Salesforce Certified Agentforce Specialist Practice Exam

What the Salesforce Certified Agentforce Specialist exam covers

• AI Agents240 questions
• Prompt Engineering130 questions
• Data 360 Fundamentals144 questions
• Testing, Deployment, and Maintenance105 questions
• Multi-Agent Orchestration48 questions
• Governance and Observability67 questions

Free Salesforce Certified Agentforce Specialist practice test questions

A sample of 10 questions with answers and explanations. Sign up free to practice all 734.

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

   The correct model is agent to topics to instructions/actions: an Agentforce agent is the top-level container that holds one or more topics, and each individual topic carries its own instructions defining scope plus the actions it is allowed to invoke. This layered structure is what lets the Atlas Reasoning Engine take a broad agent purpose and break it into narrow, classifiable jobs - when a message comes in, the engine matches it to the best-fitting topic, then uses that topic's instructions and actions to build a grounded reply. For a brand-new agent, this is the very first design decision an admin makes: how to split the intended work into topics before writing instructions or wiring up actions. The other options each invert or scramble this hierarchy in a different way. Remember the order: agent owns topics, topics own instructions and actions, never the reverse.

   Why the other options are wrong

   • AActions are always scoped to a topic, not attached to the agent directly; there is no mechanism that lets an action bypass topics entirely.
   • BThis reverses the real hierarchy: an agent owns many topics, and each topic carries its own instruction set, not one instruction set shared and owned by all topics.
   • DInstructions belong to a topic and topics belong to the agent; instructions do not contain topics, and topics do not hold the agent record.
2. Question 2AI Agents

   What is a reasonable way to think about the difference between a Flow action and a standard CRM action when both could technically update a record?

   • AThey are functionally identical with no meaningful difference
   • BA standard CRM action always requires writing Apex first
   • CA Flow action can never touch a CRM record
   • DCRM action for a simple update; Flow for logic with stepsCorrect
   ✓ Correct answer: D

   A standard CRM action is meant for a quick, common update with no extra configuration, while a Flow action becomes the better choice once conditional logic, multiple steps, or additional processing needs to wrap around that same update. This distinction, complexity and configurability rather than raw capability, is the practical way to decide between them even though both can technically touch the same record. The two are not functionally identical since their build effort and flexibility differ meaningfully, standard CRM actions are declarative and require no Apex at all, and Flow actions routinely create or update CRM records as part of their logic rather than being unable to touch them. Takeaway: match the tool to the complexity of the update, not just to whether it can technically be done.

   Why the other options are wrong

   • AThe two differ meaningfully in configurability and how much logic they can wrap around an update, even though both can affect the same record.
   • BStandard CRM actions are declarative, point-and-click configurations; they do not require writing Apex first.
   • CFlow actions frequently create or update CRM records as a core part of their logic; they are not barred from touching records.
3. Question 3AI Agents

   A customer support director wants a conversation started in embedded web chat that escalates to land with a human agent who can see the full context in the Service Console. What makes this possible?

   • AOmni-Channel routing plus a preserved transcriptCorrect
   • BThe customer restarting the conversation on the phone
   • CA brand-new case created with no history
   • DDisabling the escalation action entirely
   ✓ Correct answer: A

   Two things need to happen together for a seamless hand-off: Omni-Channel routing has to connect the conversation to an available human agent, and the transcript or context gathered so far has to be preserved and surfaced in the Service Console, so the receiving agent can see what already happened rather than starting blind. The key concept is that routing alone only solves who receives the case; without the accompanying context, the human agent would still have to ask the customer to repeat everything, defeating the purpose of a smooth escalation. This is exactly what the customer support director wants, a chat that started in embedded web chat, escalated, and lands with full visibility for the human. Restarting the conversation on the phone would throw away the very context that was gathered in chat; creating a brand-new case with no history is the opposite of preserving context; and disabling the escalation action entirely would prevent any hand-off, human-visible context or otherwise, from happening at all. Takeaway: routing plus preserved context together, not either alone, make a seamless hand-off possible.

   Why the other options are wrong

   • BRestarting the conversation on the phone would discard the existing chat context rather than preserve it for the human agent.
   • CA brand-new case created with no history is the opposite of preserving the conversation's context for the receiving agent.
   • DDisabling the escalation action entirely would prevent any hand-off, with or without context, from occurring at all.
4. Question 4AI AgentsSelect all that apply

   An insurer's claims topic is launched from a mobile app that already has the policy number and the claim ID available. (Choose TWO) Which two statements correctly describe how input variables should be used here?

