As an end-user chatting with a Pega conversational agent, you have the option to give feedback at any time (message level) by clicking on thumb up/down icons. You can also complement negative feedback with a free text comment.
User feedback items are relevant information you may leverage to, for instance :
- Follow up with customers (remediation actions against bad comments, …)
- Monitor performance KPI about your conversational agent
- Evaluate optimization directions
- Adapt agent behavior in real time
The agent conversation case (as an instance of the Pega-Autopilot-Conversation case type) is storing these feedback items for reuse, as part of the messages list held by its pyMessages property (page-list of class Pega-Autopilot-Message). You will not find them in the conversation history or the full metrics data used by the AI Tracer.
2 (message) properties are populated when a user gives feedback on a conversation message:
- pyFeedback; equal to -1 (thumb down) or 1 (thumb up)
- pyUserComment: free text
As more organizations mature their conversational channels, this granular feedback becomes a catalyst for shared learning across design, AI governance, and customer operations teams. It turns every conversation into an opportunity to improve how your agent reasons, responds, and evolves.
Feel free to share your own experience:
- Have you recently run (or are you planning to) experiment with in‑conversation feedback to improve your agent’s behavior?
- How are you currently feeding qualitative user feedback into your optimization cycles?