An approval gate introduces a controlled checkpoint into an AI workflow, preventing execution of a designated action until an authorized reviewer approves it. Option C is correct because the workflow enters a suspended or waiting state while preserving the information required to resume execution.
For example, an AI agent may prepare a network configuration change and submit it to an operations engineer for authorization.
The workflow can preserve the proposed configuration, target devices, execution context, and approval request identifier while waiting for a decision.
Once approval is received, the workflow resumes according to its defined transition rules. Rejection may trigger modification, cancellation, or escalation.
Option A is incorrect because a human approval gate does not inherently delegate the decision to another AI agent.
Option B is incorrect because a properly implemented wait state does not require continuous model inference or token consumption. The platform may maintain stored state without repeatedly invoking the model.
Option D is incorrect because suspension does not inherently terminate the workflow.
Reliable approval workflows should include persistent state, reviewer authentication, audit logging, expiration policies, and clearly defined rejection behavior.
These mechanisms support predictable execution and enforce human oversight for sensitive operations.
Specific Cisco AITECH Course Guide References:
AI-Powered Workflows and Agentic AI
Exam Objective 6.4: Human-in-the-Loop strategies
Exam Objective 5.3: AI workflow design and monitoring
===================
Submit