Theme
Gap Analysis
The Gap Analysis report surfaces queries that Agent Assist could not answer well. Use it to find missing or insufficient knowledge base content and prioritize content creation efforts.
Gap analysis table showing unanswered query topics, gap categories, frequency counts, and recommended actions
Identifying Knowledge Gaps
A knowledge gap occurs when a RAG query returns no results, returns low-relevance results, or produces a suggestion that the agent rejects. The report aggregates these cases and groups them by topic.
Unanswered Query Trends
A time-series chart tracks the volume of unanswered or poorly answered queries over the selected date range. A rising trend indicates growing coverage gaps that need attention.
| Metric | Description |
|---|---|
| No Results | Queries where the vector search returned zero matches above the relevance threshold. |
| Low Relevance | Queries where matches were returned but all scored below the confidence threshold. |
| Rejected Suggestions | Queries where the agent explicitly rejected the generated suggestion. |
Gap Categories
Gaps are grouped into categories to help you prioritize:
| Category | Description |
|---|---|
| Missing Topic | No knowledge base content exists for the query subject. |
| Outdated Content | Content exists but is stale or no longer accurate. |
| Insufficient Detail | Content exists but lacks the depth needed to answer the query. |
| Scope Mismatch | Content was retrieved from the wrong knowledge base or domain. |
TIP
Focus on the "Missing Topic" category first. These represent entirely uncovered areas where new content will have the greatest impact on acceptance rates.
Recommended Actions
For each gap category, the report suggests next steps:
- Create new KB content -- Draft articles or documents covering the missing topic. Upload them to the appropriate knowledge base.
- Update existing documents -- Revise outdated or incomplete content to address the identified shortcomings.
- Adjust KB scope -- Reconfigure knowledge base assignments or embedding parameters to reduce scope mismatches.
- Retrain classifier -- If gaps align with unclassified intents, add new intent definitions and training data.
Exporting Gap Reports
Export the gap analysis data for offline review or sharing with content authors:
- CSV -- Tabular export of all identified gaps with query samples, categories, and suggested actions.
- PDF -- Formatted report suitable for stakeholder review.
WARNING
Exported reports may contain query text from real conversations. Ensure that exports are handled according to your organization's data handling policies.
Advanced: Semantic Gap Analysis
For deeper gap analysis, administrators have access to an advanced view that uses AI-powered semantic clustering to group similar unanswered queries, auto-generate draft knowledge base articles from conversation transcripts, and review transcript excerpts for context. See the Administration Gap Analysis guide for details.