Explore the benefits of using agents
Agents help users work with a focused set of information through a conversational interface. Their value depends on well-managed source content, clear instructions, and a scenario suited to the agent's capabilities.
Benefits of knowledge-focused agents
Agents created in SharePoint or with Agent Builder can support common information tasks:
- Focused retrieval. An agent searches its configured knowledge sources instead of requiring the user to browse several sites or folders.
- Summarization. An agent condenses relevant source content and provides citations that users can review.
- Consistent guidance. Instructions define the agent's purpose and influence how it formats and scopes responses.
- Knowledge sharing. Teams can use a shared agent to ask questions against a common set of approved sources.
- Permission-aware responses. Each user receives information only from Microsoft 365 content that they can access.
Knowledge-focused agents respond when a user prompts them. They don't independently monitor files, assign tasks, update metadata, or send reminders. Scenarios that require those operations need actions, workflows, or a custom engine agent.
Use agents in SharePoint
A SharePoint site's ready-made agent helps users ask questions about site content. Users with the required access can also create an agent from selected sites, folders, or files to narrow its scope for a particular project or audience.
Consider an onboarding team that stores policies, role guides, and frequently asked questions in SharePoint. The team can create an agent grounded in those approved sources. New employees can ask questions in natural language, review cited sources, and follow the current guidance without searching each library separately.
SharePoint stores the agent as an .agent file on the site. The team can manage access to that file with SharePoint permissions and update the agent's instructions or knowledge sources as its guidance changes.
Use agents in Microsoft Copilot Chat
In Microsoft Copilot Chat, users can select available agents for specialized conversations. For example, a project team can use an agent grounded in approved project documentation to summarize decisions, compare requirements, or prepare a status overview.
Declarative agents can include actions when their authoring tool and configuration support them. More complex automation, proactive behavior, and custom orchestration are advanced scenarios that require Copilot Studio or pro-code development.
Evaluate an agent scenario
Before creating an agent, confirm that:
- The knowledge sources are current, authoritative, and appropriately permissioned.
- The agent's purpose is specific enough to guide useful responses.
- Users understand that AI-generated output can be incomplete or incorrect and should verify important information against cited sources.
- The selected authoring tool supports any required actions or integrations.