You have already completed the Test before. Hence you can not start it again.
Test is loading...
You must sign in or sign up to start the Test.
You have to finish following quiz, to start this Test:
Your results are here!! for" AB-410 Practice Test 14 "
0 of 60 questions answered correctly
Your time:
Time has elapsed
Your Final Score is : 0
You have attempted : 0
Number of Correct Questions : 0 and scored 0
Number of Incorrect Questions : 0 and Negative marks 0
Average score
Your score
AB-410
You have attempted: 0
Number of Correct Questions: 0 and scored 0
Number of Incorrect Questions: 0 and Negative marks 0
You can review your answers by clicking on “View Answers” option. Important Note : Open Reference Documentation Links in New Tab (Right Click and Open in New Tab).
Answer Review
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Answer
Review
Unattempted
Correct
Incorrect
Unattempted
Every question in this attempt was answered correctly.
Question 1 of 60
1. Question
Which “Responsible AI“ principle focuses on ensuring AI systems are reliable and do not cause unintended harm when failures occur?
Correct
Reliability and Safety emphasize graceful error handling and consistent performance under pressure.
Incorrect
Reliability and Safety emphasize graceful error handling and consistent performance under pressure.
Unattempted
Reliability and Safety emphasize graceful error handling and consistent performance under pressure.
Question 2 of 60
2. Question
Which “Responsible AI” principle focuses on keeping AI systems secure from external attacks such as prompt injection?
Correct
Privacy and Security means the AI system protects data and can withstand malicious manipulation.
Incorrect
Privacy and Security means the AI system protects data and can withstand malicious manipulation.
Unattempted
Privacy and Security means the AI system protects data and can withstand malicious manipulation.
Question 3 of 60
3. Question
Scenario: You must ensure an AI agent does not reveal internal project code names to customers. Which Azure feature helps identify those names in the modelÂ’s output?
Correct
Custom categories in Content Safety can be trained to detect and block specific internal terms or sensitive keywords.
Incorrect
Custom categories in Content Safety can be trained to detect and block specific internal terms or sensitive keywords.
Unattempted
Custom categories in Content Safety can be trained to detect and block specific internal terms or sensitive keywords.
Question 4 of 60
4. Question
Scenario: A developer needs to summarize a 50-page technical document. Which LLM parameter should be increased so the summary does not stop mid-sentence?
Correct
Max Tokens controls the length of generated output, so for long summaries this limit must be high enough.
Incorrect
Max Tokens controls the length of generated output, so for long summaries this limit must be high enough.
Unattempted
Max Tokens controls the length of generated output, so for long summaries this limit must be high enough.
Question 5 of 60
5. Question
In RAG architecture, what does the “Orchestrator” such as Semantic Kernel do?
Correct
The Orchestrator acts as the “glue” by controlling the sequence of operations needed to produce a grounded response.
Incorrect
The Orchestrator acts as the “glue” by controlling the sequence of operations needed to produce a grounded response.
Unattempted
The Orchestrator acts as the “glue” by controlling the sequence of operations needed to produce a grounded response.
Question 6 of 60
6. Question
What is the primary role of “Semantic Kernel” (SK) in the Microsoft AI ecosystem?
Correct
Semantic Kernel is the orchestration SDK for building AI agents that can “think” through the LLM and “do” through Plugins.
Incorrect
Semantic Kernel is the orchestration SDK for building AI agents that can “think” through the LLM and “do” through Plugins.
Unattempted
Semantic Kernel is the orchestration SDK for building AI agents that can “think” through the LLM and “do” through Plugins.
Question 7 of 60
7. Question
In the Azure ecosystem, which component lets you create “Skillsets” that automatically translate documents and perform OCR during search indexing?
Correct
AI Enrichment in Azure AI Search applies skillsets during ingestion to convert raw data into searchable insights.
Incorrect
AI Enrichment in Azure AI Search applies skillsets during ingestion to convert raw data into searchable insights.
Unattempted
AI Enrichment in Azure AI Search applies skillsets during ingestion to convert raw data into searchable insights.
