Salesforce Certified Agentforce Specialist Practice Exam

What the Salesforce Certified Agentforce Specialist exam covers

• AI Agents240 questions
• Prompt Engineering130 questions
• Data Cloud for Agentforce144 questions
• Development Lifecycle107 questions
• Multi-Agent Interoperability48 questions

Free Salesforce Certified Agentforce Specialist sample questions

A sample of 10 questions with answers and explanations. Sign up free to practice all 669.

1. Question 1AI Agents

   A Salesforce admin is configuring a brand new Agentforce agent for the first time. Which statement correctly describes how the agent is structured in Agent Builder?

   • AActions attach to the agent directly, bypassing topics
   • BTopics contain agents; each agent has one instruction set
   • CAgent contains topics; each holds instructions and actionsCorrect
   • DInstructions contain topics; topics hold the agent record
   ✓ Correct answer: C

   In Agent Builder, an agent is a container for one or more topics. Each topic has instructions that define its scope and one or more actions it can invoke. This layered structure lets the Atlas Reasoning Engine classify an incoming utterance to the right topic, then use that topic's instructions and actions to build a response.

   Why the other options are wrong

   • AActions are assigned to topics, not attached to the agent as a flat list that bypasses topics.
   • BThis reverses the real hierarchy; an agent owns topics, not the other way around.
   • DInstructions belong to a topic, not the reverse; topics are not nested inside instructions.
2. Question 2AI Agents

   A city government's permit-renewal topic needs to check whether a fee is outstanding before allowing a renewal request to proceed. How should this be modeled?

   • ASkip the fee check since actions cannot check conditions
   • BPerform the check only through the Trust Layer audit trail
   • CAssign a fee-lookup action, guided by the topic's instructionsCorrect
   • DRequire every citizen to email the finance department
   ✓ Correct answer: C

   By assigning a fee-status action and writing instructions that direct the topic to check for outstanding fees before proceeding, the reasoning engine can appropriately gate or inform the renewal step.

   Why the other options are wrong

   • AActions and the reasoning engine's sequencing can absolutely incorporate conditional checks like this.
   • BThe audit trail logs prompts and responses for governance; it does not perform business logic checks.
   • DRequiring a manual email step bypasses the automation the agent is meant to provide.
3. Question 3AI Agents

   An admin wants an agent to route escalated banking disputes specifically to the fraud team's queue rather than general support. What should be configured?

   • AA generic escalation with no routing configuration
   • BA new agent type just for fraud
   • COmni-Channel routing to a fraud-specific queueCorrect
   • DA Prompt Builder template for fraud
   ✓ Correct answer: C

   Omni-Channel routing lets an admin direct escalated conversations to a targeted queue or skill group, such as a fraud team, rather than a general support pool.

   Why the other options are wrong

   • AA generic escalation with no targeting would not guarantee fraud specialists receive the case.
   • BA new agent type is unrelated to configuring where escalations route.
   • DPrompt Builder templates generate content for prompts, not escalation routing.
4. Question 4AI Agents

   A Salesforce admin is troubleshooting why a service agent gave an unexpected answer to a customer. To see the exact prompt sent to the model and the exact response returned, which Trust Layer capability should the admin consult?

   • AZero data retention
   • BToxicity detection
   • CAudit trailCorrect
   • DPII masking
   ✓ Correct answer: C

   This record is the standard way to debug unexpected agent behavior, since it shows what was actually sent to and returned from the LLM.

   Why the other options are wrong

   • AZero data retention is about the provider discarding requests, not about giving Salesforce a reviewable record.
   • BToxicity detection is a real-time content filter; it does not produce a log for debugging.
   • DPII masking changes what is sent to the LLM, it does not itself provide a way to review the interaction afterward.
5. Question 5Data Cloud for Agentforce

   An automotive OEM wants dealer inventory files stored in a cloud storage bucket to flow into Data Cloud on a recurring schedule. Which component should be configured?

   • AA relationship data graph
   • BA Prompt Builder Flex template
   • CA data stream tied to storageCorrect
   • DAn Agentforce custom action
   ✓ Correct answer: C

   A data stream connected to an external cloud storage connector ingests files on a schedule, landing them as Data Lake Objects that can later be mapped and used for grounding.

   Why the other options are wrong

   • AA data graph relates already-modeled DMOs, it does not pull in raw files.
   • BA Prompt Builder template is a grounding technique, not a file ingestion tool.
   • DA custom action executes agent logic, it does not ingest external files.
6. Question 6Data Cloud for Agentforce

   A city agency's citizen-services agent uses an ensemble retriever spanning permit-application guides and zoning-code documents. During testing, answers about permits occasionally pull in irrelevant zoning content. What retriever-level change would most directly help?

   • AIncrease the LLM's temperature setting for responses
   • BRemove the search index for permit content entirely
   • CTighten filters so each source stays relevantCorrect
   • DDisable the Agent User's assigned permission set
   ✓ Correct answer: C

   When an ensemble retriever mixes content from multiple sources inappropriately, adjusting filters or relevance thresholds so each underlying retriever contributes only clearly relevant content per question directly addresses the mixing problem.

