AI-201 · Question #206
Coral Cloud Resorts needs to ensure its booking agent executes actions in a specific sequence: first retrieve available sessions, then verify customer eligibility, and finally create the booking…
The correct answer is B. Create custom variables that store completion status for each step, then implement conditional. LLMs are non-deterministic by nature and may reorder steps even with numbered topic instructions (A), because instruction-following does not guarantee execution order. The reliable approach is to implement custom variables that store the completion status of each step and use…
Question
Coral Cloud Resorts needs to ensure its booking agent executes actions in a specific sequence:
first retrieve available sessions, then verify customer eligibility, and finally create the booking. The current implementation allows the large language model (LLM) to execute these actions in any order, causing booking failures. Which approach should an AgentForce Specialist implement?
Options
- AWrite comprehensive topic instructions detailing the exact sequence of actions using numbered
- BCreate custom variables that store completion status for each step, then implement conditional
- CConfigure topic, classification description, and action instructions with priority levels and
How the community answered
(36 responses)- A19% (7)
- B72% (26)
- C8% (3)
Explanation
LLMs are non-deterministic by nature and may reorder steps even with numbered topic instructions (A), because instruction-following does not guarantee execution order. The reliable approach is to implement custom variables that store the completion status of each step and use conditional logic so each subsequent action is only available (or enabled) after the previous one has been marked complete. This creates a hard dependency chain that the agent cannot bypass. Priority-level configurations (C) in topic classification descriptions influence topic selection, not intra-topic action sequencing.
Topics
Community Discussion
No community discussion yet for this question.