Just noticed that the Conversation Detail View in the performance dashboard is returning null values for intent_confidence_score when calls are transferred from the AI Bot to a human agent. This is occurring specifically within our primary sales flow, which utilizes the latest conversational AI node configurations. The transfer logic relies on a confidence threshold of 0.85, yet the dashboard metrics display empty fields for these specific interactions, whereas direct agent calls populate the data correctly.
The environment is configured with the European data residency option, and all flows were deployed via the Architect UI without using custom API integrations. The issue appears consistent across all agents in the ‘Sales_EU’ queue during the last 48 hours. We require accurate confidence scores for our monthly quality assurance reports and WFM adherence calculations. The underlying JSON payload from the conversation history API does not seem to include the bot_turn metadata for these specific handoff events, which suggests a potential gap in how the performance module aggregates bot interaction data.
Is there a known limitation regarding intent confidence metric persistence for AI-to-Agent transfers in the current release, or is this a configuration issue within the flow definition?