THE KNOWLEDGE AGENT

The Knowledge Agent in SharePoint is a collection of AI-powered features integrated into Microsoft SharePoint Online.

  • It serves as an intelligent content curator designed to enrich, organize, structure, and maintain content, making it more effective for Microsoft 365 Copilot and other agents.

  • By addressing challenges such as unstructured data, inconsistent metadata, outdated content, and discoverability issues, it transforms SharePoint into an AI-ready knowledge hub.

Key Features of the Knowledge Agent

  • The Knowledge Agent provides role-based tools accessible via a floating button (typically with the teal SharePoint logo) in SharePoint sites and libraries.
  • Features vary by context (e.g., pages vs. document libraries) and user permissions / roles, which include Content Consumer, Content Creator, Content Manager, and Site Manager.
    • Ask a Question / Get Answers and Insights: Users query site or library content in natural language. The agent delivers grounded responses, summaries, insights, file comparisons, or audio overviews, leveraging enriched metadata for precision.
    • Organize This Library: Analyzes document libraries and suggests metadata columns (e.g., based on content analysis). It supports autofill columns for automatic extraction and population of metadata. This improves AI reasoning, search, and discoverability. Users preview suggestions before applying changes.
    • Set Up Rules: Create automations using natural language (e.g., "Notify me when invoices over $500 are added"). This simplifies workflows without requiring technical expertise.
    • Create New View: Generates custom views by sorting, filtering, or grouping content based on metadata (e.g., policies expiring soon or contracts by client).
    • Improve This Site: Helps site owners maintain freshness by identifying and retiring inactive pages, detecting broken links (with redirect options within the tenant), surfacing content gaps, and providing recommendations. Inactivity thresholds can be configured per site.
    • Create Page or News Post: Assists with content creation using templates, layout suggestions, and co-authoring capabilities. Users can upload documents for AI-assisted refinement.
    • Additional Capabilities: Includes "See More Agents" for expanded AI actions, metadata hygiene suggestions, policy alignment, and role-tailored recommendations. It enhances Copilot responses by providing better context through structured data.
  • These features reduce manual effort, improve content governance, and ensure AI tools deliver accurate, trustworthy outputs.

How SharePoint End Users Can Use the Knowledge Agent

  • End users interact with the agent directly in their SharePoint environment without additional setup:
    • 1. Navigate to a SharePoint site or document library where the agent is enabled.
    • 2. Locate the floating button (bottom-right corner, teal SharePoint logo).
    • 3. Click to open the agent interface, which presents context-aware options based on location and role.
    • 4. Select a feature (e.g., "Ask a question," "Organize this library," or "Improve this site").
    • 5. Provide natural language inputs where prompted. Review previews or suggestions before confirming actions.
    • 6. For content queries, receive AI-generated responses grounded in site data.

Best Practices for Users

  • Site owners/managers should regularly use "Improve This Site" to maintain freshness.
  • Content creators benefit from metadata organization and page creation tools.
  • All users gain from quick Q&A and custom views for efficient information retrieval.
  • Verify AI-suggested metadata and actions for accuracy, especially in governed environments.
  • Combine with Copilot for enhanced productivity.
  • The agent respects existing SharePoint permissions and focuses suggestions accordingly.

Additional Considerations

  • Licensing and Availability: Included with Microsoft 365 Copilot during preview; confirm post-GA details with Microsoft.
  • Governance: Admins should review permissions, monitor AI usage, and align with organizational policies. Features like broken link redirects and page retirement help enforce hygiene.
  • Limitations: Preview status implies potential changes; not all languages or environments (e.g., certain government clouds) are fully supported. Performance depends on content quality.
  • Future Enhancements: Microsoft continues to evolve the agent based on feedback, with deeper integration for agents and Copilot.

Here is a curated list of effective prompts for the Knowledge Agent in SharePoint Online.

These leverage its capabilities for querying content, organizing libraries, creating workflows, generating insights, and supporting content management. The agent excels with natural language inputs that reference specific documents, metadata, policies, or site elements.

1. Content Discovery and Summarization

  •  "Summarize the key points from all policies in this library."
  •  "Provide an overview of the Digital Mega-300 video camera, including its technical features and project timeline."
  •  "What are the main updates in the 2024 annual report documents?"
  •  "Compare the contracts for Client A and Client B, highlighting differences in terms and expiration dates."
  •  "List all documents related to project X that were modified in the last 6 months."

2. Metadata and Library Organization

  • "Organize this library by suggesting metadata columns for document type, expiration date, and owner."
  • "Classify all documents in this folder as either 'Policy', 'Contract', or 'Report' based on content."
  • "Create AI-generated views to group contracts by client and sort policies by expiration date."
  • "Auto-fill metadata for resumes in this library, extracting skills, experience, and education."
  • "Suggest improvements to metadata hygiene for files missing key tags."

3. Automation and Rules (Workflows)

  • "Email me when new invoices over $500 are added to this library."
  • "Set up a notification rule for documents tagged as 'High Priority'."
  • "Create a workflow to route approval requests for contracts exceeding $10,000."
  • "Notify the team when a file is added with the keyword 'urgent'."
  • "Automate archiving for documents older than 2 years."

4. Site and Content Maintenance

  • "Improve this site by identifying broken links and outdated pages."
  • "Suggest content gaps in our HR knowledge base and recommend new pages."
  • "Retire or archive pages that have not been updated in over a year."
  • "Generate a FAQ section based on the documents in this library."
  • "Fix or flag any inconsistent formatting across policy documents."

5. Content Creation and Co-Authoring

  • "Create a new news post summarizing the latest team achievements using the provided template."
  • "Help draft a page on company benefits with sections for health, leave, and retirement."
  • "Generate an audio overview of this document."
  • "Build a comparison table for our product versions based on existing specs."
  • "Act as a co-author: refine this draft with professional tone and add relevant sections."

6. Advanced Insights and Analysis

  • "Which policies expire next year?"
  • "Provide insights on trends in our contract renewals over the past three years."
  • "Summarize the site content and highlight the most accessed documents."
  • "What changes were made in the latest version of the employee handbook?"
  • "Analyze risks or compliance issues in the uploaded legal files."

Best Practices for Prompting

  • Be specific: Reference libraries, folders, document types, dates, or metadata (e.g., "in the Contracts library" or "documents with status = Approved").
  • Use natural language: The agent handles conversational requests well but benefits from context about scope or desired output format (e.g., "list", "summary", "table").
  • Iterate: Follow up on responses with refinements like "Explain that in more detail" or "Filter to Q3 only."
  • Combine actions: For complex needs, chain prompts, such as summarizing then creating a view or rule.

These prompts are drawn from Microsoft documentation and community examples. Tailor them to your site's structure and content for optimal results. Test in your environment, as capabilities may evolve with updates to the Knowledge Agent. If you provide more details about your site's focus (e.g., HR, legal, projects), I can refine this list further.