Analytics Overview
KoeIQ uses OpenAI to run four AI analytics on every transcript. Here is what each one produces and how to configure them.
The four AI analytics
| Type | Output | API endpoint |
|---|---|---|
| Summary | 2–3 sentence AI narrative of the call | POST /api/voicelogs/{id}/summary |
| Emotion | Overall sentiment (-1 to +1), dominant emotion label, per-speaker scores | POST /api/voicelogs/{id}/emotion |
| Intent | Primary call intent, urgency level, follow-up flag | POST /api/voicelogs/{id}/intent |
| Quality | 0–100 score, strengths, and improvement areas | POST /api/voicelogs/{id}/quality |
Auto-generate analytics
Enable auto-generate to run all four analytics automatically after every transcription completes.
- Go to Admin → Settings.
- Toggle Auto-Generate Analytics on.
- Click Save.
💡Auto-generate ensures every call has analytics immediately after transcription. Note that this incurs OpenAI API costs for every call.
On-demand analytics
You can trigger individual analytics from the Call Detail page.
- Click Generate in the relevant analytics panel.
- Use Analyse All to run all four analytics at once.
- Re-running an analytic overwrites the existing result.
PII protection
Before any text is sent to OpenAI, KoeIQ automatically scrubs PII from the transcript.
- Redacted entities include names, phone numbers, addresses, and credit card numbers.
- The original transcript is preserved in the database (viewable via the PII diff view in JP Compliance Pack).
- Only the redacted text is sent to OpenAI.
ℹ️PII redaction runs automatically for all analytics types — no extra configuration needed.
Where results are stored
All analytics results are persisted in the analytics_results table in PostgreSQL and displayed on the Call Detail page.