When an agent accepts a prediction from Smart Compose, whether it is the entire suggestion or just a portion of it, by pressing the tab key or the right arrow key, that action is captured in the reporting system. This information can provide insights into the effectiveness and adoption of Smart Compose within the agent workflow.

The available metrics and dimensions for reporting on Smart Compose usage are as follows:

Query Phrase - The sentence for which the prediction is shown.

Responses - The suggested prediction which may or may not be used by the user.

Used Response Phrase - The predicted text that is used/edited by the agent.

Compose Count (Smart Compose) - Number of requests sent by the user for Smart Compose.

Received Count (Smart Compose) - Number of predictions by AI model.

Rendered Count (Smart Compose) - Number of predictions presented to the user.

Accepted Count (Smart Compose) - Number of predictions used by the user.

Number of Smart Compose Predictions Predicted for Users

You can also create a widget to compare the performance SLAs, predicted CSAT scores, etc. on cases that used Smart Compose with the cases that didn't. It will help you determine the impact of Smart Compose capability on agents' productivity.

Average Case Processing SLA - When Smart Compose Used

Average Case Processing SLA - When Smart Compose Not Used

​

Average of Predicted CSAT Score - When Smart Compose Used

Average of Predicted CSAT Score - When Smart Compose Not Used

Case-wise Case Processing SLA with/without Smart Compose Usage