AB-100 Sample Questions & Answers
Deploying AI-powered solutions, including tuning, testing and the ALM process, carries nearly half the weight, ahead of planning strategy, costs and requirements, and designing the AI, agents and extensibility behind each solution.
Launch the full AB-100 simulator →Showing 10 of 20 free samples.
- Question 1Intermediate
Plan AI-powered business solutions · Provide prompt engineering guidelines, prompt library guidance, and small language model use cases
You are designing an AI architecture for a logistics company. They need an agent deployed on edge devices in delivery trucks that experience frequent network disconnections. The agent must process basic natural language commands to update delivery statuses locally.
To meet these constraints, you should recommend deploying a ___________ model to the edge devices.
Show answer & explanation
Correct answer: B
Small Language Models (SLMs), such as the Phi family, are designed with a smaller parameter count, allowing them to run efficiently on edge devices with limited compute and memory. This makes them the perfect use case for scenarios requiring offline capabilities or low-latency local processing where network connectivity is intermittent.
- Question 2Intermediate
Plan AI-powered business solutions · Select ROI criteria and create an ROI analysis for AI-powered business solutions
You are creating a Total Cost of Ownership (TCO) and Return on Investment (ROI) analysis for a proposed Copilot Studio agent intended to deflect IT helpdesk tickets.
Which combination of metrics represents the most comprehensive set of criteria for evaluating the ROI of this specific AI solution?
Show answer & explanation
Correct answer: D
A comprehensive ROI/TCO analysis must weigh the costs (platform licensing, initial development effort, and ongoing maintenance/ALM) against the tangible business benefits (ticket deflection rate, reduction in average time-to-resolution, and corresponding labor cost savings). Lines of code or simple API call counts do not translate directly to business ROI without context.
- Question 3Beginner
Plan AI-powered business solutions · Determine when to build custom agents or extend Microsoft 365 Copilot
True or False: When deciding between building a custom agent in Copilot Studio and extending Microsoft 365 Copilot, you must build a custom standalone agent if your solution requires grounding on data stored entirely in external, non-Microsoft systems.
Show answer & explanation
Correct answer: B
False. Microsoft 365 Copilot can be extended to ground on external, non-Microsoft systems using Microsoft Graph connectors or API plugins. You do not strictly need to build a custom standalone agent just because the data lives outside the Microsoft ecosystem.
- Question 4Intermediate
Plan AI-powered business solutions · Design a multi-agent solution by using platforms such as Microsoft 365 Copilot, Copilot Studio, and Microsoft Foundry
Trey Research is designing a complex solution that requires multiple specialized agents to collaborate. Agent A handles data retrieval from Dataverse, Agent B performs complex mathematical analysis using a custom Python script, and Agent C formats the final response for the user.
Which platform provides the native framework required to design and orchestrate this specific multi-agent collaboration pattern?
Show answer & explanation
Correct answer: D
Microsoft Foundry (Azure AI Foundry) provides the advanced pro-code tools and frameworks (such as Semantic Kernel or Prompt flow) required to orchestrate complex multi-agent solutions where agents perform highly specialized tasks like executing custom Python scripts and passing state between each other. While Copilot Studio supports agent creation, highly complex custom multi-agent orchestration involving code execution is best suited for Microsoft Foundry.
- Question 5Intermediate
Plan AI-powered business solutions · Organize business solution data to be available for other AI systems
You are organizing enterprise data within Dataverse so that it can be optimally consumed by various AI systems across your organization, including Copilot Studio agents and custom Foundry models.
To ensure the AI systems can automatically understand the relationships and semantic meaning of the data without extensive hardcoding, which Dataverse feature should you heavily utilize during data organization?
Show answer & explanation
Correct answer: C
Organizing data using the Common Data Model (CDM) standard entities and establishing clear table relationships (1:N, N:N) allows AI systems, like Copilots, to inherently understand the semantic structure of the data. Generative AI relies heavily on metadata and relationships to traverse data and generate accurate, context-aware responses.
- Question 6IntermediateSelect 2
Plan AI-powered business solutions · Provide prompt engineering guidelines, prompt library guidance, and small language model use cases
Your organization is establishing a corporate prompt library for generative AI solutions. As the Solutions Architect, you are writing the prompt engineering guidelines for the team.
