The Agent AI tracer view lets you analyze Pega conversational agent traces directly from the Agent widget (see Agent Observability).
However, it is not available for step agents, Agentic Channel agents, or externally called agents, meaning you will have to open the corresponding conversation case in Dev Studio where the history view is not as user-friendly and does miss some metrics.
A workaround is to expose the Agent Conversation case type (Pega-Autopilot-Conversation) in the Explore Data Landing Page.
This enables AI Tracer access for any agent conversation from a Constellation Insight or List View, simplifying the process.
Here is a step-by-step guide:
1/ Create a new Report Definition on Pega-Autopilot-Conversation class
- You may include meaningful fields like pyAssistantName (agent name), pyAssistantType (Coach or Agent) and pyContextID (business case ID, external call, ….).
- You may exclude Autopilot conversations by adding a filter on agent name (.pyAssistantName not equals to “pzDevAutopilot”)
2/ Create a new Data Page of type list that applies to Pega-Autopilot-Conversation class.
3/ Ensure the Data Page has no parameter AND query option is checked (bottom of Definition tab)
4/ Then, still from Dev Studio, open Application Definition (e.g. menu MyApp>Definition), go to the Case & Data tab, and add the Conversation Case Type.
5/ Flag pyAssistantName, pyAssistantType and pyContextID properties as relevant records for class Pega-Autopilot-Conversation (menu Configure>Application>Inventory>Relevant Records in Dev Studio)
6/ Test from an new incognito browser window to ensure it will work as expected.
You can now see the Conversation Case type in Explore Data, create Insights and open conversation cases surfaced using AI Tracer view.
Final note: You may even start enabling reporting on agent conversations, but be aware that the granular metrics are not natively exposed.
