Theme
Feedback Reports
The Feedback Reports dashboard aggregates agent feedback on suggestions to give you a clear picture of how well Agent Assist is performing. Use these metrics to identify weak spots in your knowledge bases and guide content improvements.
Feedback analytics dashboard with acceptance rate chart, rejected topics breakdown, and unanswered query list
Key Metrics
Answer Acceptance Rate
The percentage of suggestions that agents accepted (thumbs up or used in a reply) versus rejected (thumbs down or ignored). This is the primary quality indicator for your assist deployment.
| Rating | Guidance |
|---|---|
| Above 70% | Strong performance; continue monitoring |
| 50% - 70% | Review rejected queries for content gaps |
| Below 50% | Investigate RAG settings and knowledge base coverage |
Source Helpfulness by KB
Breaks down acceptance rates by knowledge base. This helps identify which KBs are performing well and which need attention. A KB with a low helpfulness score may have outdated content, poor chunking, or insufficient coverage.
Rejected Query Topics
Groups rejected suggestions by topic to reveal patterns. If agents consistently reject answers about a specific subject, that topic likely has a content gap in your knowledge bases.
Response Time Trends
Tracks the end-to-end latency from customer message to suggestion delivery. Spikes in response time may indicate infrastructure issues or overly complex RAG configurations.
Unanswered Query Tracking
Lists customer questions that did not produce any suggestion because no knowledge base chunks met the similarity threshold. These unanswered queries represent direct opportunities to expand your knowledge base content.
How Feedback Improves RAG Quality
Agent feedback feeds directly into the retrieval pipeline through two mechanisms:
Chunk suppression: Chunks that are repeatedly associated with rejected suggestions have their relevance scores penalized. See RAG Settings for details on feedback-based chunk suppression.
Gap identification: Unanswered queries and rejected topics are surfaced in the dashboard so you can create new content to fill those gaps.
TIP
Export unanswered queries as a CSV from the dashboard and use them as a content creation checklist for your knowledge base authors.
Filtering
Filter feedback data by:
- Date range: View trends over custom time periods
- Knowledge base: Isolate performance for a specific KB
- Agent: Review individual agent feedback patterns
- Queue: Filter by CCaaS queue to see performance by team or topic area