NewDumps是個可以滿足很多客戶的需求的網站。有些使用我們類比測試軟體已經通過相關IT認證考試的人成為了NewDumps的回頭客。NewDumps可以提供領先的Microsoft 培訓技術助你通過Microsoft AB-620 認證考試。
| Section | Weight | Objectives |
|---|---|---|
| Plan and configure agent solutions | 30–35% | - Plan an agent solution
|
| Test, deploy, and manage agents | 20–25% | - Deploy and manage agent lifecycle
|
| Build and extend agents in Copilot Studio | 40–45% | - Implement advanced capabilities
|
NewDumps的關於Microsoft AB-620 認證考試的針對性練習題卻是很受歡迎的。NewDumps的資料不僅能讓你獲取知識和很多相關經驗，還可以使你為考試做充分的準備。雖然Microsoft AB-620認證考試很難，但是通過做NewDumps的練習題後，你會很有信心的參加考試。放心地選擇NewDumps的高效練習題吧，為Microsoft AB-620 認證考試做一個最充分的準備。
問題 #57
A team is preparing to evaluate an agent in Copilot Studio before expanding access to additional users.
The team must choose an evaluation method that meets the following requirements:
Use a fixed set of prepared interactions.
Determine responses against a predefined baseline.
Support consistent comparison across repeated test runs.
You need to determine the evaluation method that satisfies the evaluation requirements.
答案：
解題說明：
Explanation:
Fixed prepared interactions # Use a curated prompt list; Predefined baseline # Compare against predefined expected outcomes; Repeatable comparison # Use a consistent scoring evaluation matrix.
Comprehensive and Detailed Explanation From Microsoft AB-620 Study Guide: A curated prompt list creates the controlled input population required for repeatability. Predefined expected outcomes establish what a correct or acceptable response should contain, which lets the evaluator judge performance against a baseline rather than against a reviewer ' s memory. A stable scoring matrix ensures that repeated runs are interpreted under the same thresholds and dimensions. Live conversations and post-rollout feedback can enrich the test inventory, but they cannot replace a fixed regression set because the questions, users, and context change. Publishing status confirms deployment, not response quality. General impressions are similarly unsuitable for consistent comparison. The team should select the scoring method according to the response type: semantic comparison for valid paraphrases, exact match for immutable values, keyword checks for mandatory phrases, and custom criteria for compliance or domain quality. Versioning the prompt list, expected answers, rubric, agent version, and knowledge snapshot is essential; otherwise a changed test set can be mistaken for a change in agent performance. Study Guide alignment: Test and manage agents > Evaluate agent performance > Create a test set; Choose an evaluation method.
問題 #58
You are testing an agent in a development environment and need to verify how it responds across multiple conversation paths before publishing to production. What should you use?
答案：D
解題說明：
The built-in Test panel lets builders simulate conversations, inspect variable values, and trace which topics/nodes fired, without affecting the published production agent.
問題 #59
Hotspot Question
A team uses environment variables to separate configuration from an agent so that the same solution can be moved between environments.
The team requires the following scenarios to be addressed:
- An administrator updates a sensitive configuration value and needs
the change to take effect at runtime without republishing the agent.
- A builder needs to confirm a configuration value but must not be able to modify it from Copilot Studio.
- An administrator updates a non-secret configuration value and needs
the updated value to be used by the published agent at runtime.
You need to apply the correct handling behavior for different environment-variable scenarios.
Which behavior should you apply for each scenario? To answer, select the appropriate options in the answer area.
NOTE: Each correct selection is worth one point.
答案：
解題說明：
Explanation:
Box 1: Use a secret environment variable type to allow for runtime resolution.
Runtime Resolution: Secret environment variables in Power Platform / Copilot Studio link directly to Azure Key Vault.
Zero Republishing: When the administrator updates the value in Azure Key Vault, the agent fetches the latest value securely at runtime without needing a manual republish.
Security Compliance: Sensitive data (like API keys or passwords) is never exposed in plain text within the Copilot Studio authoring canvas.
Box 2: Environment variables are read-only in Copilot Studio
You must use environment variables configured as read-only within Copilot Studio to meet this requirement.
Prevention: It prevents makers or builders from changing values directly inside the bot authoring canvas.