Question 8 of 60
8. Question
Scenario: You need to restrict an AI response to no more than 100 characters. Which parameter creates a hard stop for generation?
Correct
Max Tokens sets a hard limit on generated tokens, which directly relates to response length.
Incorrect
Max Tokens sets a hard limit on generated tokens, which directly relates to response length.
Unattempted
Max Tokens sets a hard limit on generated tokens, which directly relates to response length.
Question 9 of 60
9. Question
Which Azure AI service provides the “Vector Search” capability required to match a user’s natural language question with the most relevant company document?
Correct
Azure AI Search supports vector indexing, which is required for semantic retrieval-augmented generation (RAG) architectures.
Incorrect
Azure AI Search supports vector indexing, which is required for semantic retrieval-augmented generation (RAG) architectures.
Unattempted
Azure AI Search supports vector indexing, which is required for semantic retrieval-augmented generation (RAG) architectures.
Question 10 of 60
10. Question
Scenario: You are evaluating AI performance with a “Golden Dataset.“ Which metric measures the percentage of correct answers from the model?
Correct
Accuracy is a standard evaluation metric that shows how often the model‘s prediction matches the ground truth.
Incorrect
Accuracy is a standard evaluation metric that shows how often the model‘s prediction matches the ground truth.
Unattempted
Accuracy is a standard evaluation metric that shows how often the model‘s prediction matches the ground truth.
Question 11 of 60
11. Question
Scenario: You need to extract “Project Names” and “Budget Amounts” from a thousand unstructured email threads. Which AI Builder model is most appropriate?
Correct
Entity Extraction is specifically designed to pull structured data points from unstructured natural language text.
Incorrect
Entity Extraction is specifically designed to pull structured data points from unstructured natural language text.
Unattempted
Entity Extraction is specifically designed to pull structured data points from unstructured natural language text.
Question 12 of 60
12. Question
Scenario: You need to make sure an AI agent does not generate responses that include medical advice. Where should this restriction be defined most effectively?
Correct
The System Message sets the behavioral constraints and “safety” rules the model must follow.
Incorrect
The System Message sets the behavioral constraints and “safety” rules the model must follow.
Unattempted
The System Message sets the behavioral constraints and “safety” rules the model must follow.
Question 13 of 60
13. Question
Scenario: You are building a logistics app that must recognize “Fragile” stickers on shipping boxes using a camera. Which AI Builder model should you use?
Correct
Object Detection recognizes and locates specific visual items, such as stickers or logos, within an image.
Incorrect
Object Detection recognizes and locates specific visual items, such as stickers or logos, within an image.
Unattempted
Object Detection recognizes and locates specific visual items, such as stickers or logos, within an image.
Question 14 of 60
14. Question
What is the main purpose of “Few-shot Prompting“?
Correct
Few-shot prompting gives context through examples, helping the model understand complex or specific output requirements.
Incorrect
Few-shot prompting gives context through examples, helping the model understand complex or specific output requirements.
Unattempted
Few-shot prompting gives context through examples, helping the model understand complex or specific output requirements.
Question 15 of 60
15. Question
Scenario: You want an AI response to be exactly 50 words. Which parameter sets a hard limit on output length?
Correct
Max Tokens imposes a hard generation-length limit, helping control both response size and token usage.
Incorrect
Max Tokens imposes a hard generation-length limit, helping control both response size and token usage.
Unattempted
Max Tokens imposes a hard generation-length limit, helping control both response size and token usage.
Question 16 of 60
16. Question
Which Microsoft tool visually orchestrates an AI model lifecycle from data ingestion and labeling through deployment and monitoring?
Correct
Azure AI Studio is the unified platform developers use to build, evaluate, and deploy generative AI solutions and custom models.
Incorrect
Azure AI Studio is the unified platform developers use to build, evaluate, and deploy generative AI solutions and custom models.
Unattempted
Azure AI Studio is the unified platform developers use to build, evaluate, and deploy generative AI solutions and custom models.
