According to the Microsoft Azure AI Fundamentals (AI-900) official study guide and the Microsoft Learn module "Identify features of conversational AI workloads on Azure", a Language Understanding (LUIS) model is designed to interpret natural language input by identifying intents (the purpose of an utterance) and entities (specific data items in the utterance).
Every LUIS model automatically includes a special intent called "None." This intent is used to handle utterances that do not fall into any of the model's defined intents. Adding examples of irrelevant or out-of- scope utterances to the None intent helps the model learn to recognize when a user's input does not match any existing categories.
For example, if your e-commerce chatbot handles intents such as "TrackOrder" and "CancelOrder," but a user says "What's your favorite color?", that input should be mapped to the None intent so the bot can respond appropriately, such as "I'm not sure how to answer that." The AI-900 curriculum emphasizes that including diverse None intent examples improves model robustness and prevents false matches, thereby enhancing user experience.
Other options are incorrect:
* A. Test the model by using new utterances: Testing is important but does not define how to detect out- of-scope inputs.
* C. Create a prebuilt task entity: Entities extract specific data but are unrelated to intent classification.
* D. Create a new model: Unnecessary; handling out-of-scope utterances is done within the same model via the None intent.
# Final answer: B. Add utterances to the None intent