   • AInput variables let the calling app pass in the policy number and claim IDCorrect
   • BUsing input variables avoids unneeded questions and keeps the flow predictableCorrect
   • CInput variables are populated by actions and passed back to the caller
   • DInput variables only work inside Prompt Builder, never inside topics
   • EInput variables replace the need for the Einstein Trust Layer altogether
   ✓ Correct answer: A, B

   A and B correctly describe input variables here: they let the calling mobile app pass in the policy number and claim ID it already has directly into the claims topic, and doing so avoids the agent asking unneeded questions while keeping the conversation flow predictable from the first turn. This is the standard use of input variables, taking values the host application already knows and making them available to the topic's actions immediately, rather than having the model try to extract them from a live exchange with the customer. Input variables are not populated by actions and passed back to the caller, that describes an output variable instead, they are not restricted to Prompt Builder templates, they are a core part of configuring topics in Agent Builder, and they do not replace the Einstein Trust Layer, which handles data privacy separately and independently. C, D, and E each misdescribe input variables' direction, scope, or purpose. The takeaway: input variables bring known data in, predictably and without redundant questions.

   Why the other options are wrong

   • CThat describes an output variable, which carries a value from the topic's actions back out to the caller, not an input variable.
   • DInput variables are a core part of configuring agent topics in Agent Builder, not limited to Prompt Builder templates.
   • EVariables control conversation flow and predictability; they do not provide the data-privacy protections that the Einstein Trust Layer provides.
5. Question 5Data 360 Fundamentals

   A law firm wants an agent to pull a Matter along with its related Documents and Parties in a single grounded response, reflecting how these records relate in the business. What should the firm's Data Cloud admin build?

   • AA new custom CRM object
   • BA batch data stream
   • CA Prompt Builder Flex template
   • DA data graph rooted at MatterCorrect
   ✓ Correct answer: D

   A data graph rooted at the Matter DMO, with Document and Party nested beneath it, provides this related, hierarchical structure for the agent to ground on, matching how these records genuinely relate in the firm's business, a matter has documents and involved parties. Because the requirement explicitly calls out reflecting real business relationships in a single grounded response, a data graph is the structure purpose-built for exactly this, rather than treating Matter, Document, and Party as three unrelated lookups. The other options cannot deliver this: a new custom CRM object stores its own fields without relating existing DMOs, a batch data stream only ingests source data, and a Prompt Builder Flex template generates grounded text without defining any relationships. Takeaway: reflecting genuine business relationships across DMOs in one call is what a data graph is for.

   Why the other options are wrong

   • AA new custom CRM object stores its own fields, it does not relate existing DMOs together.
   • BA batch data stream ingests source data, it does not model relationships between DMOs.
   • CA Prompt Builder template generates grounded text, it does not define DMO relationships.
6. Question 6Data 360 Fundamentals

   A pharmaceutical company's medical-information agent must never answer clinical questions from the LLM's general knowledge alone; every answer must trace to an approved reference retrieved from Data Cloud. What configuration principle enforces this?

   • AConfigure the agent with the maximum number of topics possible
   • BGround each relevant topic's action via an approved retrieverCorrect
   • CSet the LLM's temperature to its highest possible value
   • DRely on the Testing Center to rewrite ungrounded answers
   ✓ Correct answer: B

   The only way to guarantee every answer traces to an approved reference is to ground the relevant topic's action itself with a retriever scoped to that approved content; without that grounding, the LLM is free to answer from its general training, which is precisely what must never happen for clinical information. The key concept is that grounding is applied per topic action, so each action that should be source-based needs its own retriever configured against the approved reference material. This matters here because clinical guidance carries real risk if it is wrong, so the agent needs a structural guarantee, not just good instructions. The other options, maximizing topic count, raising temperature, or relying on post-hoc testing, do not enforce sourcing and one of them, higher temperature, actively works against it. Remember: grounding is enforced per action, not by volume of topics or by testing after the fact.

   Why the other options are wrong

   • AThe number of topics configured does not, by itself, force any topic's action to be grounded in approved retrieved content.
   • CRaising the LLM's temperature increases response variability and creativity, which works against consistent, tightly source-based clinical answers rather than enforcing them.
   • DThe Testing Center evaluates results during development and QA; it does not rewrite or correct ungrounded answers once the agent is live in production.
7. Question 7Testing, Deployment, and Maintenance

   A utility company's outage-reporting agent passed 30 of 32 test cases in the Testing Center. The 2 failures both involved the topic correctly firing but the wrong action being invoked. Where is the fix most likely needed?

   • AThe action configuration or ordering within the outage topicCorrect
   • BThe permission set assigned to the Agent User
   • CThe change set used for the last deployment
   • DThe org-wide default sharing setting for Accounts
   ✓ Correct answer: A

   With 30 of 32 cases passing and the 2 failures both showing the correct topic but the wrong action, the fix is almost certainly in how that outage topic's actions are configured or ordered, since the reasoning engine already classified the utterances correctly and the problem sits downstream in action selection. Reviewing the Agent User's permission set would only affect data access, not which action the topic chooses. A prior change set used for deployment describes what was moved between orgs, it says nothing about current live action logic. Org-wide default sharing for Accounts controls record visibility, unrelated to how a topic selects among its actions. Focusing the review on the topic's own action configuration is the direct, efficient path to resolving an action-selection failure like this one.

