Free Behavioral Audit
Show us your AI. We'll show you what it's doing.
Up to 10MB of conversations (~3,000 typical). CSV or JSONL. About thirty minutes to your inbox.
no cardno loginno spam
Want to see the dashboard first? Browse a populated CX-bot deployment →
How it goes
Four steps. No onboarding.
01
Upload
CSV or JSONL. Drag, drop, done.
02
We read the human side
What your users do after the AI answers, conversation by conversation.
03
You read
Report lands in your inbox in ~30 minutes.
04
You decide
Or keep watching with a subscription.
What lands in your inbox
A headline you can't dismiss.
Your eval stack grades the model. This grades what the answer did to the person on the other end. The claim, the observation that would falsify it, and three things to ship this week. The deployment is the subject. Individual users appear only as evidence.
Sample · User Health Reportdrift detected
Trust collapses on turn 4 of the cancellation flow.
It is not the model, it is the explanation. The bot offers policy when the user wants action.
| Trust calibration | broken |
| Frustration buildup | at risk |
| Dependency drift | healthy |
| Silent abandonment | at risk |
Prediction · falsifiable
If wrong, you would see the drop on a different turn or a different cohort. It is on turn 4. It is in cancellation only.
signed · BIE engine · run 7F2A4C
BIEverified
SOURCE your uploadmeasures users, not the model~30 min to inbox
Reasonable concerns
The five questions you were going to ask.
What format do my conversations need?
CSV, JSONL, JSON, or TXT. We handle the rest.
Is my data shared anywhere?
No. Your conversations are yours. We never train on them, never share them, never use them for benchmarks.
Why is this free?
Because once you read the report, you'll know what continuous monitoring would do.
How many conversations should I upload?
Anywhere from 200 to 10,000. We adapt the analysis to what you give us.
How long does it take?
About thirty minutes. Sometimes faster. We email you when it's ready.