View Full Microsoft AB-731 Exam Dumps and Practice Test Dumps.

 

Question 361

A company wants to identify whether a generative AI solution can create useful content from employee instructions. Which test is most appropriate?

1. Check only how quickly the application opens
2. Test representative prompts and evaluate the quality of generated outputs
3. Count the number of available interface settings
4. Measure only the size of the model

Correct Answer: 2

Explanation

A generative AI solution should be evaluated using prompts and tasks that represent how employees will actually use it. The organization can assess whether generated content is relevant, accurate enough for the intended purpose, consistent, and aligned with the required format or tone. Testing only technical specifications does not show whether the solution delivers business value. Representative evaluation can also reveal situations where human review or additional safeguards are required. The results should be compared against predefined requirements so decision-makers can determine whether the capability is suitable for the intended workflow.

Question 362

An organization wants to reduce the risk of an AI assistant using irrelevant information when answering a business question. Which practice can help?

1. Include every available document in every request
2. Remove all contextual information
3. Provide relevant, authoritative context for the task
4. Allow users to access unrestricted information

Correct Answer: 3

Explanation

Relevant and authoritative context can help an AI system focus on information that directly supports the requested task. Providing unnecessary documents can increase token usage and may introduce conflicting or irrelevant information. Organizations should identify trustworthy sources and ensure that users only receive information they are authorized to access. Retrieval mechanisms can help locate relevant content instead of supplying an entire information repository. Context should also be kept current because outdated information can reduce response reliability. Good context management balances relevance, quality, security, and cost.

Question 363

A company is considering an AI application that will process customer records. Which question should be answered before the project proceeds?

1. What safeguards will protect customer information and control access?
2. Which interface contains the most buttons?
3. How can all access restrictions be removed?
4. Which model has the longest name?

Correct Answer: 1

Explanation

Customer records may contain sensitive information, so the organization should establish how the data will be protected and who is authorized to access it. Security considerations can include authentication, authorization, least privilege, data protection, monitoring, and appropriate handling throughout the AI workflow. Privacy requirements should also be assessed based on the information involved and the intended use. These considerations should be addressed before deployment rather than after the system is already processing customer information. Secure design helps reduce the potential impact of unauthorized access or inappropriate data use.

Question 364

A business wants to help users understand the limitations of an AI capability before widespread adoption. Which activity is most useful?

1. Hide known limitations from employees
2. Provide training and practical guidance based on realistic use cases
3. Tell users that all outputs are guaranteed to be correct
4. Remove opportunities for user feedback

Correct Answer: 2

Explanation

Training can help users develop realistic expectations about what an AI capability can and cannot do. Practical guidance should explain appropriate use cases, limitations, verification requirements, data-handling expectations, and situations requiring human review. Realistic examples are useful because they show employees how the capability behaves in actual workflows. Organizations should also provide feedback and reporting mechanisms so that emerging issues can be identified. Responsible adoption depends on informed users who understand that AI can be highly useful while still producing errors or unsupported information.

Question 365

Which situation is most likely to benefit from a multimodal AI capability?

1. Validating whether a number is greater than zero
2. Applying a fixed percentage to a known value
3. Analyzing a product photograph together with its written description
4. Checking whether a password meets a length requirement

Correct Answer: 3

Explanation

Multimodal AI capabilities can work with multiple forms of information, such as text and images. A workflow that requires analysis of a product photograph together with its written description is therefore a suitable candidate. The organization should evaluate the capability using representative images and text from the actual business environment. Other considerations include quality, latency, cost, security, and human review requirements. Multimodal capability alone does not establish suitability; it must provide sufficient performance for the specific task and meet the organization’s business and responsible-AI requirements.

Question 366

A company wants to use AI-generated recommendations in a business process but does not want the system to make final high-impact decisions automatically. Which design is appropriate?

1. Remove all approval steps
2. Require meaningful human review before consequential decisions
3. Give the AI unlimited authority
4. Accept every recommendation without verification

Correct Answer: 2

Explanation

Human review can provide an important safeguard when AI recommendations could have significant consequences. The reviewer should have sufficient information, authority, and time to assess the recommendation rather than simply approving it automatically. Organizations should define when human intervention is mandatory and assign clear responsibility for final decisions. The level of oversight should reflect the potential impact and risk of errors. This approach allows AI to assist with analysis or recommendations while preserving appropriate human judgment and accountability for consequential outcomes.

Question 367

A company wants to determine whether a business process is a suitable candidate for AI. Which factor should it consider first?

1. Whether the process has a meaningful business problem or opportunity that AI could address
2. Whether the process uses the newest technology
3. Whether competitors advertise a similar solution
4. Whether the process has the longest documentation

Correct Answer: 1

Explanation

AI adoption should begin with a meaningful business problem or opportunity rather than technology novelty. The organization should determine what outcome it wants to improve and whether AI is an appropriate way to achieve that outcome. It can then assess factors such as data availability, feasibility, expected value, cost, security, privacy, and risk. Starting with the business objective helps prevent unnecessary AI adoption and provides measurable criteria for evaluating success. A process should not be selected simply because another organization uses AI for a similar activity.

