On Virtual Agent Dashboard the NLU prediction is not reflecting correctlyIssue There appears to be an issue where the Correct Natural Language Understanding (NLU) percentage is showing '0' percent for the Dynamic Working and US Paid Time Off topics on the Virtual Agent Dashboard. The steps to reproduce the problem include: Navigate to the Virtual Agent DashboardClick on Topics, and check the topics Dynamic Working or US Paid Time OffThe column Correct NLU Prediction % shows as '0'.ReleaseNot release specific CauseIn the Quebec release of ServiceNow, the property com.glide.cs.intent_confidence_delta has been set to '0' by default, meaning it’s not being used. This setting is recommended, as it doesn’t impact any previously set configurations when upgrading to Quebec. Now, the ServiceNow NLU model relies on the confidence threshold to determine the correct intent. Here’s how topic discovery works: If the Virtual Agent is confident in the user's intent, it directly launches the relevant topic.If the confidence in the top two topics is close (within a set difference), the agent will prompt the user to confirm the intent.If the intent is unclear, the agent defaults to the Fallback topic. When the confidence level is close between two topics, the agent doesn’t automatically choose one but instead presents a list of topics to the user. This is where the system property com.glide.cs.intent_confidence_delta comes into play. If this value is set too high (e.g., 0.07), the Virtual Agent may not be confident enough to auto-select the first topic and will instead offer a list.ResolutionAdjust the system property com.glide.cs.intent_confidence_delta accordingly to increase confidence.