AB-731 daily web practice

An internal AI council brings cross-functional stakeholders together to align responsible AI decisions. The other options are less suitable because a project plan organizes execution but does not ensure stakeholder alignment on responsible AI strategy, data governance is important, but the question asks for broader cross-functional alignment, and a technology roadmap helps planning, but it does not create the stakeholder forum needed for alignment.

AI transformation should connect the solution to a measurable business outcome before broad investment. The other options are less suitable because buzzwords do not prove the use case has value or can be governed, prototype branding does not measure business value, and unrelated trivia does not show that the solution solves the targeted business problem.

Microsoft 365 Copilot uses Microsoft Graph signals and permissions to ground work in organizational context. The other options are less suitable because browser bookmarks do not provide permission-aware access to Microsoft 365 work data, virtual network peering connects networks and does not ground Copilot in user work content, and a spreadsheet macro automates a workbook but does not supply cross-app organizational context.

A dedicated privacy team can manage governance, data handling, and compliance concerns during AI adoption. The other options are less suitable because access controls are important, but the question asks how to address broader governance and privacy challenges, budgeting funds the work but does not create ownership for privacy decisions, and end-user training helps adoption, but it does not replace privacy governance ownership.

Ensuring proper use of AI services directly supports adherence to security and ethical guidelines in cloud architectures. The other options are less suitable because reducing data storage needs does not address security and responsible engineering practices in AI services, enhancing performance anti-patterns is unrelated to ensuring the proper use of AI services for compliance purposes, and improving retrieval-augmented generation accuracy does not ensure the responsible or secure usage of AI services.

Structured evaluations test outputs against defined quality and safety criteria before release. The other options are less suitable because one successful demo does not prove the app is reliable for broader use, icon size does not measure model output quality or safety, and skipping review raises risk for workflows that affect users or business decisions.

It accurately reflects Microsoft's comprehensive approach to responsible AI by addressing fairness, reliability, safety, privacy, security, inclusiveness, transparency, and accountability. The other options are less suitable because it focuses only on encryption and overlooks other critical aspects of responsible AI such as fairness and inclusiveness, contradicts Microsoft's emphasis on ethical and responsible practices in AI deployment, and maximizing business profits alone disregards the importance of ethical considerations like privacy, security, and accountability that are central to Microsoft’s responsible AI policies.

Retrieval-augmented generation adds relevant approved content to the prompt before the model responds. The other options are less suitable because a larger prompt without retrieval still lacks a dependable connection to the approved documents, randomized responses reduce reliability and do not ground the answer in policy content, and token reporting helps cost analysis but does not ground chatbot answers.

Azure SDKs provide developers with tools and libraries to build applications on Azure using their preferred programming languages. The other options are less suitable because it is not a service that enables development in various programming languages; it's related to chat-based assistance for productivity tasks, the Azure CLI is primarily used for command-line interface operations and does not provide language-specific tools or libraries for application development, and they relate to securing access without passwords, which has no direct relation to developing applications in preferred programming languages.

An internal AI council gives departments a shared forum for responsible AI policy, risk, and adoption decisions. The other options are less suitable because a project plan can organize work, but it does not ensure departments align on responsible AI decisions, a budget funds initiatives but does not align departmental approaches by itself, and training improves awareness, but it does not replace cross-functional strategy ownership.