1. A solutions architect needs to deploy a prebuilt Microsoft 365 Copilot agent for IT helpdesk scenarios. Before development begins, which TWO activities are MOST important when defining the solution rules and constraints for the prebuilt agent?
- A. Specifying which data sources the agent is permitted to access and which it must not access.✓ Correct
- B. Retraining the underlying large language model on company-specific IT tickets.
- C. Defining escalation rules — the conditions under which the agent must hand off to a human technician.✓ Correct
- D. Selecting the physical data center region where Microsoft stores its LLM weights.
- E. Designing a custom neural network architecture to replace the prebuilt agent's inference engine.
Explanation
Correct - Data source permissions: Defining permissible data sources ensures the agent only accesses authorized knowledge bases and does not inadvertently expose sensitive systems, which is a fundamental constraint. Correct - Escalation rules: Determining handoff conditions (e.g., unresolved tickets, security incidents) ensures the agent operates within safe boundaries and maintains human oversight where needed. | Wrong - Retraining the LLM: Prebuilt agents use existing foundation models; retraining the underlying LLM is neither required nor typically possible for prebuilt agent deployments. | Wrong - Physical data center region for LLM weights: This is a Microsoft infrastructure concern outside the architect's control and is not part of defining agent rules and constraints. | Wrong - Custom neural network architecture: Replacing the prebuilt agent's inference engine contradicts the premise of using a prebuilt agent and is out of scope for this activity.