Question 17 of 60
17. Question
Which Dataverse feature lets an AI agent determine whether a product is “In Stock“ by using a calculated sum across multiple warehouses?
Correct
Rollup columns automatically calculate aggregates, such as sum or count, from related records and provide real-time data for the AI.
Incorrect
Rollup columns automatically calculate aggregates, such as sum or count, from related records and provide real-time data for the AI.
Unattempted
Rollup columns automatically calculate aggregates, such as sum or count, from related records and provide real-time data for the AI.
Question 18 of 60
18. Question
Scenario: A developer needs an AI agent that can summarize a user‘s recent SharePoint files. Which Copilot Studio feature gives this “out-of-the-box“ connectivity?
Correct
Copilot Studio lets you add SharePoint or OneDrive as a Knowledge Source so the agent can automatically search and summarize documents.
Incorrect
Copilot Studio lets you add SharePoint or OneDrive as a Knowledge Source so the agent can automatically search and summarize documents.
Unattempted
Copilot Studio lets you add SharePoint or OneDrive as a Knowledge Source so the agent can automatically search and summarize documents.
Question 19 of 60
19. Question
Which Microsoft Dataverse feature lets you audit exactly what data an AI agent accessed during a specific session?
Correct
Audit logs track data changes and access, which is essential for compliance and security in AI applications.
Incorrect
Audit logs track data changes and access, which is essential for compliance and security in AI applications.
Unattempted
Audit logs track data changes and access, which is essential for compliance and security in AI applications.
Question 20 of 60
20. Question
What is the primary role of “Few-shot Prompting“ in a production AI application?
Correct
Few-shot prompting improves accuracy for specific tasks by showing the model the expected result.
Incorrect
Few-shot prompting improves accuracy for specific tasks by showing the model the expected result.
Unattempted
Few-shot prompting improves accuracy for specific tasks by showing the model the expected result.
Question 21 of 60
21. Question
Which tool monitors “Token Usage” and “Cost” for an Azure OpenAI resource over time?
Correct
Azure Cost Management gives detailed visibility into resource consumption and spending across Azure AI services.
Incorrect
Azure Cost Management gives detailed visibility into resource consumption and spending across Azure AI services.
Unattempted
Azure Cost Management gives detailed visibility into resource consumption and spending across Azure AI services.
Question 22 of 60
22. Question
Which tool is used to move a Copilot agent from a “Development“ environment into a “Production“ environment?
Correct
Solutions are the standard Application Lifecycle Management mechanism for moving assets between Power Platform environments.
Incorrect
Solutions are the standard Application Lifecycle Management mechanism for moving assets between Power Platform environments.
Unattempted
Solutions are the standard Application Lifecycle Management mechanism for moving assets between Power Platform environments.
Question 23 of 60
23. Question
Which Microsoft service provides a secure “Private Link” connection so traffic between your app and Azure OpenAI does not travel over the public internet?
Correct
Private Endpoints let you access Azure AI services through a private IP address inside your Virtual Network (VNet).
Incorrect
Private Endpoints let you access Azure AI services through a private IP address inside your Virtual Network (VNet).
Unattempted
Private Endpoints let you access Azure AI services through a private IP address inside your Virtual Network (VNet).
Question 24 of 60
24. Question
Which Microsoft service lets you define “Data Loss Prevention“ (DLP) policies to manage which connectors can be used with AI-driven Power Automate flows?
Correct
The Power Platform Admin Center is the central location for managing DLP policies so sensitive data is not leaked through unauthorized connectors.
Incorrect
The Power Platform Admin Center is the central location for managing DLP policies so sensitive data is not leaked through unauthorized connectors.
Unattempted
The Power Platform Admin Center is the central location for managing DLP policies so sensitive data is not leaked through unauthorized connectors.
Question 25 of 60
25. Question
In relation to LLMs, what does “Model Distillation“ mean?
Correct
Distillation produces more efficient Small Language Models (SLMs) that preserve much of the reasoning capability of larger models while reducing cost.
Incorrect
Distillation produces more efficient Small Language Models (SLMs) that preserve much of the reasoning capability of larger models while reducing cost.
