AB-410 Exam Questions

Privacy and Security focuses on protecting personal data and ensuring the system is resilient against unauthorized access.

The Prediction model analyzes historical trends to forecast binary outcomes (e.g., Will return / Will not return).

The system message provides the “core instructions“ that define how the AI acts, such as “You are a professional legal assistant.“

Grounding anchors the model‘s output in relevant, verifiable context (like internal PDFs) to prevent hallucinations.

Embeddings models transform natural language into mathematical representations used for semantic matching in vector databases.

Managed Identities provide an identity for applications to use when connecting to resources that support Microsoft Entra authentication.

Sentiment Analysis evaluates text to determine emotional tone, which is vital for escalation and customer satisfaction logic.

Azure AI Search is a specialized search engine that supports vector storage and high-speed retrieval for AI applications.

Zero-shot prompting relies on the model‘s pre-trained knowledge to fulfill a request without needing task-specific guidance.

This system trigger allows for global pre-processing, such as checking for profanity or logging interactions.

The Orchestrator (like Semantic Kernel) coordinates different agents to solve complex, multi-step problems.

Azure AI Language provides the text analytics capabilities needed to process high volumes of streaming data for specific themes and sentiment.

Dataverse Search provides intelligent, cross-table, and fuzzy matching capabilities essential for robust AI grounding.

Max Tokens sets a firm boundary on the number of tokens generated, which controls both length and cost.

Zero-shot prompting relies entirely on the model‘s pre-trained knowledge to understand and execute an instruction.

Coherence measures the structural and logical integrity of the AI‘s output to ensure it makes sense as a whole.

The Alpha parameter in hybrid search balances the weight between keyword (BM25) and vector (Semantic) search scores.

Power Automate acts as the logic engine that connects to external systems and returns data to the bot via connectors.

Adding overlap ensures that entities or concepts split across boundaries are preserved in both chunks, improving retrieval quality.

The Planner is the intelligence layer that determines the roadmap of actions required to achieve a high-level goal.

Entra ID (formerly Azure AD) handles the authentication and authorization flows for securing AI resources.

Azure AI Vision extracts visual features/tags, and Azure OpenAI (LLM) uses those tags to write a natural language description.

Azure AI Studio is the unified platform for discovering, comparing, and deploying a wide variety of foundation models.

Validation and sanitization should occur in the logic layer (Power Automate or APIM) before the data reaches the backend database.

The Prebuilt ID Document model is specifically trained to recognize and pull data from identity documents like licenses and passports.

Prediction models are designed to forecast binary or multiple outcomes based on historical data patterns.

Phi-3 models are efficient SLMs that provide high-quality reasoning in a smaller, faster, and more cost-effective package.

Implicit sharing allows users to benefit from AI features without every individual needing to be granted specific permissions on the backend service.

Inclusiveness ensures that technology is usable by everyone, regardless of their physical or cognitive abilities.

Speech-to-Text transcribes the audio, and Azure OpenAI (LLM) summarizes the resulting text.

Azure APIM allows you to publish, secure, and analyze the APIs that your intelligent applications consume.

The On-premises Data Gateway allows Microsoft Cloud services to securely access data residing in local networks.

Key Phrase Extraction (part of Azure AI Language) identifies the main talking points to determine the overall topic.

Prompt Shields are specifically designed to detect and block malicious content hidden within user inputs or retrieved data.

Model versioning prevents application breakage by allowing developers to test and migrate to newer model releases at their own pace.

Top P (Nucleus Sampling) narrows the pool of potential next words based on their combined probability, helping balance focus and variety.

Azure Cost Management + Billing provides granular reports on token consumption and associated spending.

Field-level security allows administrators to restrict access to specific high-sensitivity columns within a table.

Prompt Flow allows developers to create a DAG (Directed Acyclic Graph) to orchestrate and evaluate AI workflows.

Solutions are the ALM containers used to group all related assets of a business application for easier management and transport.

The system prompt defines the “identity“ of the bot, such as “You are a helpful assistant that only uses technical documentation as your source.“

Vector search relies on the distance between high-dimensional numerical representations of text to find semantic matches.

Category Classification is the ideal model for tagging and routing text-based documents into specific business buckets.

RBAC in Dataverse ensures that only authorized personnel can execute sensitive AI processes or access protected data.

Slot filling is the process where the bot identifies and collects specific pieces of information (entities) needed to complete a task.

Semantic Kernel is the orchestration SDK used to build AI agents that can “think“ (LLM) and “do“ (Plugins/Functions).

Stop sequences are essential for ensuring the model provides concise answers or follows a strict formatting template.

Transparency ensures that the processes and data used to make AI decisions are clear and explainable to users.

Object Detection can locate and label multiple specific items (like a vest or helmet) within a single image.

Top P limits the model‘s word choices to the most likely candidates until their combined probability reaches P, ensuring high-quality variety.

Ownership-based security roles in Dataverse allow for granular control over who can see, edit, or delete specific records.

Trigger phrases are the examples provided to the bot to help it recognize the user‘s intent and launch the correct logic.

Solutions are the standard Application Lifecycle Management (ALM) mechanism for packaging and transporting assets in Power Platform.

Power Automate can check the current time and use conditional logic to decide whether to proceed with the agent‘s action.

Azure AI Vision‘s OCR capabilities can extract text from static images as well as video frames in real-time.

Speech-to-Text is the core service for transcribing spoken audio into machine-readable text formats.

Prompt Shields (part of Content Safety) detect and block malicious instructions hidden within retrieved data or user prompts.

Fine-tuning is used when a model needs to learn a specific vocabulary, style, or deep domain knowledge not present in the base training.

Language Detection identifies the source language, which then allows the bot to route to the correct translation or response logic.

Azure AI Studio includes evaluation tools that calculate metrics like “Groundedness“ to detect factual inconsistencies.