Messaging AI debugging information
Several environmental factors might prevent Messaging AI from working as expected. Use the strategies described in this section to debug issues in your Messaging AI instance.
Agent desktop issues
If agents do not receive suggestions from Messaging AI even if the developer tools on the browser show that the desktop is receiving transcripts, the issue might be related to Messaging AIconfigurations in the Digital Messaging interface.
Agents not receiving case suggestions
This section describes how to debug issues with Messaging AI case suggestions.
Messaging AI suggests a case to an agent when:
- a customer utterance results in a topic being identified by the channel, and a case is associated with that topic,
- the case is not already in progress, and has not already been suggested in the current interaction, and,
- the intelligent guidance feature is enabled in your application, and the limit for how many suggested actions can be displayed is not exceeded.
Based on these factors, follow these debugging steps for issues that might cause case suggestions not to work properly:
Topic not identified or case not associated with the topic
If a customer utterance results in a topic being identified by the channel and if a case is associated with that topic, the developer tool logging will display the suggested case and the utterance that triggered the suggestion. The event in the log should look like the following:
Cases suggested: PegaCA-Work-AddressChange;
If a log does not show this type of an event, use the preview pane on your Digital Messaging interface configuration landing page to verify that a given utterance results in the expected topic and that a case is associated with that topic.
If the preview pane shows the correct topic being identified, and if the channel configuration has been changed less than one hour ago, find and expire the D_ChannelConfiguration node’s scoped data page to ensure that the queue processors are using the correct channel.
Case in progress or already suggested
Messaging AI does not suggest a case that has already been suggested or launched in the current interaction based on an earlier trigger.
Suggested actions settings not allowing suggestions
In App Studio, enable the intelligent guidance feature, and set the number of suggested actions to a value greater than 1. For example, if you want the intelligent guidance widget to display 4 suggestions at a time, set the value to 4. If the limit that you set is reached, the widget will not display further case suggestions.
Agents not receiving knowledge suggestions
This section describes how to debug issues with Messaging AI knowledge suggestions.
Messaging AI suggests an article to an agent when:
- a customer utterance results in a topic being identified by the channel, and a knowledge article is associated with that topic,
- the article has not already been suggested or viewed in the current interaction, and,
- the display of Pega Knowledge articles is enabled in your application, and the limit for how many articles can be displayed is not exceeded.
Based on these factors, follow these debugging steps for issues that might cause knowledge suggestions not to work properly:
Topic not identified or article not associated with the topic
If a customer utterance results in a topic being identified by the channel and a knowledge article is associated with that topic, developer tool logging will display the suggested article and the utterance that triggered the suggestion. The event in the log should look like the following:
Articles suggested: {
"pxObjClass":"Code-Pega-List"
,"pyThisPageAsJSON":""
,"pxResults":[
{
"ArticleAbstract":"Uplus Bank waives the annual service charge on all our
supported credit cards for the first two years! Check your specific card
account information for details on standard monthly fees and APR."
,"ArticleTitle":"Credit card fees & charges"
,"pyID":"KC-41"
}
]
}
If a log does not show this type of an event, use the preview pane on your Digital Messaging interface landing page to verify that a given utterance results in the expected topic and that a knowledge article is associated with that topic.
If the preview pane shows the correct topic being identified, and if the channel configuration has been changed less than one hour ago, find and expire the D_ChannelConfiguration node’s scoped data page to ensure that the queue processors are using the correct channel.
Case in progress or already suggested
Messaging AI does not suggest an article that has already been suggested or viewed in the current interaction based on an earlier trigger.
Knowledge management settings not allowing suggestions
In App Studio, enable the display of Pega Knowledge articles, and set the number of knowledge articles to a value greater than 1. For example, if you want the intelligent guidance widget to display 4 knowledge articles at a time, set the value to 4. If the limit that you set is reached, Messaging AI does not suggest further knowledge articles.
Form autofill not working
This section describes how to debug issues with form autofill within a launched case.
Messaging AI autofills detected entities into input fields in the case when:
- dialogs are enabled,
- an entity is detected,
- the entity has not expired,
- a service case is in progress,
- entities have been mapped to input fields in the service case,
- the input field is currently on the screen, and,
- the input field is currently empty.
Based on these factors, follow these debugging steps for issues that might cause the form autofill feature not to work properly:
Dialogs not enabled
Entities detected during a conversation are displayed in the section of the Interaction Portal that also displays dialogs. If dialogs are disabled, this section is not displayed, and entities are not populated in the case fields. Therefore, for form autofill to work, you must enable dialogs for inbound messages.
Entity not detected
You must configure and train your Messaging AI model to detect entities. Out-of-the-box entity models (such as pySystemEntities) do not require training. However, new custom entity types, such as car models in the case of a car manufacturing company, require additional training.
Entity has expired
You can configure entities to expire within a specified time after being detected. If an entity was detected, but its expiration time has lapsed, the entity is not autofilled. The entity is still shown in the entity panel for agents to manually insert. For more information, see the description of the Entity expiration property in
Service case not in progress
Entities are only autofilled for service cases that are in progress.
Entities not mapped to service case properties
Entities are only filled if the entity types have been mapped to the case properties in the Digital Messaging interface on the Entities extraction tab of the case suggestion.
Input fields not visible on screen
Messaging AI only autofills fields that are visible on the screen as part of the current flow action. The list of entities is evaluated when a screen loads and new entities are detected while the screen is visible.
Input fields not empty
Messaging AI only fills entities into empty input fields. If the input field has a preset value, Messaging AI does not replace the value.
Previous topic Configuring Messaging AI in the Digital Messaging interface Next topic Web Messaging