Courseiva

CCNA Agentforce Ai Questions

29 questions · Agentforce Ai topic · All types, answers revealed

Practice these questionsExam hubAll questions
1
MCQmedium

A Salesforce administrator is building an Agentforce Agent to assist customers with product recommendations. The agent must use a large language model to generate personalized suggestions based on customer purchase history. The administrator wants to ensure that sensitive customer data, such as credit card numbers, is not exposed to the external LLM. Which feature should the administrator configure?

A.Field-Level Security
B.Named Credentials
C.Einstein Trust Layer
D.Shield Platform Encryption
AnswerC

The Einstein Trust Layer is designed to securely handle data sent to external LLMs. It provides data masking, zero-data retention, and audit trails. By configuring masking rules for sensitive fields like credit card numbers, the administrator ensures that such data is replaced with placeholder tokens before being sent to the LLM. This protects customer privacy while still allowing the agent to generate recommendations based on non-sensitive data.

Why this answer

The Einstein Trust Layer is the correct feature to prevent sensitive data from being exposed to external LLMs. It allows administrators to define masking rules that replace sensitive information with placeholders before the data is sent. This ensures that the LLM never sees actual credit card numbers while still enabling the agent to generate personalized recommendations based on other data.

Exam trap

The trap here is confusing data-at-rest encryption or field-level security with data-in-transit masking for LLM interactions, which is uniquely handled by the Einstein Trust Layer.

Practice this question →
2
MCQeasy

A Salesforce administrator is configuring an Agentforce Agent to help sales reps find and update opportunities. The agent needs to use an action that updates the Opportunity Stage field. The administrator wants to ensure the agent only updates the stage when the user explicitly requests it. Which action type should the administrator use to achieve this with minimal development effort?

A.Prompt Template action
B.Flow action
C.Apex action
D.Einstein Trust Layer action
AnswerB

Flow actions are declarative and can be built quickly using Flow Builder. They allow the administrator to create a record update without code. By exposing a Flow as an action in Agent Builder, the agent can invoke it when the user requests a stage change. This meets the requirement with minimal development effort and is easy to maintain.

Why this answer

Flow actions are the most efficient way to enable an agent to update a field because they are declarative and require no code. The administrator can build a Flow that updates the Opportunity Stage and expose it as an action in Agent Builder. This allows the agent to invoke the Flow when the user requests it, satisfying the requirement with minimal effort.

Exam trap

The trap here is overcomplicating a simple update by choosing Apex when a declarative Flow action would suffice.

Practice this question →
3
Multi-Selectmedium

An administrator is preparing to activate a new Agentforce Service Agent that handles password reset requests for an internal help desk. Before go-live, the administrator wants to validate that the agent behaves correctly and safely. Which TWO testing practices should the administrator perform? (Choose two.)

Select 2 answers
A.Grant the agent's running user Modify All Data to ensure it can reset any user's password.
B.Activate the agent in a production org and monitor real customer conversations for the first week.
C.Use the Agentforce testing center to run predefined test cases against the agent and review pass/fail results.
D.Disable all topics except the password reset topic to force the agent to always use the correct flow.
E.Preview the agent in Agent Builder and manually converse with it using representative password reset utterances.
AnswersC, E

The Agentforce testing center lets admins define test cases with expected utterances and outcomes, then run them against the agent to catch regressions before activation. It provides structured, repeatable validation rather than ad hoc chat. For a password reset agent, this is the right way to confirm the agent routes to the correct topic and action without exposing credentials.

Why this answer

Pre-activation validation combines automated test cases in the Agentforce testing center with manual conversation in the Agent Builder preview. Together they verify topic routing, action execution, and response quality against realistic password reset utterances. Activating in production, disabling topics, or granting excessive permissions are not testing practices and introduce risk rather than confidence.

Exam trap

The trap here is confusing configuration changes or production monitoring with actual pre-activation testing, which must be done in a controlled environment.

