Smart Comprehend Features and Functionalities
Updated
This article outlines key features and functionalities of the Smart Comprehend module.
Recommending Language-Variant Articles Based on Agent Platform Language
Smart Comprehend supports recommending knowledge-base article variants based on the agent’s platform language, even when the case language is different.
This enhancement helps multilingual support teams, especially those handling cases in languages agents may not understand by surfacing article recommendations in the agent’s platform language.
Example Use Case
Some organizations receive customer messages in multiple languages. For example, customers may write in German.
But agents may only understand English. With this enhancement, Smart Comprehend recommends articles in English automatically without having to manually switch the language, if existed.
For every predicted article:
If an article variant exists that exactly matches the agent's locale (example: Agent locale = English (US) → Article variant must be English (US)) - Smart Comprehend displays that variant to the agent.
If a matching article variant does NOT exist, the article is shown in the originally predicted case language.
Note: Smart Comprehend selects article variants only when the variant’s locale exactly matches the agent’s platform locale. The locale of the language must match exactly. This ensures that agents receive articles in the precise language and regional format they are expected to work in. Example, if the agent locale on the platform is English (US) and the article variant is English (UK), the Smart Comprehend will not show the variant language article. Instead, agent will see original case-language article.
UI Behavior of the Knowledge Base Card
When Smart Comprehend displays an article variant that matches the agent’s platform language:
The article title will appear in agent's locale in the KB Card. The snippet within the recommendation card will not be displayed.
When opened, the full article displays in the agent’s platform language, as expected.
Articles that do not have a matching variant continue to show both title and snippet in the original case language.
