What are Gaps?
Gaps represent questions your AI couldn’t answer or actions it couldn’t perform. Thunder automatically detects these moments and categorizes them, helping you understand where your AI falls short. There are two types of gaps:| Type | Description | Example |
|---|---|---|
| Knowledge Gap | Information the AI doesn’t know but users expect | ”What’s your refund policy?” when the AI wasn’t trained on refund info |
| Tool Gap | Actions the AI can’t perform but users request | ”Can you book an appointment for me?” when the AI can’t access calendars |
Why Gaps Matter
Gaps are your roadmap for AI improvement. They tell you:- What knowledge to add - Specific topics users ask about that your AI can’t answer
- What tools to build - Capabilities users expect but don’t exist yet
- How often they occur - Which gaps affect the most users
- Recent trends - What users are asking about now
Gap Metrics
| Metric | Description |
|---|---|
sessions | Conversations where this gap was detected |
users | Unique users who encountered this gap |
messages | Number of times this gap occurred (a gap can appear multiple times per session) |
Querying Gaps
Most Common Gaps
Find your most frequent unanswered questions:Filter by Gap Type
Find only knowledge gaps or only tool gaps:Recent Gaps
Find gaps from the last 7 days to see what users are asking about now:Gap Trends Over Time
Track how gaps change week over week:Finding Sessions with Gaps
See the actual conversations where a gap occurred:Use Cases
Knowledge Base Priorities
Use knowledge gap frequency to decide what documentation or training data to add first.
Feature Roadmap
Tool gaps show what integrations and capabilities users want most.
Measure AI Improvement
Track gap frequency over time to verify that updates are working.
Topic Correlation
Cross-reference gaps with topics to understand which subject areas have the most missing information.
See Querying Metrics for more query patterns and the full field reference.