Practice this question →
4
MCQhard

Northern Trail Outfitters has an Agentforce Service Agent that escalates complex cases to human agents. After go-live, the administrator notices the agent sometimes answers questions that belong to a different Topic, producing irrelevant responses. The administrator wants to control which Topic handles each utterance and reduce cross-topic confusion. Which configuration directly addresses this?

A.Refine each Topic's classification description and scope so the agent can distinguish which Topic an utterance belongs to.
B.Disable all but one Topic so the agent always uses the same instructions.
C.Lower the agent's model temperature to zero so it always picks the first matching Topic.
D.Increase the agent's temperature setting so it generates more varied responses.
AnswerA

Topics act as the agent's routing and scoping mechanism. Clear, distinct classification descriptions and well-defined scope help the model assign each utterance to the correct Topic, which directly reduces cross-topic answers. This is the supported lever for controlling topic selection during a conversation.

Why this answer

Topic selection depends on how clearly each Topic is described and scoped. When descriptions overlap or are vague, the agent can route an utterance to the wrong Topic and answer out of scope. Refining classification descriptions and tightening each Topic's boundaries gives the model the signal it needs to choose correctly, which directly reduces cross-topic responses.

Exam trap

The trap here is assuming that model generation settings such as temperature control which Topic an agent selects, when Topic routing is actually driven by classification descriptions and scope.

Practice this question →
5
MCQmedium

Which of the following is a primary benefit of using Agentforce over traditional chatbot solutions?

A.It requires less initial configuration time than traditional chatbots.
B.It provides autonomous decision-making and dynamic response generation.
C.It is cheaper to implement because it doesn't use Salesforce data.
D.It prevents the need for any human involvement in support processes.
AnswerB

Unlike traditional chatbots that rely on predefined decision trees and static responses, Agentforce uses generative AI to understand context and generate natural, dynamic responses. It can also autonomously decide which actions to take, making it far more capable of handling complex, non-linear customer interactions effectively.

Why this answer

Agentforce is built on generative AI, allowing for dynamic, context-aware responses rather than static, rule-based decision trees. This approach enables the agent to understand complex user intent, retrieve data from across the Salesforce platform, and perform actions autonomously. This evolution from static scripts to intelligent, functional agents significantly improves customer satisfaction and reduces the burden on human support staff by resolving complex issues independently.

Exam trap

Candidates often equate Agentforce with traditional rule-based chatbots, missing that its primary advantage lies in generative AI and autonomous decision-making capabilities.

Practice this question →
6
Multi-Selecthard

An administrator at Cosmic Enterprises is preparing an Agentforce Service Agent for a new return policy topic. Before activation, the security team requires that the agent never expose customer payment card digits and that every interaction be traceable for compliance review. (Choose two.)

Select 2 answers
A.Reduce the agent's topic scope so it only answers questions about store hours and shipping times.
B.Enable the Einstein Trust Layer audit trail to capture prompts, responses, and masking events.
C.Assign the agent user a profile with Modify All Data so it can retrieve any record a customer references.
D.Create a validation rule on the payment object that blocks saving records containing full card numbers.
E.Configure Data Masking rules in the Einstein Trust Layer for the fields that hold card digits.
AnswersB, E

The audit trail records interactions between Salesforce, the model, and connected systems, including which masking rules fired. This creates the traceability the security team demands, allowing reviewers to reconstruct what the agent was asked, what it returned, and whether sensitive data was masked, which is precisely the compliance visibility the scenario requires.

Why this answer

Meeting the security team's demands requires two distinct controls: masking to keep raw card digits away from the model, and auditing to prove what happened in each interaction. Data Masking rules in the Einstein Trust Layer handle the first, while the Trust Layer audit trail handles the second. Broad permissions, validation rules, and topic narrowing address adjacent concerns but provide neither the runtime protection nor the traceability the scenario specifies.

Exam trap

The trap here is accepting data-hygiene measures such as validation rules or narrower topics as substitutes for runtime masking and interaction auditing.

Practice this question →
7
MCQmedium

Northern Trail Outfitters has an Agentforce Service Agent whose order-status topic uses both a Flow action that queries order records and a prompt template that summarizes results. Users report the agent sometimes invents shipping dates that do not exist in the order record. The administrator verifies the flow returns correct data and the topic is correctly scoped. What should the administrator do to reduce this hallucination?

