Praxis Academy
Home › Platform Administrator › Agentforce AI

Platform Administrator practice questions: Agentforce AI

Agentforce AI is 8% of the Salesforce Certified Platform Administrator exam. It covers Einstein Trust Layer, topics, actions and agent testing. Try these 4 free questions, then open each answer to see why every option is right or wrong.

01

Why does the same test run of a prompt template produce slightly different wording each time?

  1. Templates only call the LLM after deployment.
  2. The system simulates responses locally.
  3. Prompt Builder sends a fresh request to the LLM on every test run.
  4. Prompt Builder caches earlier outputs, causing differences.
Show answer

Answer: C

A. Templates only call the LLM after deployment.Tests do call the LLM.
B. The system simulates responses locally.Responses are not simulated.
C. Prompt Builder sends a fresh request to the LLM on every test run. ✓Correct.
D. Prompt Builder caches earlier outputs, causing differences.Caching would give identical results.

Why: Each test sends a real request to the model, and LLM output is probabilistic.

02

Titan Industrial's admin is building a prompt template for ticket summaries and wants it to include the customer's account name, equipment serial number and issue details from the case. What pulls record data into the template?

  1. Global picklist value sets
  2. Merge fields
  3. Custom metadata types
  4. Autolaunched flow variables
Show answer

Answer: B

A. Global picklist value setsValue sets share picklist values.
B. Merge fields ✓Correct.
C. Custom metadata typesCustom metadata holds configuration records.
D. Autolaunched flow variablesFlow variables are for flows, not prompt templates.

Why: Merge fields in Prompt Builder insert record data dynamically into the prompt.

03

In an agent's reasoning loop, every attempted path gives unsatisfactory or incomplete results. What does the agent do next?

  1. Ask the user for more information or clarification.
  2. Invent an answer to satisfy the request.
  3. Show a fatal system error.
  4. End the session automatically.
Show answer

Answer: A

A. Ask the user for more information or clarification. ✓Correct.
B. Invent an answer to satisfy the request.Fabricating answers is unsafe.
C. Show a fatal system error.Failures do not crash the agent.
D. End the session automatically.It does not terminate the session.

Why: With missing context or ambiguity, the agent asks the user for clarification.

04

Meridian Health deploys Agentforce Service Agents. What is the most important prerequisite for a complete, real-time unified view of customer data?

  1. Set up Email-to-Case routing addresses.
  2. Make sure Data Cloud is implemented and connected.
  3. Assign custom profiles to the agent.
  4. Design custom utility bar components.
Show answer

Answer: B

A. Set up Email-to-Case routing addresses.Routing addresses do not unify data.
B. Make sure Data Cloud is implemented and connected. ✓Correct.
C. Assign custom profiles to the agent.Profiles do not give a unified view.
D. Design custom utility bar components.Utility bar components are UI only.

Why: Data Cloud unifies data from many systems into a real-time profile that grounds the agent.

Find out if you're ready to pass

Take the free Platform Administrator diagnostic: 20 questions, no account needed. Then unlock all 320 questions, full practice exams and the study guide.

Start the free diagnosticSee plans

Other sections of this exam