Advice for questions

A question contains specific evaluation criteria that will be assessed when Auto QM evaluates a conversation. Questions should be written as clear, actionable statements that focus on observable agent behaviors, in terms of something the agent did or didn't do explicitly or implicitly. Each question should examine only one behavior or concept to ensure precise feedback. While questions can be phrased in various formats, framing them in a way that requires a yes or no answer typically provides the most consistent results.

The recommended format begins with "Did the agent...", such as the following examples:

• Did the agent properly authenticate the customer before discussing account details?

• Did the agent recap the outcome of the conversation to ensure the customer's understanding?

• Did the agent express empathy and provide reassurance when the customer mentioned their problem?

• Did the agent offer additional assistance before ending the conversation?

Related topics

• Create forms for AI evaluation
• View automated evaluation scores
• About the Auto QM evaluation criteria