A.Add a second Flow action that queries the same order object to cross-check the first result.
B.Increase the agent's temperature setting so responses vary more naturally.
C.Enable Data Masking on the order's shipping date field so the agent cannot read it.
D.Adjust the prompt template instructions to require answers only from the flow action's returned values and to state when data is unavailable.
AnswerD

Hallucination in grounded topics usually stems from instructions that let the model fill gaps. Rewriting the prompt template to constrain the agent to the flow action's returned fields, and to say the information is unavailable rather than guessing, directly addresses the invented shipping dates while keeping the verified flow and topic scope unchanged.

Why this answer

When the underlying data source is verified and the topic is correctly scoped, fabricated details come from prompt behavior rather than retrieval. Constraining the prompt template to rely strictly on the flow action's returned values, and to acknowledge missing information instead of guessing, is the targeted remedy. Temperature changes, duplicate queries, and masking all act on the wrong layer and leave the model free to embellish.

Exam trap

The trap here is blaming the data source and adding redundant queries when the fabrication originates in unconstrained prompt instructions.

Practice this question →
8
MCQmedium

An administrator at Get Cloudy Consulting is building an Agentforce Service Agent to help customers reset their portal passwords. The agent must verify the customer's identity before allowing the reset. The administrator wants the agent to collect a one-time code sent to the customer's registered email and validate it against a record in Salesforce. Which Agentforce configuration should the administrator use to collect and validate the code within the conversation?

A.Create an Agent Action that invokes an Apex class annotated with @InvocableMethod, and add it to the relevant Topic.
B.Enable Einstein Activity Capture on the customer's contact record so the agent can read the code.
C.Add the code as a Prompt Template variable and let the agent compare it to the customer's email address.
D.Configure a Data Cloud segment that includes the customer's code and attach it to the agent's Topic.
AnswerA

An @InvocableMethod Apex action can accept the one-time code as an input, query the record, and return a validation result to the agent. This is the supported way to run custom server-side logic from an Agentforce Topic, so the agent can branch on the returned value. No other listed mechanism performs a server-side record check during the conversation.

Why this answer

Collecting a value in conversation and validating it against a Salesforce record requires deterministic server-side logic. An Apex action exposed through @InvocableMethod can take the collected code as input, query the record, and return a result the agent can act on. Prompt templates, Data Cloud segments, and Einstein Activity Capture do not perform real-time record validation.

Exam trap

The trap here is assuming that any Agentforce feature that can reference customer data, such as a prompt template or Data Cloud segment, can also perform a deterministic validation check.

Practice this question →
9
MCQmedium

An administrator at Northern Trail Outfitters created an Agentforce Service Agent that resolves billing inquiries. After activation, the agent successfully handles English chats but ignores Spanish chats entirely, even though the org's Einstein Translation and Language Settings are enabled. The administrator confirmed the agent's Employee Agent is assigned and the Messaging channel is live. What should the administrator do to allow the agent to respond to Spanish-speaking customers?

A.Enable the Einstein Translation for Messaging setting on the Messaging channel and refresh the embedded deployment.
B.Create a duplicate Service Agent with the Spanish locale and route Spanish chats to it using Omni-Channel routing rules.
C.Add the Spanish language to the agent's supported languages in Agentforce Studio and republish the agent.
D.Assign the agent's Einstein Agent User a permission set that includes the 'Agentforce Multilingual' user permission.
AnswerC

Agentforce agents maintain their own language configuration independent of org-wide translation settings. Adding Spanish under the agent's language settings in Agentforce Studio, then republishing, makes the agent capable of understanding and replying in Spanish. Org-level translation only translates metadata, not agent runtime conversations, so this agent-specific change is required for the scenario.

Why this answer

Agentforce agents have their own language settings that must be explicitly configured for each supported language, separate from org-wide translation features. Adding Spanish in Agentforce Studio and republishing makes the agent bilingual. Channel-level translation and duplicate agents do not change the agent's own language capability, so they would not resolve the scenario.