Which TWO practices should you include in the guidelines to ensure prompts yield consistent and safe results? (Select TWO)
Show answer & explanation
Correct answers: B, D
Effective prompt engineering relies on clear instructions, contextual framing, and specific output formatting. Additionally, including few-shot examples (positive and negative) helps the model understand the exact expectations, leading to more consistent and accurate results.
Effective prompt engineering relies on clear instructions, contextual framing, and specific output formatting. Additionally, including few-shot examples (positive and negative) helps the model understand the exact expectations, leading to more consistent and accurate results.
- Question 7Intermediate
Design AI-powered business solutions · Design topics, agents, agent flows, and prompt actions in Copilot Studio, including fallback
You are designing an agent in Copilot Studio for a retail company. The agent must handle highly unpredictable user queries, extract entities dynamically without predefined slots, and generate conversational responses based purely on a set of external knowledge bases.
Which orchestration mode should you configure in Copilot Studio to meet these requirements?
Show answer & explanation
Correct answer: A
Generative AI orchestration allows the agent to dynamically route intents, extract entities without rigid slot filling, and generate contextual responses based on knowledge bases. Standard NLP and CLU require more rigid, predefined intents and entities, which does not suit highly unpredictable user queries.
- Question 8Advanced
Design AI-powered business solutions · Design agent extensibility in Copilot Studio, including Model Context Protocol and Computer Use
Your organization is building a Copilot Studio agent that needs to securely query a proprietary, on-premises vector database maintained by your data science team. The database exposes a standardized interface designed for AI agents to retrieve context dynamically during conversations.
To integrate this external data source seamlessly into your Copilot Studio agent's reasoning loop, which extensibility framework should you implement?
Show answer & explanation
Correct answer: A
The Model Context Protocol (MCP) is an open standard that allows AI models and agents to securely connect to external data sources and tools. Implementing an MCP server for the proprietary vector database allows the Copilot Studio agent to dynamically fetch context during its reasoning loop without hardcoding complex API integration logic.
sequenceDiagram participant Agent as Copilot Studio Agent participant MCP as MCP Server participant DB as Vector Database Agent->>MCP: Request Context (Query) MCP->>DB: Fetch relevant vectors DB-->>MCP: Return data MCP-->>Agent: Standardized Context - Question 9Advanced
Design AI-powered business solutions · Design Copilot customizations, business terms, and connectors for Dynamics 365 customer experience, service, and Sales
CASE STUDY
Company Background:
Contoso Telecommunications operates a global support center using Dynamics 365 Customer Service and Dynamics 365 Contact Center. They experience high call volumes regarding billing discrepancies.Current Situation:
Human agents spend an average of 8 minutes per call manually verifying customer identity, checking billing history in an external system, and explaining the charges before resolving the issue.Requirements:
- Implement an AI agent to handle the initial customer interaction over voice channels.
- The agent must be able to securely authenticate the caller.
- The agent must query the external billing system to retrieve real-time data.
- If the agent cannot resolve the billing issue, it must seamlessly transfer the call to a human agent, passing along the full conversation context and gathered data.
flowchart LR Cust[Customer] -->|Voice Call| VoiceAgent[Channel Agent] VoiceAgent Auth[Auth Service] VoiceAgent Billing[Billing API] VoiceAgent -->|Transfer with Context| Human[Human Agent in D365]Which combination of technologies should you design to meet these requirements?
Show answer & explanation
Correct answer: A
To meet the requirements, you need a Copilot Studio agent with voice capabilities enabled. By integrating it as a channel agent within Dynamics 365 Contact Center, it can handle incoming voice calls, use Power Automate/Connectors to hit the billing API, and use native Omnichannel handoff capabilities to transfer the call (and context) to a human agent in D365 Customer Service.
- Question 10Beginner
Design AI-powered business solutions · Design task agents, autonomous agents, and prompt and response agents
You are designing an AI solution that monitors an inbox, reads incoming vendor invoices, extracts the data, and automatically enters the data into an ERP system without requiring human intervention (unless an anomaly is detected).
Which type of agent does this scenario describe?
Show answer & explanation
Correct answer: C
An autonomous agent operates independently in the background, triggering on events (like an email arriving), making decisions based on its instructions, and executing actions (entering data) without requiring a human to prompt it or guide its workflow. A prompt and response agent requires direct user interaction.
Ready for the real thing?
The full AB-100 simulator has every exam-style question, timed mode, and instant scoring.