Free AB-100 sample questions

Correct answer: Plan

In the Cloud Adoption Framework for AI, the 'Plan' phase involves evaluating technical feasibility, selecting the first workloads, and building a proof of value (PoV) or proof of concept. The Strategy phase is for defining business justification and expected outcomes. The Ready phase focuses on preparing the environment (landing zones), and the Adopt phase is where full-scale development and deployment occur.

Correct answers: Implement a model router in Microsoft Foundry to dynamically direct prompts.; Deploy a combination of Large Language Models (LLMs) and Small Language Models (SLMs).

To optimize TCO, a model router dynamically directs incoming prompts to the most appropriate model based on complexity, intent, or cost parameters. By pairing this with a mix of Large Language Models (LLMs) for complex tasks and Small Language Models (SLMs) for basic tasks, the organization avoids paying premium LLM costs for simple summarization.

To optimize TCO, a model router dynamically directs incoming prompts to the most appropriate model based on complexity, intent, or cost parameters. By pairing this with a mix of Large Language Models (LLMs) for complex tasks and Small Language Models (SLMs) for basic tasks, the organization avoids paying premium LLM costs for simple summarization. flowchart LR Req[User Request] --> Router{Model Router} Router -->|Complex Reasoning| LLM[GPT-4o] Router -->|Basic Summarization| SLM[Phi-3 Mini] LLM --> Resp[Response] SLM --> Resp

Correct answer: Cleanliness

Cleanliness refers to the absence of duplicates, formatting errors, and conflicting information in the dataset. Poor cleanliness directly impacts an agent's ability to provide coherent and accurate responses when grounding on that data. Timeliness refers to data being up-to-date, relevance refers to the data matching the domain, and availability refers to the system's ability to access the data.

Correct answer: Extend Microsoft 365 Copilot by creating a declarative agent in Copilot Studio with an API plugin connected to the on-premises engine via an on-premises data gateway.

Extending Microsoft 365 Copilot using a declarative agent with an API plugin is the optimal choice. It meets the requirement to surface in Teams, leverages existing M365 Copilot investments (avoiding the cost of building a fully custom UI/orchestrator), and can securely access the on-premises pricing engine via a standard API plugin routed through an on-premises data gateway. Building a fully custom agent ignores the requirement to leverage existing M365 investments.

Correct answer: Establishing a centralized repository for responsible AI guidelines, reusable agent templates, and cross-departmental AI governance.

A Microsoft AI Center of Excellence (CoE) focuses on establishing governance, sharing best practices, providing reusable assets (like templates and prompt libraries), and ensuring responsible AI adherence across the organization. It is meant to empower fusion teams, not bottleneck all AI development to a single centralized developer team.

Correct answer: Small Language (SLM)

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.

Correct answer: Platform licensing, development effort, and maintenance costs vs. helpdesk ticket deflection rate and average time-to-resolution savings.

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.

Correct answer: False

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.

Correct answer: Microsoft Foundry (via Azure AI Foundry tools and orchestrators)

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.

Correct answer: Common Data Model (CDM) standard entities and well-defined table relationships

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.