Exam trap

The trap here is assuming that enabling Einstein Translation at the org or channel level automatically makes an Agentforce agent multilingual, when agent language support is configured per agent.

Practice this question →
10
MCQmedium

An administrator is configuring an Agentforce Agent to answer questions about company policies. The agent must only use information from a specific set of Knowledge articles and must not generate answers based on general knowledge. Which feature should the administrator enable to restrict the agent's responses to the approved Knowledge articles?

A.Topics
B.Einstein Trust Layer
C.Prompt templates
D.Knowledge grounding
AnswerD

Knowledge grounding allows the agent to retrieve and use information from specified Knowledge articles. By configuring the agent to use only certain articles, the administrator can restrict its responses to that approved content. This feature ensures the agent does not generate answers from general knowledge, aligning with the requirement.

Why this answer

Knowledge grounding is the feature that connects the agent to specific Knowledge articles and ensures its responses are based on that content. By setting up knowledge grounding with a defined set of articles, the administrator can prevent the agent from using general knowledge, thus meeting the requirement.

Exam trap

The trap here is assuming that topics or Einstein Trust Layer automatically restrict knowledge sources, when in fact knowledge grounding is the specific mechanism for controlling the agent's information sources.

Practice this question →
11
MCQmedium

An administrator at a retail company has built an Agentforce Service Agent that handles returns and exchanges. During testing, the agent correctly answers product questions but fails to initiate a return because it cannot find the 'Initiate Return' action. The action was created as an Apex class implementing the @InvocableMethod annotation. Which step should the administrator take to make the action available to the agent?

A.Create an invocable action in Agent Builder and reference the Apex class.
B.Add the Apex class to the agent's permission set and enable 'Agentforce Action Access'.
C.Deploy the Apex class to the agent's connected app and add it as a remote site setting.
D.Add the Apex class to the agent's Einstein Trust Layer configuration.
AnswerA

In Agent Builder, actions are defined by selecting an invocable method (Flow, Apex, or Prompt Template). The administrator must create a new action, choose 'Apex' as the action type, and select the class that implements @InvocableMethod. This registers the action with the agent so it can be invoked during conversations. Without this step, the agent has no knowledge of the Apex method.

Why this answer

Agentforce agents can use actions built from Flows, Apex, or Prompt Templates, but these must be explicitly added as actions in Agent Builder. Simply creating an Apex class with @InvocableMethod does not make it available; the administrator must create an action that references the class. This ensures the agent can discover and invoke the method during conversations.

Exam trap

The trap here is assuming that any invocable Apex method is automatically available to an agent without explicitly adding it as an action in Agent Builder.

Practice this question →
12
MCQmedium

Universal Containers wants to deploy an Agentforce Service Agent that autonomously responds to customer inquiries about order status. The agent needs to securely retrieve real-time data from an external order management system via REST API. Which configuration step must the administrator complete first in Agentforce?

A.Create a custom Apex trigger to intercept inbound customer chat transcripts and route them to the external order management endpoint.
B.Configure an External Service using a Named Credential to define the schema and authentication parameters for the external order API.
C.Write an LWC component that calls the external order system directly and injects the returned JSON data into the chat window state.
D.Enable Einstein Copilot Studio and assign the Standard User profile permissions to execute external REST requests.
AnswerB

External Services leverage OpenAPI specifications and Named Credentials to automatically map external endpoints into invocable actions. This declarative bridge enables the Agentforce reasoning engine to understand API payloads, authenticate securely, and fetch live data during a customer conversation.

Why this answer

Before configuring topics and actions in Agentforce, an administrator must establish secure connectivity to the external system using Named Credentials and External Services. This foundational setup allows the generative AI model to safely interact with external APIs, ensuring that authentication tokens and endpoint configurations are properly managed within Salesforce security frameworks.

Exam trap

Candidates often jump straight to creating the Agent Action. They forget that the security layer (Named Credentials) must exist before the Agent can safely authenticate with an external API.