Separation: It keeps the configuration layer decoupled from the conversational logic.
Security: It ensures the agent behaves consistently when moved across development, testing, and production environments.
Box 3: The agent automatically pulls the updated Current Value from the environment at the start of every session To ensure the published agent uses the updated configuration value at runtime after an administrator changes it, the agent must automatically pull the updated current value from the environment at the start of every session.
Dynamic Fetching: Fetching the variable at the start of a session ensures the agent never relies on stale, hardcoded data.
Zero Redeployment: Administrators can change values (like a support email or API endpoint URL) in the environment settings without needing to edit, recompile, or republish the agent itself.
Environment Isolation: The agent's logic remains identical across Development, Testing, and Production, changing its behavior strictly based on the environment context it loads.
Reference:
https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-variables-about
問題 #60
Case Study 1 - Blue Yonder Airlines
Background
Blue Yonder Airlines is a global carrier headquartered in Los Angeles, California, operating domestic and international flights. The company serves millions of passengers annually through its website, mobile app, and call centers. To improve customer service efficiency and reduce call center volume, Blue Yonder is deploying an AI agent in Microsoft Copilot Studio.
The agent will handle customer inquiries across multiple channels - web chat, mobile app, and Microsoft Teams (for internal support staff). It will answer questions, retrieve data from enterprise systems, and escalate to human agents when needed.
The project is led by a cross-function team:
- Product manager: Defines requirements and success metrics.
- Lead agent author: Designs topics, intents, and generative behavior.
- Flow designers: Build agent flows and integrations.
- IT/security and compliance: Oversees identity, data protection, and Responsible AI (RAI) compliance.
Current environment
Channels
Public website: Embedded web chat
Mobile app: In-app chatbot
Microsoft Teams: Internal support agent access
Identity and access
Customers: Anonymous access for general inquiries (e.g., flight status, baggage policy).
Authentication is required for personal data access (e.g., bookings, loyalty points).
Internal staff: Authenticate via Microsoft Entra ID.
Data sources
Reservation and Ticketing System (internal): REST API, no prebuilt connector with custom enterprise database.
Flight Status and Weather APIs (external): REST APIs with API keys.
Customer Support Knowledge Base: SharePoint library with PDFs and policy documents.
Loyalty Program Data: Stored in Dynamics 365 and Dataverse.
Travel Advisory Content: Uses REST API with partner services.
Integration mechanisms
Custom connectors must be used for internal APIs that lack prebuilt connectors.
HTTP request nodes may be used for lightweight external APIs.
Knowledge sources must be used for unstructured content.
Agent flows must be used to encapsulate reusable logic (e.g., rebooking).
Business requirements
Omnichannel support
Deploy the agent across web, mobile, and Teams with a consistent user experience. The Teams deployment must also support internal staff.
Self-service capabilities
The agent must handle common inquiries such as:
- Flight status
- Booking and rebooking
- Loyalty program questions
- Travel policies and baggage rules
Human escalation
If the agent cannot resolve an issue or the user requests help, it must:
- Escalate to a human agent.
- Transfer the conversation transcript and relevant context.
- Redact any sensitive personal data before escalation.
Knowledge integration
The agent must use scalable methods for knowledge integration and must not rely on manually authored Q&A topics for each document.
Performance metrics
First-contact resolution: +25%
Tier-1 call deflection: ≥20%
Response time: 90% of queries answered within 30 seconds
Accuracy: ≥95% for known FAQs
CSAT: ≥85% for AI-handled interactions
Technical requirements
Platform constraints
No custom code is permitted; only Copilot Studio's built-in tools may be used.
All backend logic must be implemented using agent flows.
Markdown must be used for formatting (e.g., bold, bullet points); HTML is not supported.
Authentication
Sign-in is required for personal data access.
Anonymous access is allowed for general inquiries.
User identity must be used for data access; shared or builder credentials must not be used.
Compliance and security
Power Platform DLP policies must be enforced to block unauthorized data flows.
Responsible AI content moderation filters must be enabled.
Prompt modifications must be added to enforce tone, disclaimers, and refusal behavior.
Disclaimers must be applied consistently across all generative responses. Manual edits to individual topics must be avoided.
Monitoring and maintenance
All conversations and actions must be logged for auditing.
Weekly reviews of transcripts and metrics must be conducted.