Unattempted
Distillation produces more efficient Small Language Models (SLMs) that preserve much of the reasoning capability of larger models while reducing cost.
Question 26 of 60
26. Question
Which AI Builder feature is appropriate for identifying specific objects, such as “Solar Panels,“ in high-resolution satellite imagery?
Correct
Object Detection is intended to locate and label particular visual entities within an image or video frame.
Incorrect
Object Detection is intended to locate and label particular visual entities within an image or video frame.
Unattempted
Object Detection is intended to locate and label particular visual entities within an image or video frame.
Question 27 of 60
27. Question
Which AI Builder model should extract the “Total Balance” from a photo of a bank statement with a unique layout?
Correct
The Unstructured Document model is designed to locate data points in documents when the layout is not fixed or consistent.
Incorrect
The Unstructured Document model is designed to locate data points in documents when the layout is not fixed or consistent.
Unattempted
The Unstructured Document model is designed to locate data points in documents when the layout is not fixed or consistent.
Question 28 of 60
28. Question
What does the “Presence Penalty“ parameter discourage in a model completion?
Correct
Presence penalty makes the model more likely to introduce new topics instead of returning to topics already discussed.
Incorrect
Presence penalty makes the model more likely to introduce new topics instead of returning to topics already discussed.
Unattempted
Presence penalty makes the model more likely to introduce new topics instead of returning to topics already discussed.
Question 29 of 60
29. Question
What does the “Top P“ parameter, also called Nucleus Sampling, specifically limit during LLM generation?
Correct
Top P limits consideration to the most likely tokens whose cumulative probability reaches the P threshold, which improves quality.
Incorrect
Top P limits consideration to the most likely tokens whose cumulative probability reaches the P threshold, which improves quality.
Unattempted
Top P limits consideration to the most likely tokens whose cumulative probability reaches the P threshold, which improves quality.
Question 30 of 60
30. Question
Scenario: You are building an agent for a global HR department. The agent must recognize whether a user’s question contains a “Vacation Request” intent. Which service handles this?
Correct
CLU is the specific Azure AI Language feature that maps natural language utterances to defined business intents.
Incorrect
CLU is the specific Azure AI Language feature that maps natural language utterances to defined business intents.
Unattempted
CLU is the specific Azure AI Language feature that maps natural language utterances to defined business intents.
Question 31 of 60
31. Question
Which “Responsible AI“ principle is violated when an AI hiring tool consistently ranks male candidates above female candidates for technical roles?
Correct
Fairness means AI systems should not introduce or reinforce bias based on protected characteristics such as gender.
Incorrect
Fairness means AI systems should not introduce or reinforce bias based on protected characteristics such as gender.
Unattempted
Fairness means AI systems should not introduce or reinforce bias based on protected characteristics such as gender.
Question 32 of 60
32. Question
You are using AI Builder to extract data from invoices, and the invoice layout is the same across all vendors. Which model type should you use?
Correct
Structured Document models, also known as form processing, are optimized for documents with consistent layouts, such as standard invoices or applications.
Incorrect
Structured Document models, also known as form processing, are optimized for documents with consistent layouts, such as standard invoices or applications.
Unattempted
Structured Document models, also known as form processing, are optimized for documents with consistent layouts, such as standard invoices or applications.
Question 33 of 60
33. Question
A developer wants a Power Automate flow to trigger when a customer sends an image of a damaged part so the part number can be identified. Which model is required?
Correct
Object Detection can be trained to recognize and locate specific products or components in an image.
Incorrect
Object Detection can be trained to recognize and locate specific products or components in an image.
Unattempted
Object Detection can be trained to recognize and locate specific products or components in an image.
Question 34 of 60
34. Question
In Azure AI Search, what is the “Search Index“?
Correct
The Search Index is the optimized set of data that the search engine queries to quickly find relevant results.
Incorrect
The Search Index is the optimized set of data that the search engine queries to quickly find relevant results.
