How does Agentforce select the correct action to resolve a user's request?

Question

Options

• AEach topic contains a list of the matching action's user utterances so that the agent can map the
• BThe large language model (LLM) selects the right topic and action, if they exist. If there are no
• CThe reasoning engine identifies the agent action to be executed by its name and action input

• A
  10% (3)
• B
  86% (25)
• C
  3% (1)

Explanation

Agentforce uses a Large Language Model (LLM) as its core reasoning engine. When a user sends a message, the LLM evaluates all available topics and their associated actions, then selects the best-matching topic and action based on the natural language understanding of the user's intent. If no topic or action adequately matches the request, the LLM can either respond using its general knowledge, ask a clarifying question, or indicate that it cannot help. This design means Agentforce does not rely on rigid keyword-to-utterance mapping (Choice A) or solely on action names (Choice C); instead, it leverages the LLM's reasoning ability for flexible, intent-driven routing.

No community discussion yet for this question.