AB-410 Exam Questions

Azure APIM allows you to secure, throttle, and monitor the APIs that your intelligent applications and plugins consume.

Transparency ensures that the logic and data behind AI decisions are clear and understandable to stakeholders.

Vector search uses embeddings to find information based on semantic meaning and context, even if the keywords differ from the query.

This trigger allows for global logic, such as data logging, PII masking, or pre-processing text, every time a user sends a message.

Top K refers to the number of the most relevant context pieces retrieved from the index to be sent to the LLM for grounding.

Azure AI Studio provides evaluation tools to calculate metrics like groundedness, coherence, and relevance.

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

Frequency penalty reduces the likelihood of the model repeating tokens that have already appeared frequently in the output.

Custom Connectors allow Power Platform tools to communicate with any service that has a RESTful API.

Microsoft Purview provides data governance and auditing to track how sensitive data is used within AI applications.

Object Detection is designed to recognize and locate specific visual entities and their counts within an image or video frame.

Entity Extraction is specifically designed to pull structured pieces of data (entities) from unstructured natural language text.

Zero-shot prompting relies on the model’s internal general knowledge to fulfill a request without specific task guidance.

Lowering the temperature (approaching 0) makes the model more focused on the most probable tokens.

Implicit sharing allows the app creator to share the “capability“ of the AI without granting users direct access to the service keys.

Max Tokens sets a firm boundary on the number of tokens generated, which directly controls the physical length of the output.

The Azure AI Studio Playground allows developers to iteratively test, compare, and refine prompts across different models.

Stop sequences prevent the model from rambling by terminating the response once a specific character (like a closing tag) is hit.

Custom Connectors allow the Power Platform to wrap REST or SOAP APIs, providing a bridge for Copilot to perform actions in legacy systems.

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

Microsoft Purview is the unified data governance solution that helps manage and track data lineage, classification, and compliance.

AI Builder identifies the visual object, and Dataverse stores the metadata and links to the associated manuals.

The Alpha parameter determines how much the search engine relies on semantic (vector) scores vs. traditional keyword scores.

Power Automate can connect to SMS providers (like Twilio or ACS) to handle the transactional messaging required for verification.

Audit logs track data changes and access, which is critical for compliance and security in enterprise AI applications.

The system message provides the “inner core“ instructions, defining the AI‘s identity and boundaries.

The Power Platform Admin Center is used to manage DLP policies that prevent sensitive data from being moved between unauthorized connectors.

Sentiment Analysis evaluates text to determine emotional tone, which is vital for escalation logic in support bots.

Higher temperature results in more creative/random output; lower temperature (near 0) makes the model more deterministic.

Document Translation is a specific feature that translates entire files while preserving original formatting and structure.

System topics are the foundational logic blocks built into every bot to handle common conversational scenarios.

Entity Extraction is specifically built to identify and pull structured data points from unstructured natural language text.

Security Roles combined with Business Units define the horizontal data access boundaries within an organization.

Fairness focuses on ensuring that AI systems treat all people equitably and do not perpetuate or introduce social bias.

Global trigger phrases ensure that specific intents can interrupt and redirect the conversation flow from any point in the session.

The Evaluation feature allows developers to run bulk tests and calculate metrics like Groundedness and Coherence against specific datasets.

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

Azure AI Vision can extract text (OCR) from images and video frames in real-time, even in motion.

Presence penalty increases the likelihood of the model introducing new topics rather than staying on existing ones.

Azure AI Studio‘s Model Catalog is the central hub for discovering and deploying both frontier and open-source models.

The System Message is where you define the tone, persona, and stylistic constraints for the AI model.

Few-shot prompting provides context through 2–5 examples, helping the model understand complex or highly specific output requirements.

The Canvas is the visual drag-and-drop environment used to orchestrate the conversational logic of an agent.

Alpha determines how much the search engine values keyword matching versus semantic similarity.

Power Automate flows act as the “skills“ that allow an agent to perform actions across 1,000+ available connectors.

The embedding model converts text into numerical vectors to enable similarity matching against indexed data.

Grounding ensures that the AI’s responses are “anchored“ to your specific business data rather than general training knowledge.

Hallucinations occur when a model generates plausible-sounding but false information not supported by its training.

Azure Monitor provides telemetry, logs, and alerts for Azure AI services to ensure operational reliability and performance tracking.

CLU is the specific service used to map natural language utterances to specific business intents or actions.

The System Message defines the core behavior, persona, and linguistic constraints (like language) for the AI session.

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

Power Automate flows act as the bridge between Copilot Studio and external systems via standard or custom connectors.

Copilot Studio allows you to add SharePoint or OneDrive as a Knowledge Source to enable generative answers from your files.

Power Automate acts as the logic engine that allows an agent to perform multi-step actions across different connectors.

Zero-shot prompting relies on the model‘s pre-trained knowledge to fulfill a request without being given specific examples.

The Azure AI Speech service includes unified capabilities for Speech-to-Text, Text-to-Speech, and real-time Speech Translation.

Variables act as the bot‘s short-term memory, allowing it to remember details provided by the user earlier in the session.

Max Tokens sets a hard limit on the generation length, which is crucial for managing both output size and token consumption costs.

The embedding model converts text into numerical vectors so it can be compared against the indexed data.