Unattempted
The Search Index is the optimized set of data that the search engine queries to quickly find relevant results.
Question 35 of 60
35. Question
A developer is creating a multi-agent system where one agent retrieves data and another summarizes it. Which component determines which agent should be called first based on the user‘s goal?
Correct
The Planner is the architectural component that breaks down a complex goal into a sequence of steps or agent calls.
Incorrect
The Planner is the architectural component that breaks down a complex goal into a sequence of steps or agent calls.
Unattempted
The Planner is the architectural component that breaks down a complex goal into a sequence of steps or agent calls.
Question 36 of 60
36. Question
Which Azure AI service provides a dedicated environment for testing and comparing how different LLMs perform with the same prompt?
Correct
The Azure AI Studio Playground lets developers iteratively test, compare, and refine prompts across different models.
Incorrect
The Azure AI Studio Playground lets developers iteratively test, compare, and refine prompts across different models.
Unattempted
The Azure AI Studio Playground lets developers iteratively test, compare, and refine prompts across different models.
Question 37 of 60
37. Question
Scenario: You are building an intelligent app for a bank. The app must summarize a customer‘s monthly transaction history but must never show the full account number. Which AI technique should be applied first?
Correct
PII, or Personally Identifiable Information, detection identifies sensitive data such as account numbers so it can be masked before the summary is generated.
Incorrect
PII, or Personally Identifiable Information, detection identifies sensitive data such as account numbers so it can be masked before the summary is generated.
Unattempted
PII, or Personally Identifiable Information, detection identifies sensitive data such as account numbers so it can be masked before the summary is generated.
Question 38 of 60
38. Question
Scenario: An AI agent must access a customer‘s private order history in Dataverse. Which security layer stops the agent from viewing data that belongs to other customers?
Correct
Row-level security makes sure that the AI, acting on behalf of the user, retrieves only the records that the specific user is allowed to see.
Incorrect
Row-level security makes sure that the AI, acting on behalf of the user, retrieves only the records that the specific user is allowed to see.
Unattempted
Row-level security makes sure that the AI, acting on behalf of the user, retrieves only the records that the specific user is allowed to see.
Question 39 of 60
39. Question
In AI governance, what does the principle of “Accountability“ mean?
Correct
Accountability makes sure there is human oversight and responsibility for AI system behavior and outcomes.
Incorrect
Accountability makes sure there is human oversight and responsibility for AI system behavior and outcomes.
Unattempted
Accountability makes sure there is human oversight and responsibility for AI system behavior and outcomes.
Question 40 of 60
40. Question
Which Azure AI service feature supports “Speech Intent Recognition,“ where the system identifies a user‘s goal directly from spoken audio?
Correct
Integrating Azure AI Speech with CLU lets the system bypass the text-only phase and map spoken language directly to specific business intents.
Incorrect
Integrating Azure AI Speech with CLU lets the system bypass the text-only phase and map spoken language directly to specific business intents.
Unattempted
Integrating Azure AI Speech with CLU lets the system bypass the text-only phase and map spoken language directly to specific business intents.
Question 41 of 60
41. Question
What is the primary role of the “System Role“ (System Prompt) in an enterprise Chat Completion API?
Correct
The system role supplies the foundational instructions that control how the AI interacts and which rules it must follow.
Incorrect
The system role supplies the foundational instructions that control how the AI interacts and which rules it must follow.
Unattempted
The system role supplies the foundational instructions that control how the AI interacts and which rules it must follow.
Question 42 of 60
42. Question
Which Responsible AI principle helps ensure an AI system does not deliver different quality of service depending on a userÂ’s demographic group?
Correct
Fairness is concerned with reducing bias and providing equitable treatment for every user.
Incorrect
Fairness is concerned with reducing bias and providing equitable treatment for every user.
Unattempted
Fairness is concerned with reducing bias and providing equitable treatment for every user.
Question 43 of 60
43. Question
What is the main purpose of “Few-shot Prompting“?
Correct
Few-shot prompting supplies context and examples inside the prompt, helping the LLM understand complex patterns without a full fine-tuning process.