Practice this question →
13
MCQmedium

Refer to the exhibit. A customer asks for a $750 refund. The Agent refuses to process the request despite the customer being eligible. What is the most likely cause for this behavior?

A.The Knowledge Base is missing information about high-value refunds.
B.The RefundFlow has a hardcoded limit of $500.
C.The Agent is adhering to the configured guardrail limit.
D.The Agent is lacking the correct permission to update OrderHistory.
AnswerC

The agent is functioning as designed by enforcing the $500 guardrail configured in the Topic. Guardrails are critical for AI governance, ensuring agents do not execute actions that exceed company policy. The agent successfully recognized the input amount and applied the constraint, preventing a policy violation.

Why this answer

The Agent is operating within the strict constraints defined in its topic configuration. Because the guardrail explicitly limits the maximum refund amount to $500, the Agent correctly identifies that a $750 request violates this policy. This demonstrates how Agentforce uses guardrails to ensure AI agents operate within predefined corporate risk limits, preventing unauthorized or excessive financial actions even when a request appears valid.

Exam trap

Candidates often assume the Agent is malfunctioning or that the Flow is broken. They fail to realize that the Agent is behaving exactly as configured by the defined guardrail limits.

Practice this question →
14
MCQeasy

Which feature in Agentforce allows administrators to test how an agent will respond to a specific user input before deploying it to production?

A.Salesforce Optimizer.
B.Agent Builder Preview.
C.Einstein Discovery Predictions.
D.Deployment Manager.
AnswerB

The Agent Builder Preview window provides a direct interface for administrators to interact with their configured agent. It allows for testing and refining prompts, topics, and actions within the development environment, ensuring the agent's logic is sound before it is published for general user access.

Why this answer

The Agent Builder includes a built-in testing console that allows administrators to simulate conversations. This feature is essential for validating the agent's behavior, ensuring it understands intents correctly, and verifying that it triggers the appropriate actions. By testing in a sandboxed environment, administrators can refine instructions and configurations to optimize performance and prevent unexpected behavior before the agent interacts with live customers.

Exam trap

Candidates frequently confuse deployment validation tools with the interactive testing console, failing to recognize that the Agent Builder Preview is meant for real-time simulation.

Practice this question →
15
MCQmedium

An administrator at Get Cloudy Consulting has built an Agentforce Service Agent that answers questions about warranty coverage by retrieving records from a custom object. During testing, the agent replies with policy text that was updated two weeks ago, even though the source record was edited yesterday. The administrator confirms the record edit saved successfully and the agent's topic is active. Which configuration should the administrator check to resolve the stale response?

A.The Prompt Template version bound to the topic and whether it was activated after the record edit.
B.The Einstein Trust Layer Data Masking rule applied to the warranty custom object.
C.The Data Cloud or CRM search index used by the agent's retrieval, and whether it has been refreshed since the edit.
D.The agent user's profile and whether it has Read access to the warranty object's fields.
AnswerC

Agentforce topics that ground responses in records rely on a search or retrieval index, and that index is refreshed on its own schedule rather than instantly on every record save. If the index still holds the pre-edit version of the warranty record, the agent will confidently quote stale policy text even though the underlying record is current, which matches the reported symptom exactly.

Why this answer

The agent is returning content that exists but is no longer current, which points to a retrieval layer rather than permissions, masking, or prompt wording. Search and semantic indexes behind grounding are updated asynchronously, so a record edited yesterday may still be represented by its previous indexed version. Refreshing or rebuilding that index lets the topic retrieve the current warranty record and produce an accurate answer.

Exam trap

The trap here is assuming any record save instantly changes what a grounded agent retrieves, when the retrieval index refreshes on its own schedule.

Practice this question →
16
MCQhard

Refer to the exhibit. An agent attempts to trigger an order lookup, but it fails with the provided error. What is the most likely cause of this 403 error?

A.The flow is currently inactive.
B.The agent's user profile lacks execution permissions for the Flow.
C.The flow has an invalid input parameter.
D.The Agentforce platform is experiencing a service outage.
AnswerB

The 403 Forbidden error indicates an authorization failure. For an agent to invoke a flow, the user profile associated with the agent must have the 'Run Flow' permission. Without this explicit authorization, the platform blocks the agent from initiating the requested process, resulting in the observed error.