Topics, flows, and knowledge sources must be updated as policies or systems evolve.
Issues and constraints
API rate limits: External APIs (e.g., flight status) have usage limits. Agent flows must handle retries and caching to avoid exceeding quotas.
Knowledge base limits: Copilot Studio has limits on the number and size of indexed documents.
Large files must be split or summarized.
Generative answer risks: Generative responses must be constrained to avoid policy violations.
Prompt modifications and filters must be used to enforce tone, safety, and compliance.
User input variability: Users phrase questions in diverse ways. Topics must include varied trigger phrases and fallback handling.
Authentication UX: The agent must clearly explain when sign-in is required and handle transitions smoothly across channels.
Problem statement
Blue Yonder Airlines must deploy a secure, scalable, and policy-compliant AI agent using Microsoft Copilot Studio. The agent must deliver accurate, helpful, and safe responses across multiple channels, integrate with enterprise systems, and support both anonymous and authenticated users. It must adhere to strict data protection and Responsible AI standards while improving customer service efficiency and satisfaction.
Drag and Drop Question
You need to configure the agent in Copilot Studio to meet Blue Vender's Responsible AI and data protection requirements.
Which three actions should you perform in sequence? To answer, move the appropriate actions from the list of actions to the answer area and arrange them in the correct order.
答案：
解題說明：
Explanation:
Scenario: Problem statement: The agent must adhere to strict data protection and Responsible AI standards while improving customer service efficiency and satisfaction.
Step 1: Define Data Loss Prevention (DLP) policies
Security and compliance must be established at the foundation level. Admins use the Power Platform admin center to configure DLP policies. This establishes environmental guardrails by restricting unauthorized data connectors or preventing the agent from leaking sensitive data.
Step 2: Add custom prompt modifications
Once the environment is secure, the maker configures the specific behavior of the agent. Custom prompt modifications and agent instructions shape the tone, define conversational topics, and establish explicit boundaries for the generative model (e.g., instructing the model to strictly stay within its knowledge base).
Step 3: Conduct adversarial testing
This is the final evaluation phase before deployment. After the guardrails and custom prompts are in place, you simulate real-world attacks (such as jailbreaks, prompt injections, or malicious user inputs) to verify that the agent reliably adheres to your corporate guidelines and Microsoft's Responsible AI standards.
Reference:
https://learn.microsoft.com/en-us/troubleshoot/power-platform/copilot-studio/generative-answers/agent-response-filtered-by-responsible-ai
問題 #61
Here is the complete text extracted from the image:
A tenant administrator configures a Copilot connector that indexes content from an external knowledge system into Microsoft Graph.
A builder is building an agent that must:
Use the indexed enterprise knowledge during answer generation.
Avoid duplicating or manually uploading the same content.
Avoid invoking the system as a transactional tool.
You need to configure the agent so that responses are grounded in the authorized knowledge sources.
What should you do?
答案：B
解題說明：
Comprehensive and Detailed Explanation From Microsoft AB-620 Study Guide: The tenant administrator has already configured a Copilot connector that indexes the external system into Microsoft Graph. The agent should consume that approved enterprise knowledge source from its Knowledge configuration. This lets generative answers retrieve authorized indexed content without recreating the connector, uploading duplicate documents, or treating the source as a callable transaction. Copilot connectors are designed for enterprise search and knowledge grounding; the user's access is evaluated so that responses surface only content the user is permitted to see. A custom connector is appropriate when the agent must call an API or perform an operation, which the scenario explicitly excludes. Adding the connector as a tool would likewise shift the design from retrieval to invocation. Manual file upload would create a second, potentially stale copy and break the administrator-managed indexing lifecycle. After adding the source, the builder should provide a clear name and description, test representative questions, confirm citations or source attribution, and verify that restricted users cannot retrieve unauthorized material. Study Guide alignment: Integrate and extend agents in Copilot Studio > Connect to enterprise knowledge sources > Connect to Copilot connectors.
問題 #62
......
我們會在互聯網上免費提供部分關於Microsoft AB-620 認證考試的練習題讓嘗試，您會發現NewDumps的練習題是最全面的，是你最想要的。
AB-620學習資料: https://www.newdumpspdf.com/AB-620-exam-new-dumps.html