The AI Knowledge Agent generates answers directly from your knowledge base. The quality, structure, and freshness of your documentation determine how well the agent performs. Well-organized content with clear headings and concise information produces precise, reliable answers. Disorganized or outdated content leads to vague or incorrect responses.
Think of the agent as a new team member reading your wiki for the first time — the easier your docs are to understand, the better the agent serves your team.
Headings help the agent identify which sections are relevant to a question. Use specific headings that describe the content that follows.
Effective headings:
Less effective headings:
Break long pages into multiple sections with descriptive headings rather than writing large blocks of unstructured text.
Page titles are one of the strongest signals the agent uses to find relevant content. A title like "Password Reset Process" is far more useful than "IT Doc #47" or "Updated Procedure."
When naming pages:
The agent treats your wiki content as the current source of truth. Outdated pages can cause it to surface incorrect or obsolete information.
Deprecated pages that remain in your wiki compete with current information. The agent may cite an outdated page alongside a current one, creating confusion.
The agent performs best when content is organized around the questions people ask. Consider these formats:
Organize information as explicit questions and answers:
Q: How many vacation days do new employees receive?
A: New employees receive 15 vacation days per year, prorated based on start date.
Use numbered lists for processes:
How to submit a purchase request:
- Open the Procurement portal.
- Click New Request.
- Fill in the required fields.
- Attach quotes from at least two vendors.
- Submit for manager approval.
When documenting terminology, define terms clearly and concisely so the agent can provide accurate definitions.
For long or detailed pages, include a summary at the top that captures the key points. The agent often pulls from summaries when generating concise answers. A two-to-three sentence overview at the beginning of a page significantly improves answer relevance.
Kowiki generates AI metadata for your pages automatically. In cases where the auto-generated metadata does not capture the page's intent accurately, add or refine tags and summaries manually.
Well-chosen tags help the agent associate content with relevant topics, especially for pages where the title alone does not convey the full scope. Learn more in AI-Generated Metadata.
When the same information exists on multiple pages, the agent may produce answers that cite several sources saying the same thing — or worse, cite an outdated duplicate. Maintain a single source of truth for each topic and link to it from other pages rather than copying content.
Dense, concise paragraphs produce better answers than verbose, meandering text. The agent synthesizes content more accurately when the source material is clear and direct.
Pay attention to which questions the agent struggles with. These gaps indicate areas where your documentation needs improvement. Over time, refining content based on real usage patterns creates a feedback loop that continuously improves answer quality.