Why this answer

A 403 Forbidden error in Salesforce indicates that the executing user—in this case, the Agentforce service user—lacks the necessary permissions to access the specified resource. In the context of Flow-based actions, this means the agent's profile or permission set does not include 'Run Flow' access for the specific flow mentioned, preventing it from executing the requested automation safely and securely.

Exam trap

Candidates frequently mistake a 403 Forbidden error for an invalid record ID or bad API endpoint, ignoring the service user's missing execution permissions.

Practice this question →
17
MCQhard

An administrator at AW Computing is building an Agentforce Service Agent that must answer questions about open cases using a Flow action. The flow's input variables are declared as Text, but the agent must pass a Case Id value. Testing shows the action fails to return results. Which change should the administrator make to the action configuration?

A.Add a formula resource in the flow that concatenates the Case Id with an empty string before use.
B.Change the flow's input variable from Text to an Id-compatible type and ensure the agent action maps the Case Id to that input.
C.Convert the flow's input variable to a Record variable typed to the Case object and map the Case Id into it.
D.Grant the agent user the View All Data permission so it can read any case regardless of ownership.
AnswerB

The failure occurs because the agent passes a Case Id that the action must accept in a compatible input. Aligning the flow's input variable with an identifier-capable type and explicitly mapping the Case Id in the agent action configuration ensures the value is delivered correctly, allowing the flow to query and return open-case results as intended.

Why this answer

The action fails because the value the agent supplies and the input the flow expects do not align. Correcting the flow's input variable to an identifier-compatible type and explicitly mapping the Case Id in the agent action configuration lets the identifier flow through, after which the flow can query and return open-case details. Record variables, formula concatenation, and broad permissions all target the wrong layer of the problem.

Exam trap

The trap here is expanding permissions or reshaping values when the real defect is a mismatched input type and missing mapping between the action and the flow.

Practice this question →
18
MCQmedium

What is the purpose of the 'Human Agent Handoff' feature in Agentforce?

A.To force the user to pay for premium support services.
B.To ensure complex or unresolved issues are escalated to a human.
C.To automatically terminate the chat session if it takes too long.
D.To allow the agent to learn from the human agent's responses in real-time.
AnswerB

The human agent handoff is essential for managing the limitations of AI. By identifying when an agent is struggling or when the user specifically asks for help, the feature transitions the chat to a live support agent, ensuring that the customer gets the assistance they need without disruption.

Why this answer

The human agent handoff is a critical safety mechanism that ensures that if an agent cannot resolve a query or if a customer requests a person, the interaction is seamlessly transitioned to a human support agent. This maintains continuity of service, preserves the context of the conversation, and prevents customer frustration by ensuring that human intervention is available when AI capabilities are reached.

Exam trap

Candidates often assume handoff is for technical errors only. They miss that handoff is a standard business process for customer satisfaction when the AI reaches its functional or emotional limitations.

Practice this question →
19
MCQeasy

Universal Containers is preparing to activate an Agentforce Service Agent for a customer-facing help site. Before activation, the administrator must ensure the agent only surfaces knowledge articles that the requesting customer is permitted to see. Which configuration should the administrator verify?

A.The running user's sharing settings and object permissions for the knowledge articles the agent retrieves.
B.The agent's Einstein Trust Layer data masking rules for the knowledge object.
C.The Einstein Trust Layer audit trail retention period for the agent.
D.The agent's response length limit in the Agent Builder settings.
AnswerA

When an agent retrieves knowledge on behalf of a customer, the records returned honor the access of the context in which the action runs. Verifying sharing settings and object and field permissions ensures the agent cannot surface articles the customer is not entitled to see. This is the correct control for record-level access.

Why this answer

Knowledge retrieval by an agent respects the permissions and sharing of the context in which the retrieval runs. To prevent customers from seeing articles they are not entitled to, the administrator must verify sharing settings and object and field permissions for the knowledge records the agent can return. Masking, response length, and audit retention do not govern record-level access.