   Why the other options are wrong

   • ATemperature affects the wording style of the generated response, not which grounding content is retrieved.
   • BRemoving the index eliminates needed permit content entirely rather than fixing the mixing issue.
   • DPermission sets govern data visibility, not topical relevance mixing between two content sources.
7. Question 7Multi-Agent Interoperability

   An airline's own rebooking agent needs to jointly rebook a codeshare passenger together with a partner airline's separate agent, built on a different vendor's platform. What enables this cross-vendor collaboration?

   • ASharing one Data Cloud org
   • BSharing one list view
   • CUsing the A2A protocolCorrect
   • DOne shared Prompt Template
   ✓ Correct answer: C

   Because the two agents are built by different vendors, they need a shared protocol, not shared org data, to hand off and coordinate on the joint codeshare rebooking task.

   Why the other options are wrong

   • ASharing a Data Cloud org does not give two independently built agents a way to coordinate a task.
   • BA list view is a UI record display, unrelated to cross-vendor agent coordination.
   • DA single shared Prompt Template does not let two separate agent platforms hand off a task to each other.
8. Question 8Development Lifecycle

   A car rental company's roadside-assistance agent, while sandbox testing, occasionally responds with information from a policy that was deprecated last quarter. What is the underlying lesson for interpreting this test result?

   • AGrounded answers are only as current as the underlying source content, so source data must be kept up to dateCorrect
   • BThe agent invented the deprecated policy without any source, which no amount of testing could ever catch
   • CThe Testing Center itself introduced the outdated content
   • DThis indicates a permission set misconfiguration
   ✓ Correct answer: A

   Because grounding pulls from configured content, an outdated answer is a signal to update the Knowledge article, Data Cloud record, or other source, not a sign of a platform defect or a security issue.

   Why the other options are wrong

   • BGrounded responses are generated from real configured sources, not invented from nothing.
   • CThe Testing Center evaluates behavior, it does not inject content into sources.
   • DPermission sets govern data access, not whether content is current.
9. Question 9Development LifecycleSelect all that apply

   An esports company's Agentforce team wants to formalize a mature process for iterating on its live fan-support agent. Which three practices should be part of that process? (Choose THREE)

   • AValidate proposed changes in a sandbox before promoting themCorrect
   • BRe-run the regression test suite after each changeCorrect
   • CMaintain a record of configuration changes over timeCorrect
   • DApply all changes directly in production to save time
   • EAvoid testing changes that seem minor
   ✓ Correct answer: A, B, C

   These three practices ensure changes are verified before release, do not break existing behavior, and remain traceable over time, which is the discipline expected for an agent under continuous, ongoing iteration.

   Why the other options are wrong

   • DApplying changes directly in production skips verification and increases risk regardless of time saved.
   • EEven seemingly minor changes can have unintended effects and should still be tested.
10. Question 10Prompt Engineering

    A quality reviewer wants to confirm a prompt template's grounded merge fields resolve correctly for a specific test record before it goes live. Where should this be verified?

    • AIn the Testing Center for agent regression tests
    • BIn the test panel within Prompt Builder itselfCorrect
    • CIn the Data Cloud data stream monitor
    • DIn the Salesforce mobile app settings
    ✓ Correct answer: B

    The test panel in Prompt Builder runs the template against a chosen record and shows the resolved, generated output, which is exactly what the reviewer needs to check merge field resolution.

    Why the other options are wrong

    • AThe Testing Center is used for agent-level regression testing with utterances, not previewing a single template's merge fields.
    • CA data stream monitor tracks Data Cloud ingestion, not prompt output preview.
    • DMobile app settings have nothing to do with previewing prompt output.

Who this Salesforce Certified Agentforce Specialist practice exam is for

This practice set is for anyone preparing for the Salesforce Certified Agentforce Specialist exam - from first-time candidates building a foundation to experienced Salesforce practitioners doing a final review before test day. If you learn best by working through realistic questions and reading why each answer is right or wrong, it is built for you.

How to use this Salesforce Certified Agentforce Specialist practice exam

1. Start with the free sample questions above to gauge your current baseline.
2. Read the full explanation on every question, including why each wrong option is wrong.
3. Track your weak domains and focus your study where you are losing the most marks.
4. Once you are scoring consistently well, take a timed, full-length mock exam.
5. Use your readiness score to decide when you are ready to book the real Salesforce Certified Agentforce Specialist exam.

Related Salesforce resources

• Salesforce Certified Agentforce Specialist study guideKey concepts
• Salesforce practice examsAll Salesforce
• Certification pathWhere this fits
• Certification exam guides & tipsBlog
• Plans & pricingFree & paid
• Salesforce Certified Platform Administrator (ADM-201) practice examRelated
• Salesforce Certified Platform Developer I practice examRelated
• Salesforce Certified Platform App Builder practice examRelated

Salesforce Certified Agentforce Specialist practice exam FAQ