   Why the other options are wrong

   • BThe Agent User's permission set governs data access, not which action a correctly classified topic chooses to invoke.
   • CA change set from a prior deployment records what was moved between orgs; it has no bearing on live action-selection logic.
   • DOrg-wide default sharing for Accounts controls record visibility, not which action a topic selects once it has classified correctly.
8. Question 8Testing, Deployment, and MaintenanceSelect all that apply

   A school district's admin is preparing a change set to promote a new 'Enrollment Status' topic to production. Which two components must be included so the topic works once deployed? (Choose TWO)

   • AThe prompt template the topic invokes for grounded answersCorrect
   • BThe custom actions the topic relies onCorrect
   • CThe district's fiscal year calendar setting
   • DThe default language of the Setup menu
   • EThe color theme of the parent portal
   ✓ Correct answer: A, B

   A topic is only as functional as what it depends on, so both the prompt template it invokes for grounded answers and the custom actions it relies on must be included in the same change set for Enrollment Status to work once deployed. The key concept is metadata dependency completeness: a topic's own definition can deploy successfully while still being non-functional if a referenced component is left out of the package. For a school district promoting a new topic, missing either the template or an action would leave part of the topic's behavior broken in production even though the topic itself appears to deploy fine. A fiscal year calendar, the Setup menu language, and a portal color theme are unrelated org-level or cosmetic settings, none of which affect whether the topic's own metadata dependencies are complete. Takeaway: include every prompt template and action a topic depends on in its deployment package, not just the topic definition itself.

   Why the other options are wrong

   • CThe district's fiscal year calendar setting has no bearing on whether the Enrollment Status topic's own metadata dependencies deploy successfully.
   • DThe Setup menu's default display language is an admin preference unrelated to whether the topic's actions or prompt template function correctly.
   • EThe parent portal's color theme is a cosmetic styling choice with no connection to the topic's underlying metadata dependencies.
9. Question 9Prompt Engineering

   A company's compliance team wants assurance that giving a marketing user the ability to run prompt templates does not let that user see Opportunity data they are not otherwise permitted to view. What principle addresses this concern?

   • APrompt template execution respects the running user's existing accessCorrect
   • BPrompt templates always run as a dedicated integration user
   • CMarketing users automatically inherit administrator access
   • DField-level security is disabled during prompt execution
   ✓ Correct answer: A

   The principle that addresses this concern is that prompt template execution respects the running user's existing access, meaning giving a marketing user the ability to run a template does not grant them any new visibility into Opportunity data beyond what their own permissions already allow. This is the same record-level and field-level security model that governs every other feature, a prompt template is just another way of surfacing data, not a separate access path that bypasses existing rules. The compliance team's concern is exactly about unintended data exposure, and the correct reassurance is that the marketing user's own object, record, and field permissions still fully determine what the generated output can include. Templates do not run as a separate integration user by default, there is no automatic inheritance of admin access, and field-level security remains enforced throughout execution. Running a prompt template never grants access beyond what the user already has.

   Why the other options are wrong

   • BWhile an agent's own internal actions may sometimes run under a dedicated agent user in certain contexts, that is a different scenario from a user directly running a template themselves.
   • CThere is no automatic inheritance of administrator access for marketing users or any other user role.
   • DField-level security remains fully enforced during prompt execution, it is not disabled just because a template is running.
10. Question 10Prompt Engineering

    A prompt template written specifically around Account fields cannot be reused for Contact records without a rewrite. What change would make the template reusable across both objects?

    • AIncrease the character limit
    • BGeneralize wording and rely on input variablesCorrect
    • CAdd a stricter persona instruction
    • DSwitch to Data Cloud retriever grounding
    ✓ Correct answer: B

    The fix is to generalize the wording and drive the template's field-specific behavior through input variables, which Prompt Builder resolves against whichever object the resource is configured for. A template's reusability across object types depends on how it references data: if the instructions hardcode Account-specific concepts and field names, the template breaks the moment Contact records are passed in instead. Rewriting the wording in generic terms and letting variables carry the object-specific values means the same template design serves both objects without a rewrite. This is exactly the gap in the scenario, the template was authored around Account fields specifically. Character limits, stricter persona instructions, and switching the grounding retriever do not touch this object-coupling problem, only the wording and variable design do. Remember: build templates around inputs, not literal object names.

    Why the other options are wrong

    • AIncreasing the character limit changes how much text the template holds, not whether its wording depends on Account-specific fields.
    • CA stricter persona instruction changes tone and formality, it does nothing about the field references that couple the template to one object.
    • DSwitching to Data Cloud retriever grounding changes where data comes from, not whether the instructions themselves are written generically enough to work across objects.

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
• How these questions are written and reviewedMethodology
• Report a problem with a questionCorrections
• Salesforce Certified Business Analyst practice examRelated
• Salesforce Certified Data 360 Consultant practice examRelated
• Salesforce Certified Platform Administrator (ADM-201) practice examRelated

Salesforce Certified Agentforce Specialist practice exam FAQ