AI evaluations and predicted CSAT
An AI evaluation reads a conversation and scores it out of 100, says what went well and what to improve, and predicts the rating the customer would have left. It is off until you turn it on.
Turn it on
In the Supportman app in Slack, go to Settings → AI evaluation and answer two questions:
- Which conversations? None (the default), Rated conversations (only ratings that pass your filters), or Every closed conversation, rated or not.
- Handled by: Humans and Fin, Humans only, or Fin only.
How conversations are scored
The overall score out of 100 is made up of:
- Brand voice and tone, out of 20
- Technical knowledge, out of 30
- Problem resolution, out of 30
- Communication, out of 20
Each evaluation also lists strengths, areas to improve, and a predicted CSAT with how confident it is and why.
Where results appear
For rated conversations, the evaluation posts as a thread reply on the rating message. For every closed conversation, it posts in your main channel; very short conversations are skipped. In the web app, open any conversation to see its scores and predicted CSAT.
What is predicted CSAT?
Most customers never rate a conversation. Predicted CSAT is the AI’s estimate of the rating the customer would have left, so unrated conversations still carry a satisfaction signal. It is always shown separately from real ratings.
What is IQS?
IQS is the overall quality score out of 100 from the AI evaluation. Once evaluations are on, it appears on rating messages, in the web app, and against the IQS target you set in Team Settings.
Can I change the scoring categories?
Not from the settings. A custom rubric with your own categories and weights can be set up for your team as a custom implementation: email team@supportman.io to talk it through.
Still stuck?
Email team@supportman.io or use the chat bubble on this page, and a person on the team will help.