Auto-Detected Signals

Automatic Signal Detection

Auto-detected signals are automatically identified by our AI system as your recordings are processed. The system analyzes the full transcript, understands context, and identifies signals across all 14 domains without any configuration required.

How It Works:

1. Recording Processing: When a recording is uploaded and transcribed, the signal detection engine automatically analyzes the content
2. Context Analysis: The AI understands context, not just keywords, to identify meaningful signals
3. Domain Classification: Signals are automatically categorized into the appropriate domain
4. Evidence Extraction: Relevant snippets from the transcript are captured as evidence
5. Confidence Scoring: Each signal receives a confidence score indicating detection accuracy

Signal Confidence Scores

Every auto-detected signal includes a confidence score that indicates how certain the system is about the detection:

• High Confidence (80-100%): Very likely to be accurate, strong evidence present
• Medium Confidence (50-79%): Likely accurate, but may require review
• Low Confidence (0-49%): Uncertain detection, should be reviewed carefully

Confidence scores help you prioritize which signals to review first. High-confidence signals can often be trusted without review, while lower-confidence signals may need human verification.

Signal Evidence

Each signal includes evidence snippets—exact quotes from the transcript that support the detection. Evidence snippets provide:

• Context: The surrounding conversation that led to the signal detection
• Verification: Direct quotes that you can verify against the recording
• Clarity: Understanding of what was actually said

Evidence snippets typically include 2-3 sentences of context before and after the key statement, giving you a complete picture of the conversation.

Signal Context

Beyond evidence snippets, signals include rich context information:

• Timestamp: Exact time in the recording when the signal was detected
• Speaker: Who said it (if speaker identification is available)
• Sentiment: The emotional tone (positive, negative, neutral)
• Related Signals: Other signals detected in the same conversation
• Recording Information: Which recording and meeting the signal came from

This context helps you understand not just what was detected, but when, who said it, and how it fits into the broader conversation.

Signal Review

While auto-detected signals are highly accurate, you may want to review them:

1. Confirm Signals: Mark signals as confirmed if they're accurate
2. Dismiss Signals: Remove signals that are false positives
3. Edit Signals: Modify signal details if needed
4. Add Notes: Add your own notes or context to signals

Reviewing signals helps improve the system's accuracy over time and ensures your signal results are reliable.