New Release AI-900 Microsoft Certified: Azure AI Fundamentals Questions

According to the Microsoft Azure AI Fundamentals (AI-900) official study guide and the Microsoft Learn module “Identify features of common machine learning types”, the standard workflow for creating and deploying a machine learning model — especially within Azure Machine Learning Designer — follows a structured sequence of steps to ensure that the model is trained effectively and evaluated correctly.

Here’s the detailed breakdown of the correct order:

Import and prepare a dataset:This is always the first step in the machine learning lifecycle. The dataset is imported into Azure Machine Learning and cleaned or preprocessed. Preparation might include handling missing values, normalizing data, removing outliers, and encoding categorical variables. This ensures the dataset is ready for modeling.

Split the data randomly into training data and validation data:The dataset is then divided into two parts — the training set and the validation (or testing) set. Typically, around 70–80% of the data is used for training and 20–30% for validation. This step ensures that the model can be evaluated on unseen data later, preventing overfitting.

Train the model:During this stage, the machine learning algorithm learns patterns from the training data. Azure Machine Learning Designer provides multiple algorithms (classification, regression, clustering, etc.) that can be applied using “Train Model” components.

Evaluate the model against the validation dataset:Finally, the trained model’s performance is tested using the validation dataset. Evaluation metrics such as accuracy, precision, recall, or RMSE (depending on the model type) are calculated to assess how well the model generalizes to new data.

The incorrect option — “Evaluate the model against the original dataset” — is not used in proper ML workflows, because evaluating on the same data used for training would give misleadingly high accuracy due to overfitting.

According to the Microsoft Azure AI Fundamentals (AI-900) official study guide and Microsoft Learn modules under the topic “Describe features of common AI workloads” and “Identify capabilities of Azure AI services”, QnA Maker is the service designed specifically to create a knowledge base (KB) or question-and-answer bot from existing content such as FAQ documents, product manuals, support pages, or structured knowledge sources.

QnA Maker enables developers to take semi-structured text (for example, an FAQ document or webpage) and automatically generate a knowledge base of pairs of questions and corresponding answers. This knowledge base can then be connected to a chatbot, typically through the Azure Bot Service, so that users can interact with it conversationally. The key advantage is that the process does not require deep machine learning or programming expertise. The service uses natural language processing (NLP) to match user queries with the most relevant pre-defined answers in the knowledge base.

In the AI-900 curriculum, this falls under the Conversational AI workload—creating intelligent bots that can respond naturally to user questions. Microsoft’s training content explains that “QnA Maker extracts pairs of question and answer from your content and builds a knowledge base that can be queried by bots and other applications.” The output, as shown in the example diagram, demonstrates how user input (the question) triggers a request to the QnA Maker API, which returns a JSON response containing the best-matched answer.

The other options are not correct because:

B. Language Understanding (LUIS) is used to interpret user intent and extract entities, not to create FAQs.

C. Text Analytics performs text extraction, sentiment analysis, and key-phrase detection but does not build a Q & A knowledge base.

D. Speech handles speech-to-text or text-to-speech, not Q & A matching.

Therefore, per the AI-900 study guide and Microsoft Learn, the verified and correct answer is A. QnA Maker.

According to Microsoft’s Responsible AI principles, one of the key guiding values is Reliability and Safety, which ensures that AI systems operate consistently, accurately, and safely under all intended conditions. The AI-900 study materials and Microsoft Learn modules explain that an AI system must be trustworthy and dependable, meaning it should not produce results when the input data is incomplete, corrupted, or significantly outside the expected range.

In the given scenario, the AI system avoids providing predictions when important fields contain unusual or missing values. This behavior demonstrates reliability and safety because it prevents the system from making unreliable or potentially harmful decisions based on bad or insufficient data. Microsoft emphasizes that AI systems must undergo extensive validation, testing, and monitoring to ensure stable performance and predictable outcomes, even when data conditions vary.

The other options do not fit this scenario:

Inclusiveness ensures that AI systems are accessible to and usable by all people, regardless of abilities or backgrounds.

Privacy and Security focuses on protecting user data and ensuring it is used responsibly.

Transparency involves making AI decisions explainable and understandable to humans.

Only Reliability and Safety directly address the concept of an AI system refusing to act or returning an error when it cannot make a trustworthy prediction. This principle helps prevent inaccurate or unsafe outputs, maintaining confidence in the system’s integrity.

Therefore, ensuring an AI system does not produce predictions when input data is incomplete or unusual aligns directly with Microsoft’s Reliability and Safety principle for responsible AI.

In a Language Understanding (LUIS) application, an utterance represents an example of what a user might say to the bot. According to Microsoft Learn – “Build a Language Understanding app”, an utterance is a sample phrase that helps train the LUIS model to recognize user intent.

In the given example — “Which act is playing on the main stage?” — the statement is an utterance that a user might say to find out about show schedules. LUIS uses utterances like this to identify the intent (the user’s goal, e.g., GetShowInfo) and to extract any entities (e.g., main stage) that provide additional details for fulfilling the request.

To clarify the other elements:

Intent: The overall purpose or action (e.g., “FindShowDetails”).

Entity: Specific information in the utterance (e.g., “main stage”).

Domain: A general subject area (e.g., entertainment, events).

Thus, “Which act is playing on the main stage?” is an utterance used to train the LUIS model to understand natural language input.