Conversation analytics
Conversation reviews with a reason and evidence for every score
Choose a period, channels and staff. Before starting you see how many conversations will be analysed and the estimated usage; results arrive in the background.
Criteria and default weights
- Accuracy 25%8/10
- Business rules 20%5/10
- Communication 15%7/10
- Resolution 20%6/10
- Sales opportunity 10%4/10
- Response time 10%7/10
Default weights; an admin can change them in a new rubric version.
Evidence · message #7
“Okay, write to us again if you decide.”
The customer asked about booking; the operator did not offer a time or ask for contact details.
Review: pending
Greeting-only chats are marked “insufficient data” without a score.
What you get
Score and six criteria
A 1–10 score across accuracy, business rules, communication, resolution, sales opportunity and response time.
Evidence on messages
Each finding links to the exact message. High-risk conclusions without evidence are not accepted.
Human review
Confirm, dismiss or correct a score with a comment; the original AI result stays in history.
Problem queue
Wrong information, unresolved request, rude tone, missed next step — filtered by channel and staff member.
Fair split
AI and human parts are scored separately; after a handoff, staff are not scored for someone else's replies.
How it works
- Step 1Choose period and filters
- Step 2Check the estimate
- Step 3Run in the background
- Step 4Review the problems
Frequently asked questions
Is this a verdict on an employee?
No. It assesses a specific conversation, not a person's overall skills.
What about very short chats?
If there is not enough context, the result is “insufficient data” and no score is given.
Can a reviewer change a score?
Yes, with a comment. The original AI result is kept, and a re-analysis is a new record.
See it on your own scenario
We will walk through your channels, team and lead flow in a short demo.