📘 Free AB-731 Sample Questions
Your company plans to build a generative AI solution based on internal data.
You recommend using Microsoft Foundry as a starting point to develop and manage the solution.
What is a key benefit of using Microsoft Foundry for this project?
A
Provides a scalable platform for developing and deploying generative AI solutions.
B
Removes the need to select or configure the underlying AI model.
C
Enables business users to build generative AI solutions.
D
Offers a low-code platform for developing generative AI solutions.
Correct Answer:
A. Provides a scalable platform for developing and deploying generative AI solutions.
Explanation:
A. Provides a scalable platform for developing and deploying generative AI solutions.
Microsoft Foundry (now commonly known as Azure AI Foundry) is a comprehensive unified platform designed
to manage the entire lifecycle of AI development. Its primary value proposition for building generative AI
solutions on internal data includes:
bloomcs.com
bloomcs.com
+2
End-to-End Management: It provides a "factory" environment to discover, build, test, and deploy AI models
and agents at scale.
Scalable Infrastructure: It leverages Azure's enterprise-grade infrastructure to handle production workloads,
ensuring that solutions can scale as business needs grow.
Integrated Tooling: It unifies various tools—such as a model catalog, prompt flow for orchestration, and
observability features—into a single workspace, reducing the complexity of moving from experimentation to
full-scale deployment.
Microsoft Learn
Microsoft Learn
+6
Why other options are incorrect
and use case.
Copilot Studio).
B. Removes the need to select or configure the underlying AI model. Microsoft Foundry actually
encourages model selection and configuration. It provides a model catalog where developers can compare,
fine-tune, and select from a variety of foundation models (e.g., GPT, Llama, Phi) to best suit their internal data
C. Enables business users to build generative AI solutions. While Microsoft offers tools for non-technical
users, Microsoft Foundry is primarily a developer-focused platform. Tools specifically meant for business
users to build AI without deep technical knowledge are typically found in the Microsoft Power Platform (e.g.,
D. Offers a low-code platform for developing generative AI solutions. Although it includes some low-
code/no-code interface elements (like visual prompt flows), it is not classified as a "low-code platform". It is a
high-control pro-code/unified AI platform. The dedicated low-code solution for building AI in the Microsoft
ecosystem is Microsoft Copilot Studio.
HOTSPOT -
For each of the following statements, select Yes if the statement is true. Otherwise, select No.
NOTE: Each correct selection is worth one point.
A
Correct Answer:
A.
Explanation:
1. Microsoft Foundry helps organizations securely build and manage generative Al solutions in a governed
environment.
Answer: Yes
Azure Al Foundry is designed to help organizations:
Build, test, and deploy generative Al models
Manage Al solutions across environments
Apply governance, compliance, and security controls
Monitor model performance and responsible Al usage
It provides enterprise-grade controls such as:
Role-based access control (RBAC)
Network isolation
Data protection
Responsible Al tooling
So this statement is correct.
2. Microsoft Foundry provides built-in scalability to enable organizations to expand Al workloads as usage
increases.
Answer: Yes
Azure Al Foundry runs on Azure infrastructure, which provides:
Elastic scaling
Managed compute resources
High availability
Enterprise-grade performance
As Al usage grows (e.g., more inference calls, more users), workloads can scale automatically using Azure's
cloud capabilities.
Therefore, this statement is correct.
3. Microsoft Foundry can be used for image recognition and computer vision tasks.
Answer Yes
HOTSPOT -
Select the answer that correctly completes the sentence.
A
Correct Answer:
A.
Explanation:
Azure Machine Learning,
Azure Machine Learning is used to:
Train custom machine learning models
Perform regression, classification, and forecasting
Work with structured datasets (like historical sales)
Build time-series forecasting models
Manage model training, validation, deployment, and monitoring
Demand forecasting is a classic ML scenario, and Azure ML is designed exactly for this purpose.
Which business requirement most closely relates to grounding a generative AI model?
A
supporting multiple languages
B
measuring the number of user interactions per day
C
enabling users to interact by using natural language queries
D
ensuring that verified company data sources are used for response generation
Correct Answer:
D. ensuring that verified company data sources are used for response generation
Explanation:
Why D—Ensuring that verified company data sources are used for response generation—is the best fit
Grounding a generative AI model means anchoring its output to authoritative, pre-validated information so
that the answers reflect the organization’s verified knowledge base rather than generic or fabricated content.
This requires a data-source layer that is curated, version-controlled, and mapped to the model’s retrieval or
prompt-augmentation mechanisms.
