When to Use Knowledge Base
Knowledge base works best for information that already exists in written form and does not require back-and-forth conversation.| Scenario | Why knowledge base works |
|---|---|
| Company policies | Agent quotes directly from official documentation |
| FAQs | Common questions get consistent, accurate answers |
| Reference material | Technical specs, contact lists, and lookup tables |
| Onboarding info | Benefits, office locations, and team structures |
How Content Gets Into the Knowledge Base
Connected documentation sources
The most common path is to connect a supported app under Knowledge Base, browse folders or spaces, and choose what to include. Serval syncs on a schedule (and you can trigger a manual sync), indexes the text, and keeps documents up to date when the source changes. Examples of sources you can connect include Notion, Confluence, Google Drive, Microsoft 365 (SharePoint and files via Microsoft Graph), ServiceNow, Zendesk, and Freshservice. The exact catalog can grow over time—see the Integrations overview for apps your team can connect.Custom knowledge ingestion workflows
When you need content from an API, an internal wiki, or a system without a first-party connector, workflow authors can implement knowledge ingestion workflows. Those workflows run during knowledge sync for a connected application (often a Custom App), fetch or transform source data into documents, and return markdown text plus stable IDs and citation URLs. Serval chunks and embeds that content like any other knowledge source. Use native connectors when they exist; use custom knowledge ingestion when you control the API or need logic that a packaged integration does not provide. Technical details and when to choose this path are documented under Knowledge ingestion workflows in Workflow types.Setting Up and Testing Your Knowledge Base
Choose how to bring content in
Option A — Connected source: In the sidebar, open Knowledge Base and connect a supported documentation or help-center integration.Option B — Custom ingestion: Connect the relevant app (for example a Custom App), publish a knowledge ingestion workflow that implements your sync logic, and complete any setup your team agreed on so sync can run. See Workflow types.You can use both approaches together: multiple sources and workflows can feed the same knowledge base.
Select content (connected sources)
For first-party connectors, browse your source’s folder or space structure and choose what to include:
- Select entire folders or individual files
- Exclude drafts, internal notes, or outdated content
- Changes in your source sync to Serval according to the integration’s sync behavior
| Source | What you can sync (examples) |
|---|---|
| Notion | Pages and databases |
| Confluence | Spaces and pages |
| Google Drive | Docs, Sheets, and PDFs |
| Microsoft 365 | SharePoint sites, libraries, files |
| ServiceNow | Help center articles |
| Zendesk | Help center articles |
| Freshservice | Knowledge articles |
Only sync documentation you want the help desk agent to reference. For custom ingestion, scope and filters are defined in the workflow and app configuration instead of this UI.
Test with sample questions
Before going live, verify the agent finds the right answers:
- Open a test ticket or use the Serval chat
- Ask questions your users would ask (for example, “What is our PTO policy?”)
- Confirm the agent returns accurate answers with links to the source document
- Adjust synced content, workflow logic, or filters if answers are missing or incorrect
Managing Knowledge Sources from the Sidebar
The Knowledge Base section in the sidebar gives you a centralized view of connected knowledge sources. From here you can:- View all sources — See native integrations and sources backed by custom ingestion in one place
- Filter by sync status — Find sources that are syncing, up to date, or in an error state
- Trigger manual syncs — Run an immediate sync instead of waiting for the next scheduled run
Permission-Respecting Answers
Serval respects document permissions from connected knowledge sources when the integration supports propagating access controls. When a user asks a question, the help desk agent only returns answers from documents that user is allowed to access, so sensitive or restricted content is not surfaced to unauthorized users.Permission enforcement depends on the connected source supporting document-level permissions. Sources such as Google Drive, Confluence, SharePoint, and Notion propagate their access controls to Serval where supported.
Maintaining Your Knowledge Base
Toggle Visibility
Control whether each document is available to the help desk agent:| Status | Behavior |
|---|---|
| Visible | Agent searches this document and includes it in responses |
| Hidden | Document is synced but not used by the agent |
- Seasonal content outside of relevant periods
- Drafts under review that are not ready for users
- Deprecated policies you want to keep but not cite
Review Regularly
Outdated information erodes trust. Review knowledge base documents:- Quarterly for general accuracy
- Immediately after major policy changes
- When users report incorrect answers
Best Practices
Write clear titles
Help the agent match documents to questions. Use “Remote Work Policy 2025” not “Policy.”
Structure for search
Use headings that match how users ask. Put key info in the first paragraph of each section.
Include term variations
Add synonyms users might search: “PTO,” “paid time off,” and “vacation days” in the same doc.
One topic per doc
Focused documents perform better than 50-page handbooks. Split by policy or topic.
Keep content current
Review quarterly and after policy changes. Outdated answers erode trust.
Know what belongs elsewhere
Use workflows for automated actions, guidance for troubleshooting, access requests for provisioning.