Incorrect
Few-shot prompting supplies context and examples inside the prompt, helping the LLM understand complex patterns without a full fine-tuning process.
Unattempted
Few-shot prompting supplies context and examples inside the prompt, helping the LLM understand complex patterns without a full fine-tuning process.
Question 44 of 60
44. Question
What does “Grounding“ mean for an AI model?
Correct
Grounding makes sure the model bases its responses on trusted data, such as your company‘s knowledge base, instead of relying on its general training data.
Incorrect
Grounding makes sure the model bases its responses on trusted data, such as your company‘s knowledge base, instead of relying on its general training data.
Unattempted
Grounding makes sure the model bases its responses on trusted data, such as your company‘s knowledge base, instead of relying on its general training data.
Question 45 of 60
45. Question
In Microsoft Copilot Studio, what is the purpose of the “Global“ variable scope?
Correct
Global variables let the bot retain information, such as a user‘s ID or preference, as the conversation moves between topics.
Incorrect
Global variables let the bot retain information, such as a user‘s ID or preference, as the conversation moves between topics.
Unattempted
Global variables let the bot retain information, such as a user‘s ID or preference, as the conversation moves between topics.
Question 46 of 60
46. Question
Which Microsoft Dataverse column type should store geographic coordinates, Latitude and Longitude, for an AI-driven logistics app?
Correct
Floating Point or Decimal numbers provide the precision needed to store and calculate geographic coordinates accurately.
Incorrect
Floating Point or Decimal numbers provide the precision needed to store and calculate geographic coordinates accurately.
Unattempted
Floating Point or Decimal numbers provide the precision needed to store and calculate geographic coordinates accurately.
Question 47 of 60
47. Question
What does the “Presence Penalty“ parameter specifically discourage in AI-generated output?
Correct
Presence Penalty makes the model more likely to introduce new topics by penalizing tokens according to whether they have already appeared.
Incorrect
Presence Penalty makes the model more likely to introduce new topics by penalizing tokens according to whether they have already appeared.
Unattempted
Presence Penalty makes the model more likely to introduce new topics by penalizing tokens according to whether they have already appeared.
Question 48 of 60
48. Question
A developer must make sure that a Copilot agent answers questions only from the companyÂ’s internal policy documents. What is this concept called?
Correct
Grounding means giving the AI a specific knowledge base, or context, so its responses stay factually accurate and relevant to the organization.
Incorrect
Grounding means giving the AI a specific knowledge base, or context, so its responses stay factually accurate and relevant to the organization.
Unattempted
Grounding means giving the AI a specific knowledge base, or context, so its responses stay factually accurate and relevant to the organization.
Question 49 of 60
49. Question
Which component is used to securely store and manage API keys for Azure AI Services when deploying an application to Azure App Service?
Correct
Azure Key Vault is the industry-standard service for securely storing secrets, keys, and certificates.
Incorrect
Azure Key Vault is the industry-standard service for securely storing secrets, keys, and certificates.
Unattempted
Azure Key Vault is the industry-standard service for securely storing secrets, keys, and certificates.
Question 50 of 60
50. Question
Which Azure AI service would you use to identify specific celebrities or famous landmarks in a set of travel photos?
Correct
Azure AI Vision contains pre-trained models for domain-specific content, including celebrities and landmarks.
Incorrect
Azure AI Vision contains pre-trained models for domain-specific content, including celebrities and landmarks.
Unattempted
Azure AI Vision contains pre-trained models for domain-specific content, including celebrities and landmarks.
Question 51 of 60
51. Question
In a “RAG” architecture, which component ensures that the user’s query is matched to the most relevant local data?
Correct
The Vector Search Index applies high-dimensional mathematics to identify documents that are semantically related to the userÂ’s query.
Incorrect
The Vector Search Index applies high-dimensional mathematics to identify documents that are semantically related to the userÂ’s query.
Unattempted
The Vector Search Index applies high-dimensional mathematics to identify documents that are semantically related to the userÂ’s query.