Question 368

A company wants to reduce unnecessary model costs for a simple classification task. What should it investigate?

1. Whether a smaller or less expensive model can meet the required quality
2. Whether every request should contain additional documents
3. Whether monitoring can be removed
4. Whether the largest model should always be selected

Correct Answer: 1

Explanation

Model capability should be matched to the complexity and requirements of the workload. If a smaller or less expensive model can meet the required quality, reliability, and performance levels, it may reduce operating costs without materially affecting business outcomes. Organizations should validate this through representative testing rather than assuming that a smaller model will always be sufficient. Other factors such as latency, context requirements, security, and scalability should also be considered. Cost optimization is most effective when it preserves the quality required by the business process.

Question 369

An organization wants to ensure that important AI decisions can be traced to an accountable business owner. Which governance principle is most relevant?

1. Scalability
2. Accountability
3. Token optimization
4. Multimodality

Correct Answer: 2

Explanation

Accountability establishes clear responsibility for AI systems, decisions, and outcomes. Organizations should identify who owns a particular AI capability, who approves its use, who monitors it, and who is responsible for addressing problems. Clear ownership helps prevent situations where employees assume that responsibility belongs to the AI system or another department. Accountability should be supported by documented roles, governance processes, monitoring, and escalation procedures. This is particularly important for high-impact applications where errors or inappropriate use can have significant consequences for individuals or the organization.

Question 370

A company wants to make internal information easier for an AI application to retrieve while keeping access controlled. Which approach is appropriate?

1. Make all documents publicly accessible
2. Index relevant information while preserving appropriate permissions
3. Remove all document ownership
4. Give every user administrative access

Correct Answer: 2

Explanation

Relevant organizational information can be indexed to support search and retrieval, but access controls must remain in place. The AI application should respect the permissions associated with the underlying information so that users do not receive content they are not authorized to access. Organizations should also maintain source quality and ownership and monitor how information is retrieved and used. Making all information publicly accessible is not an appropriate solution to improve retrieval. Secure retrieval combines useful indexing with authentication, authorization, and appropriate data-governance controls.

Question 371

A company wants to introduce an AI solution gradually across its organization. Which adoption strategy can help manage change?

1. Deploy every capability to every employee at once
2. Use phased adoption with feedback and lessons from earlier stages
3. Avoid measuring employee experiences
4. Remove training after the first department adopts the tool

Correct Answer: 2

Explanation

Phased adoption allows an organization to introduce AI capabilities in manageable stages and learn from each stage before expanding further. Early users can provide feedback about usability, workflow integration, training needs, and unexpected risks. The organization can then refine guidance, governance, support, and implementation plans before adding more users or departments. This approach can reduce the impact of unresolved issues and make adoption more manageable. Phased deployment does not eliminate risk, but it creates opportunities to learn and improve before broader organizational expansion.

Question 372

A business wants to decide whether to purchase an existing AI capability or create a customized solution. Which question is important?

1. Can an existing capability meet the requirement with acceptable cost, security, and functionality?
2. Does customization always provide greater business value?
3. Can maintenance requirements be ignored?
4. Should the most complicated solution always be selected?

Correct Answer: 1

Explanation

Build-versus-buy decisions should begin by determining whether an existing capability can satisfy the actual business requirement. Organizations can compare functionality, customization needs, integration, security, cost, implementation effort, scalability, and ongoing maintenance. Purchasing an existing capability may reduce implementation time, while customization may be justified when specialized requirements cannot be met otherwise. Neither approach is automatically preferable for every scenario. A structured comparison helps decision-makers select an option that provides appropriate business value while keeping implementation and operational complexity manageable.

Question 373

A company wants an AI system to provide answers based on a large collection of frequently changing internal documents. Which capability is particularly relevant?

1. Retrieval-augmented generation
2. Static password validation
3. Image resizing
4. Manual spreadsheet formatting

Correct Answer: 1

Explanation

Retrieval-augmented generation can combine information retrieval with generative AI so that responses can use relevant content from approved sources. This is useful when an organization has large collections of internal information that change over time. The quality of the solution depends on factors such as source accuracy, indexing, retrieval relevance, access controls, and response evaluation. RAG does not guarantee that every generated answer is correct, so important information may still require verification. Its main advantage is providing current, domain-specific context to the generation process.

Question 374

An organization wants to identify why employees are not adopting a newly deployed AI tool. Which action should the adoption team take?

