ADM-201 - Agentforce - Section 8.2

Maintain agents in Agent Builder and manage prompt templates, including configuring topics, actions, and prompt instructions for reliable agent responses.

Use Agent Builder to configure an agent's topics, instructions, and actions, and the Prompt Builder to create and manage grounded prompt templates that the agent and users invoke. Recognise how testing in the conversation preview and refining instructions improves the accuracy and safety of an agent's responses before it is activated.

Agent BuilderPrompt BuilderPrompt templatesAgent topics and actions

Practice question for this objective

Free sampleAgentforcemedium

At Albany Health, before activating a new Agentforce agent for customers, the administrator wants to trial sample utterances and confirm that the agent selects the correct topics and actions, without exposing it to real users. Which approach should the administrator use?

  • AUse the built-in conversation preview and testing tools in Agent Builder while the agent is still inactive. Correct
  • BActivate the agent in production for a short window and monitor its live conversations for mistakes.
  • CLoad a set of sample utterances with Data Loader and review the records it creates.
  • DEnable debug logs on the underlying flow and infer the agent's behaviour from those log entries.
Agent Builder's conversation preview and testing tools let an administrator validate topic and action selection before activation. Agent Builder includes a preview and testing capability for sending trial utterances and observing which topic and action the agent chooses. This validates behaviour while the agent is inactive, avoiding any exposure to real users that production activation would cause.

Why A is correct: Correct: Agent Builder provides a conversation preview and testing surface that lets the administrator send sample utterances and inspect topic and action selection before the agent goes live.

Why B is wrong: Activating in production does expose the agent to real users, which is exactly what the requirement rules out, so it is the wrong way to trial it safely.

Why C is wrong: Data Loader is a bulk data tool for importing and exporting records and has no role in exercising an agent's reasoning or routing.

Why D is wrong: Flow debug logs show only what one action did if it ran, and cannot demonstrate whether the agent classified the request into the right topic in the first place.

See more ADM-201 practice questions with worked answers.

More in this domain

Back to all Agentforce objectives, or the ADM-201 cert hub.

Examworthy is not affiliated with or endorsed by Salesforce. Original, blueprint-aligned practice material only.