Question 52 of 60
52. Question
In Copilot Studio, what is the purpose of the “On Message Received“ system trigger?
Correct
This trigger supports global logic, including logging or text pre-processing, whenever the user sends a message.
Incorrect
This trigger supports global logic, including logging or text pre-processing, whenever the user sends a message.
Unattempted
This trigger supports global logic, including logging or text pre-processing, whenever the user sends a message.
Question 53 of 60
53. Question
Which tool evaluates an AI response’s “Groundedness” by comparing the response with the retrieved context?
Correct
Azure AI Studio includes automated evaluation tools that calculate metrics such as groundedness, coherence, and relevance.
Incorrect
Azure AI Studio includes automated evaluation tools that calculate metrics such as groundedness, coherence, and relevance.
Unattempted
Azure AI Studio includes automated evaluation tools that calculate metrics such as groundedness, coherence, and relevance.
Question 54 of 60
54. Question
Scenario: You need a bot to identify whether a user is “Angry,” “Neutral,” or “Happy” so it can route the user to a human agent. Which service is best suited for this?
Correct
Sentiment Analysis reviews text to identify emotional tone, which is important for escalation logic.
Incorrect
Sentiment Analysis reviews text to identify emotional tone, which is important for escalation logic.
Unattempted
Sentiment Analysis reviews text to identify emotional tone, which is important for escalation logic.
Question 55 of 60
55. Question
Scenario: You must build an AI agent that performs real-time “What-If” analysis on a financial spreadsheet. Which tool should handle the logic?
Correct
Power Automate flows function as the “skills” or “actions” that a Copilot agent can call for complex calculations or data lookups.
Incorrect
Power Automate flows function as the “skills” or “actions” that a Copilot agent can call for complex calculations or data lookups.
Unattempted
Power Automate flows function as the “skills” or “actions” that a Copilot agent can call for complex calculations or data lookups.
Question 56 of 60
56. Question
What is a “Hallucination” in a Large Language Model?
Correct
Hallucinations happen when a model produces information that sounds plausible but is false and is not supported by its training or context.
Incorrect
Hallucinations happen when a model produces information that sounds plausible but is false and is not supported by its training or context.
Unattempted
Hallucinations happen when a model produces information that sounds plausible but is false and is not supported by its training or context.
Question 57 of 60
57. Question
Which tool monitors the real-time “Health” and “Latency” of an Azure OpenAI endpoint?
Correct
Azure Monitor supplies telemetry, logs, and alerts for Azure AI services to maintain operational reliability.
Incorrect
Azure Monitor supplies telemetry, logs, and alerts for Azure AI services to maintain operational reliability.
Unattempted
Azure Monitor supplies telemetry, logs, and alerts for Azure AI services to maintain operational reliability.
Question 58 of 60
58. Question
Scenario: You are building a customer service agent in Copilot Studio and want it to automatically greet the user by first name. Where should the name be stored?
Correct
Global variables, or User variables, let the bot store and reuse information across topics during a single session.
Incorrect
Global variables, or User variables, let the bot store and reuse information across topics during a single session.
Unattempted
Global variables, or User variables, let the bot store and reuse information across topics during a single session.
Question 59 of 60
59. Question
In a RAG architecture, what does “Top K“ retrieval refer to?
Correct
Top K is the count of the most relevant context pieces retrieved and passed to the LLM for grounding.
Incorrect
Top K is the count of the most relevant context pieces retrieved and passed to the LLM for grounding.
Unattempted
Top K is the count of the most relevant context pieces retrieved and passed to the LLM for grounding.
Question 60 of 60
60. Question
Which tool is used to move an intelligent application from a Development environment to a Production environment?
Correct
Solutions are the standard Application Lifecycle Management mechanism for bundling and transporting Power Platform assets.
Incorrect
Solutions are the standard Application Lifecycle Management mechanism for bundling and transporting Power Platform assets.
Unattempted
Solutions are the standard Application Lifecycle Management mechanism for bundling and transporting Power Platform assets.
X
Use Page numbers below to navigate to other practice tests