1. Investigate feedback, usage patterns, workflow barriers, and training gaps
2. Assume employees are resistant without collecting evidence
3. Remove support resources
4. Measure only the number of licenses assigned

Correct Answer: 1

Explanation

Low adoption can result from many factors, including unclear value, insufficient training, workflow friction, lack of trust, technical problems, or uncertainty about responsible use. Adoption teams should gather evidence through employee feedback, usage patterns, support requests, and discussions with relevant business groups. This information can help identify specific barriers and guide targeted improvements. License assignment alone does not prove that employees are using a tool effectively. Understanding the reasons behind low adoption allows the organization to improve communication, training, support, or workflow integration rather than making unsupported assumptions.

Question 375

A company wants to ensure that an AI assistant uses the latest approved version of a business policy. Which source-management practice is important?

1. Assign ownership for maintaining and updating authoritative policy sources
2. Keep every obsolete version equally active
3. Remove dates from policy documents
4. Allow employees to modify official policies without review

Correct Answer: 1

Explanation

AI systems that rely on organizational information need trustworthy and current sources. Assigning ownership helps ensure that official policies are reviewed and updated when requirements change. Organizations should also establish how obsolete versions are handled so that retrieval systems do not return outdated information as if it were current. Change management, document metadata, access controls, and validation processes can further support source reliability. Maintaining authoritative information is an ongoing responsibility and can be just as important as the AI model itself when answers depend on organizational policies.

Question 376

A company wants to determine whether an AI solution continues to provide value after deployment. Which activity should it perform regularly?

1. Monitor business outcomes and review performance against defined success measures
2. Stop measuring once the first deployment is complete
3. Measure only the number of model updates
4. Ignore employee feedback

Correct Answer: 1

Explanation

AI value should be monitored after deployment because business conditions, user behavior, data, and workflows can change. Organizations can compare actual outcomes with the success measures established during planning or the pilot. Relevant measures may include productivity, processing time, quality, cost, adoption, or customer outcomes depending on the use case. Employee feedback can provide additional insight into practical value and emerging problems. Ongoing measurement helps organizations determine whether the solution should be maintained, improved, expanded, redesigned, or eventually retired.

Question 377

A company is concerned that an AI system may perform differently across groups of users. Which responsible AI principle should receive attention?

1. Fairness
2. Scalability
3. Cost optimization
4. Token management

Correct Answer: 1

Explanation

Fairness involves considering whether an AI system produces appropriately equitable outcomes across relevant groups. Organizations should identify which groups matter for the particular use case and conduct representative evaluations where appropriate. Differences in performance can reveal issues with data, model behavior, system design, or workflow implementation. Fairness should be assessed alongside other responsible-AI principles such as reliability, safety, privacy, security, and inclusiveness. The appropriate evaluation depends on the business context, and identified issues may require changes to data, models, safeguards, or human oversight.

Question 378

A company wants to ensure employees understand who is responsible for maintaining an AI application after launch. Which practice is most appropriate?

1. Define operational ownership and responsibilities before deployment
2. Leave ownership unspecified
3. Assume the vendor owns every business decision
4. Make every employee equally responsible for maintenance

Correct Answer: 1

Explanation

Operational ownership should be defined before an AI solution enters regular use. Responsibilities may include monitoring, maintenance, incident handling, source updates, user support, security review, governance, and performance evaluation. Different responsibilities can be assigned to different teams, but ownership should be clear enough that issues are not overlooked. Assuming that a vendor owns every organizational responsibility can create gaps because the business still needs to manage its own processes, data, users, and governance. Clear operational roles support accountability and sustainable AI adoption.

Question 379

A business wants to understand whether AI can improve a repetitive process before investing in a large deployment. What should it conduct?

1. A focused pilot with measurable objectives
2. A full organization-wide deployment immediately
3. An evaluation with no defined outcome
4. A test using only hypothetical information

Correct Answer: 1

Explanation

A focused pilot can provide evidence about whether AI can improve a repetitive business process before the organization commits significant resources to broad deployment. The pilot should define measurable objectives, realistic scenarios, relevant users, and appropriate evaluation criteria. Results can reveal potential productivity gains as well as limitations, security concerns, training needs, and workflow issues. Testing a realistic process provides more useful evidence than relying only on hypothetical examples. Pilot findings should then inform decisions about improvement, scaling, governance, or whether the proposed solution should be reconsidered.

Question 380

A company wants to establish a long-term approach for managing AI opportunities, risks, and investments. Which activity is most appropriate?

1. Treat each AI project as completely unrelated to organizational strategy
2. Establish an AI strategy aligned with business goals, governance, adoption, and ongoing review
3. Focus only on selecting models
4. Allow technology teams to determine all business priorities independently

Correct Answer: 2

Explanation

A long-term AI strategy should connect AI investments with business objectives while addressing governance, security, adoption, responsible use, and ongoing measurement. It can provide a framework for identifying opportunities, prioritizing initiatives, allocating resources, and reviewing results over time. Technology selection is an important component but does not define the entire strategy. Business leaders and relevant stakeholders should participate in determining priorities and responsibilities. An aligned strategy helps organizations manage AI as an ongoing business capability rather than as a collection of disconnected technical projects.