Option D directly addresses this need by insisting on “verified company data sources” as the basis for
responses, which aligns with the technical grounding process (e.g., vector stores, knowledge graphs, or
semantic indexes that retrieve factual data before the model generates text).
the grounded content.
about data provenance or grounding quality.
answers.
References
A (supporting multiple languages) concerns model multilingual capability but does not impact the fidelity of
B (measuring user interactions per day) is an operational metric focused on usage volume; it tells nothing
C (enabling natural-language queries) improves usability but still depends on the underlying data being
trustworthy; without verified sources, natural-language interaction could produce inaccurate or invented
Therefore, the requirement that explicitly guarantees the use of vetted data aligns most closely with the
technical grounding concept and is critical for responsible, trustworthy AI in a corporate environment.
MicrosoftDocs: Responsible AI – Grounding LLMs with company data
https://learn.microsoft.com/en-us/azure/ai-services/content-safety/concepts/grounding
MicrosoftDocs: Retrieval-augmented generation (RAG) with Azure Cognitive Search
https://learn.microsoft.com/en-us/azure/search/search-howto-rag-azure-portal
You need to create a custom Azure Machine Learning model. The data used to train the model is consistent and
uniform.
What should you do first?
A
Prepare the training data.
B
Evaluate the model.
C
Train the model.
D
Tune hyperparameters.
E
Deploy the model.
Correct Answer:
A. Prepare the training data.
Explanation:
A.Prepare the training data – This is the mandatory first step. Uniform, consistent data must be cleaned,
validated for missing values, and properly formatted before any modeling can begin. Without a reliable
dataset, subsequent steps (evaluation, training, hyper-parameter tuning, deployment) would be built on faulty
assumptions and produce inaccurate models.
C.Train the model – Training presupposes that the data has already been prepared; otherwise the training
process fails or yields meaningless results.
E.Deploy the model – Deployment is only viable after a trained and validated model exists, meaning data
preparation and model training must precede it.
B.Evaluate the model – Evaluation compares model performance against ground-truth metrics; it requires a
trained model and therefore cannot be performed before training.
trained baseline model and its evaluation; it is not an initial action.
Hence, the correct first action is Prepare the training data.
References
learning/
D.Tune hyperparameters – Hyper-parameter optimization is part of model refinement and presupposes a
Microsoft Learn – Prepare data for Azure Machine Learning: https://learn.microsoft.com/en-
us/azure/machine-learning/how-to-prepare-data
Microsoft Learn – Azure Machine Learning documentation: https://learn.microsoft.com/en-us/azure/machine-
Prepared with a focus on technical precision and alignment with Azure Machine Learning best practices.
Your company uses a non-reasoning generative AI model to create textual content.
You discover that the model’s responses are inconsistent and do NOT meet expectations.
You need to improve the prompts.
What should you do? More than one answer choice may achieve the goal. Select the BEST answer.
A
Provide the prompts with extensive examples of the expected output.
B
Add the context, sources, and expectations to the prompts.
C
Use technical terms in the prompts to enhance AI comprehension.
D
Add only a single concise requirement to the prompts.
Correct Answer:
B. Add the context, sources, and expectations to the prompts.
Explanation:
Correct answer –B. Add the context, sources, and expectations to the prompts
Modern non-reasoning LLMs generate outputs based on the literal scope of the prompt. Supplying explicit
context, reference material, and a clear articulation of desired behavior reduces ambiguity and aligns the
model’s inference with business requirements.
This approach enables the model to prioritize relevant information (e.g., brand voice, style guide) rather than
relying on generic patterns, which directly addresses the observed inconsistency.
Why A is insufficient
Extensive examples can help but they often lead to longer prompts without guaranteeing better control over
all dimensions (tone, citation, length). The model may over-fit to the examples and still fail when faced with
novel input variations.
Why C is not optimal
Introducing technical jargon may confuse a non-reasoning model that lacks true semantic understanding of
domain terms, resulting in further hallucination or mis-interpretation.
Why D is inadequate
A single concise requirement ignores the need for supporting context and explicit expectations; the model
would likely revert to its default behavior, typically producing generic or off-target text.
consistent outputs.
References
Therefore, Option B is the most effective strategy for improving prompt quality and achieving reliable,
1. Microsoft Learn – Prompt engineering for generative AI: https://learn.microsoft.com/en-us/azure/ai-
services/forms/patterns/prompt-engineering
2. Azure Cognitive Services Documentation – Designing effective prompts:
https://learn.microsoft.com/en-us/azure/ai-services/add-content-to-prompts
HOTSPOT -
For each of the following statements, select Yes if the statement is true. Otherwise, select No.
A
Correct Answer:
A.
Explanation:
A generative AI model guarantees factually accurate responses if the model is trained on a large dataset.
