AI Analysis
Understand how ScenIQ processes your meeting recordings to extract actionable insights.
Processing Pipeline
When you upload a meeting recording, ScenIQ runs it through a multi-stage AI pipeline:
Audio Extraction
The audio track is separated from the video file for processing.
Speaker Diarization
AI identifies distinct speakers and segments their speech.
Transcription
Speech is converted to text with speaker labels and timestamps.
Task Extraction
NLP models identify action items, owners, deadlines, and priorities.
What Gets Extracted
For each meeting, the AI attempts to identify and extract the following:
| Element | Details |
|---|---|
| Action Items | Tasks mentioned during the meeting with title, description, assigned owner, priority, and due date when available. |
| Key Decisions | Important decisions made during the meeting, surfaced in the Summary tab. |
| Key Insights | Notable discussion points, risks, or blockers identified by AI. |
| Evidence Spans | Exact transcript excerpts that support each extracted task, linking tasks back to the conversation. |
Confidence Scores
Each extracted task includes an AI confidence score (0–100%) indicating how certain the model is that this is a genuine action item. Tasks with lower confidence may need manual review.
Review AI output
Processing Time
Analysis time depends on the length and quality of the recording:
- Short meetings (under 15 min): ~1 minute
- Standard meetings (15–60 min): 1–3 minutes
- Long meetings (60+ min): 3–5 minutes
The meetings list auto-refreshes while processing is in progress. You can also check status from the Dashboard.
Error Handling
If processing fails, the meeting card will display a Failed badge with an error message. Common causes include:
- Corrupted or empty video files
- Unsupported codec within a supported container format
- No audible speech in the recording
You can dismiss the failed meeting and re-upload with a corrected file.