Select an agent — you'll see how many calls are waiting
Click Analyze all → to unlock the button
Click Analyze Conversations (~1–3 min)
Review the score and issues
Click ✨ Generate Updated Prompt (~40s)
Copy → paste into ElevenLabs → Save
Wait 2–3 days → analyze again → score should improve
🟢 What are "calls not yet analyzed"?
These are real calls made by pros to this agent that have never been scored. The tool fetches the last 20 calls from ElevenLabs and checks which ones are missing from the score history.
Only calls over 30 seconds are counted — shorter ones are too brief to analyze meaningfully.
When it says "All recent calls analyzed", it means every call in the last 20 is already in the score history. Come back after the next batch runs.
Test Scripts are for fix verification only — do not use them to measure scores.
Wait at least 2–3 days of real batch calls before re-analyzing after a prompt update.
An agent is stable when it scores 85+ on 3 consecutive analyses.