshold, it may identify alternate suppliers, create purchase requests, notify stakeholders, and track resolution.
  • Customer support: an autonomous agent triages incoming cases, gathers supporting information, coordinates with specialized agents, resolves routine issues autonomously, and escalates only high-priority cases to human staff.
Key point: autonomous agents function as intelligent process managers and orchestrators rather than single-task executors.

​
Progression and considerations

There is a natural progression from retrieval → task → autonomous agents:
  • Retrieval agents answer questions and surface evidence.
  • Task agents perform actions and automate defined workflows.
  • Autonomous agents manage processes end-to-end and make contextual decisions.
As you move from level one to level three, complexity and business value increase — but so do governance, security, and oversight requirements. Many organizations begin with retrieval agents, expand to task automation, and adopt autonomous agents selectively in high-value or well-governed scenarios. This staged approach allows a controlled, auditable shift from AI-assisted productivity to AI-driven automation.
As automation increases, include clear governance, audit trails, and human-in-the-loop checkpoints to manage risk and maintain compliance.
An infographic titled "Agents – An Overview" shows a central turquoise circle labeled "Agents" branching into three numbered colored nodes. Each node summarizes a type—Retrieval Agents (simple), Task Agents (intermediate), and Autonomous Agents (advanced)—with brief role descriptions.
This lesson also explores how Microsoft 365 Copilot and agents work together, and how organizations can use them to accelerate processes while maintaining governance and security. Links and references:

Watch Video