Answer:
Explanation
No
Even when trained on very large datasets, generative AI models do not guarantee factual accuracy.
Reasons:
LLMs generate responses based on probability, not verified facts.
They can produce hallucinations (confident but incorrect information).
Training data may contain errors, outdated data, or bias.
Models do not inherently perform fact-checking.
Therefore, large datasets improve performance but do NOT guarantee accuracy.
? Correct choice: No
Content filtering and responsible AI safeguards help a generative AI model generate safe and inoffensive
content.
Answer:
Yes
Explanation
Responsible AI systems include safety layers such as:
Content filters
Prompt shields
Safety policies
Toxicity detection
Moderation models
These mechanisms help:
Block harmful content
Prevent offensive responses
Reduce misuse
Improve safety of AI outputs
Example in Azure:
Azure OpenAI Content Filtering
Responsible AI policies
? Correct choice: Yes
3 ? A generative AI model always produces fair and unbiased results when the training data has been properly
prepared and reviewed for fairness.
Answer:
No
Explanation Even with carefully prepared datasets, bias can still occur because:
Bias may exist in subtle patterns in the data.
Model training can amplify hidden biases.
Context and prompts can influence outputs.
Fairness cannot be guaranteed, only mitigated.
Responsible Al practices aim to reduce bias, but cannot eliminate it completely.
Correct choice: No
HOTSPOT -
Select the answer that correctly completes the sentence.
A
Correct Answer:
A.
Explanation:
Generative.
The key phrase in the question is:
“create new and original content.”
This is the defining characteristic of Generative AI.
Generative AI models:
Learn patterns from large datasets
Produce content such as:
Text
Images
Generate new outputs that were not in the training data
Audio
Code
Video
Your company plans to use generative AI to help project managers and engineers work with construction
blueprints stored as PDF files.
You need to recommend a generative AI solution that meets the following business requirements:
Processes both images and text -
Summarizes the design of a building
Answers user questions about a building’s design
Extracts information from blueprints, such as the location of electrical, heating, and plumbing systems
What should you recommend?
A
a multi-modal solution
B
an optical character recognition (OCR) solution
C
a text completion solution
D
a document summarization solution
Correct Answer:
A. a multi-modal solution
Explanation:
A. a multi-modal solution.
Text
Images
Diagrams
Multi-modal AI can process multiple types of data simultaneously, such as:
Documents
In which scenario is Azure Machine Learning most likely to deliver strategic value for an organization?
A
Using historical sales data to forecast demand across product categories.
B
Digitizing a paper-based process to reduce errors.
C
Entering customer feedback into a spreadsheet to understand sentiment.
D
Sending personalized emails to customers based on the customer location.
Correct Answer:
A. Using historical sales data to forecast demand across product categories.
Explanation:
Option A – Forecast demand using historical sales data aligns directly with core Azure Machine Learning
Leverages time-series regression models, automated ML pipelines, and experiment tracking to ingest large,
structured historical sales tables.
Provides quantitative forecasts (e.g., demand per SKU, product category) that drive inventory planning,
revenue optimization, and capacity allocation—key strategic levers for mature organizations.
Scales elastically on Azure compute, supports model versioning, and integrates with downstream business
intelligence tools for actionable insights.
Option B – Digitizing a paper-based process focuses on workflow automation rather than predictive
analytics; traditional RPA or document-processing services (e.g., Azure Form Recognizer) would be more
appropriate, but the scenario lacks the predictive or optimization component that Azure ML uniquely delivers.
Option C – Entering feedback into a spreadsheet is a static data-collection activity. While sentiment analysis
is possible, the described approach does not exploit ML for model training, model deployment, or continuous
insight generation, and thus offers limited strategic impact.
Option D – Sending personalized emails based on location can be achieved with rule-based marketing
automation or basic segmentation; it does not require the advanced modeling, feature engineering, or
model-driven personalization that Azure ML enables for truly strategic, data-driven customer engagement.
Why Option A Is Optimal
References
The scenario demands predictive forecasting from historical data, a classic use case where Azure Machine
Learning’s automated ML, experiment tracking, and model deployment pipelines deliver measurable ROI
through demand-driven decision making. The other options address operational efficiency or simple
segmentation without the predictive, model-centric value proposition that defines strategic Azure ML
adoption.
Azure Machine Learning documentation: https://learn.microsoft.com/azure/machine-learning/
Automated ML forecasting guide: https://learn.microsoft.com/azure/machine-learning/how-to-auto-ml-
forecasting
Questions: 1-10 out of 100
Continue Full Practice..
GET ALL 100 QUESTIONS