AB-731 Practice Test

Sample AB-731 Practice Questions

10 real questions from the free AB-731bank. Pick an answer in your head, then open “Show answer” for the explanation.

1. 1. Adatum Corporation is making the business case for deploying Microsoft 365 Copilot across 300 knowledge workers. The CFO asks for a concise summary of the primary productivity benefit for knowledge workers. Which statement best captures the core business value proposition?

   • A. Microsoft 365 Copilot replaces junior knowledge worker roles by autonomously completing research, analysis, and communication tasks without human supervision or review.
   • B. Microsoft 365 Copilot reduces the time knowledge workers spend on routine content creation and information retrieval - such as drafting emails, summarizing meetings, and researching topics - so they can focus on higher-value judgment and creative work.
   • C. Microsoft 365 Copilot automatically enforces compliance policies in employee communications, flagging and blocking non-compliant content before it is sent or stored in any Microsoft 365 application.
   • D. Microsoft 365 Copilot eliminates the need for email and meeting management by automatically redirecting communications to the most appropriate recipient and scheduling meetings on employees' behalf without requiring their input.
   Show answer

   Correct answer: B

   The core productivity value of Microsoft 365 Copilot for knowledge workers is time reallocation - reducing time on routine tasks (drafting, summarizing, searching) so employees can focus on higher-value work requiring human judgment, creativity, and expertise. Option A overstates AI capabilities - Copilot augments, not replaces, knowledge workers. Option C describes compliance monitoring, a different Microsoft 365 capability. Option D exaggerates AI autonomy and describes functionality beyond current Copilot capabilities.

2. 2. Contoso Financial Services is building an internal AI assistant to answer employee questions about HR policies, benefits, and compliance requirements. Answers must always be grounded in current official policy documents rather than general AI knowledge. The IT director is evaluating retrieval-augmented generation (RAG). Which benefit of RAG most directly addresses this requirement?

   • A. RAG connects the AI model to a searchable knowledge base of official documents, so the model retrieves relevant, current content before generating a response, keeping answers grounded in authoritative source material rather than potentially outdated training data.
   • B. RAG removes the need for any knowledge base by training the AI model directly on the HR and compliance documents, permanently encoding policy information into the model's parameters so it can answer questions without retrieval.
   • C. RAG enables the AI model to access the internet in real time, searching official government and industry websites to cross-reference HR policies against legal requirements before providing each response.
   • D. RAG ensures the AI model ignores its training data entirely and relies solely on keyword matching within uploaded documents, functioning as a traditional search engine rather than a generative AI system.
   Show answer

   Correct answer: A

   RAG combines a retrieval component - which searches a curated knowledge base for relevant documents - with a generative model that produces answers grounded in those retrieved documents. This ensures responses reflect current official policies rather than the model's training data, which may be outdated. Option B describes fine-tuning, not RAG. Option C describes live internet search access, a different capability. Option D incorrectly describes RAG as eliminating generative AI and functioning only as keyword search.

3. 3. Contoso's IT department runs a customer-facing generative AI chatbot handling 50,000 queries per month via an API-based large language model, costing $4,500/month. The CFO wants a 30% cost reduction without significantly degrading response quality. Which approach is most likely to achieve this cost target?

   • A. Reduce the number of users who can access the chatbot by introducing a paid access model, limiting use to paying customers to cut total query volume below the cost target.
   • B. Switch from the API-based model to a fully on-premises AI model running on existing corporate servers, removing token-based API costs and providing unlimited queries for a fixed infrastructure investment.
   • C. Increase the maximum output token limit for each response, on the theory that longer responses reduce follow-up queries from users and therefore lower total API call volume and monthly cost.
   • D. Audit prompt templates to remove unnecessary context, implement caching for frequent query-response pairs, and consider routing lower-complexity queries to a smaller, more cost-efficient model variant.
   Show answer

   Correct answer: D

   Reducing prompt verbosity, caching common responses, and routing simpler queries to more cost-efficient models are the most practical strategies for reducing API costs without removing functionality. Option A reduces usage by restricting access, which may harm business objectives and customer satisfaction. Option B assumes existing servers can run a capable LLM at comparable quality - a significant infrastructure assumption. Option C is incorrect - longer responses increase output token costs rather than reducing them.

4. 4. Seventh Seq executives are discussing AI strategy. A board member asks how a large language model like the one in Microsoft 365 Copilot actually generates responses. Which explanation best describes the process at a business-relevant level?

   • A. The model performs an exact search of the websites it was trained on, retrieves the most relevant article, and presents a direct quote as the response to the user.
   • B. The model was trained on vast quantities of text to learn statistical patterns in language; given a prompt, it predicts the most contextually appropriate continuation to produce a coherent, relevant response.
   • C. The model connects to a live database of pre-approved answers and selects the closest match to the user's question using traditional keyword search algorithms.
   • D. The model applies a deterministic algorithm with fixed rules to translate structured data inputs into natural language text, similar to how a mail-merge template system works.
   Show answer

   Correct answer: B

   Large language models learn statistical patterns across enormous text datasets during training. At inference time, they predict the most probable continuation of the input text based on those learned patterns - producing coherent, contextually relevant responses. This is fundamentally different from retrieval, keyword search, or rule-based templates. Option A is incorrect - LLMs do not search source articles at inference time. Option C describes a retrieval/search system. Option D describes deterministic rule-based generation, not AI-driven language modeling.

5. 5. Contoso Retail operates 500 stores across three continents. Regional managers need weekly performance commentary - narratives that interpret store KPIs and suggest operational adjustments. The BI platform already maintains real-time KPI dashboards. The CTO wants AI-generated narrative commentary based directly on structured KPI data. Which approach best serves this requirement?

   • A. Deploy Microsoft Copilot Studio to build a standalone conversational chatbot allowing managers to ask natural language questions about their stores, without any direct connection to the existing BI platform.
   • B. Use a large language model via the Microsoft Foundry API with a prompt instructing the model to invent plausible performance narratives based on general retail industry benchmarks rather than actual store data.
   • C. Integrate a generative AI model with the existing BI platform through an API or data connector so the model receives actual KPI values as context and generates personalized, data-grounded narrative commentary for each store.
   • D. Replace the existing BI dashboards with a generative AI chatbot that answers questions verbally, removing structured data visualization and relying on natural language responses for performance reporting.
   Show answer

   Correct answer: C

   The requirement is for commentary grounded in actual KPI data from the existing BI platform. The correct approach integrates the AI with the data source so it generates commentary based on real figures. Option A creates a standalone chatbot disconnected from the BI platform, not meeting the data-grounded requirement. Option B uses fabricated narratives rather than real data, which is operationally dangerous. Option D replaces a functional platform unnecessarily and removes structured visualization that managers depend on.