Exam trap

The trap here is confusing Einstein Trust Layer data masking, which protects sensitive values in prompts, with record-level access control that determines which knowledge articles the agent can retrieve.

Practice this question →
20
MCQmedium

When designing instructions for an Agentforce Agent, what is the best practice for ensuring the agent remains on-topic and maintains a professional tone?

A.Use a complex, multi-page instruction document to cover every possible edge case.
B.Provide clear, concise instructions defining the agent's persona and allowed scope.
C.Allow the agent to dynamically learn from user input to improve its tone over time.
D.Hard-code all possible responses into the Agent Builder to ensure perfect tone.
AnswerB

Clear, concise instructions define the guardrails for the agent. By explicitly stating the persona and the limitations of its knowledge, you create a controlled environment where the model is less likely to deviate from its intended purpose, resulting in more professional and reliable customer support outcomes.

Why this answer

Effective system instructions (often called 'System Prompts') should be clear, concise, and prescriptive. By explicitly defining the agent's persona, boundaries, and required behavior in the Agent Builder, you minimize hallucinations and ensure consistent service. This is a critical administrative task because well-crafted instructions directly correlate to the quality of the AI's output and the overall customer experience during automated support interactions.

Exam trap

Candidates often assume that leaving instructions open-ended allows the AI to be more helpful, failing to realize that precise guidelines and personas prevent hallucinations.

Practice this question →
21
Multi-Selecthard

A large financial services company is implementing an Agentforce Service Agent to handle customer inquiries about account balances and recent transactions. The administrator must ensure the agent can securely access customer data and provide accurate responses. Which two configurations are required to enable the agent to retrieve data from Salesforce objects? (Choose two.)

Select 2 answers
A.Add the objects as data sources in the agent's Topic configuration and define the relevant fields.
B.Configure a Named Credential for each Salesforce object the agent needs to access.
C.Enable the Einstein Trust Layer and set data masking rules for all fields.
D.Assign the agent's running user a profile with the 'Modify All Data' permission.
E.Create a permission set that grants the agent's running user read access to the necessary objects and fields.
AnswersA, E

In Agent Builder, each Topic can include data sources that specify which Salesforce objects and fields the agent can use to answer questions. By adding the necessary objects (e.g., Account, Contact) and their fields to the Topic, the agent can retrieve and reason over that data. This configuration is essential for the agent to know what data is available and how to query it.

Why this answer

To enable an Agentforce agent to retrieve data from Salesforce objects, the administrator must ensure the agent's running user has read access via a permission set, and the objects and fields must be added as data sources in the agent's Topic configuration. These two steps work together: the permission set grants the underlying access, while the Topic configuration tells the agent what data it can use.

Exam trap

The trap here is confusing external authentication mechanisms like Named Credentials with internal data access, or assuming that broad permissions like 'Modify All Data' are needed instead of least-privilege access.

Practice this question →
22
MCQmedium

An administrator needs to ensure that an Agentforce agent only accesses specific knowledge articles relevant to a customer's issue. Which configuration step is required to enforce these boundaries?

A.Enable Einstein GPT for Knowledge in the Setup menu.
B.Assign the 'Agent Knowledge Reader' permission set to the agent profile.
C.Add Knowledge as a data source in Agent Builder and define retrieval filters.
D.Update the sharing rules for all knowledge articles to public read-only.
AnswerC

Defining Knowledge as a data source within the Agentforce builder allows for the application of specific filters. These filters ensure the agent only queries articles matching defined criteria, which is the standard mechanism for restricting the retrieval-augmented generation scope and ensuring the AI provides accurate, contextually relevant information.

Why this answer

To control agent access, administrators must configure Knowledge as a data source within the Agentforce builder. By applying filters based on category or publication status, you ensure the AI retrieves only verified, relevant documentation. This is critical for maintaining data accuracy and preventing the agent from hallucinating based on irrelevant or outdated internal documents while interacting with external customers.

Exam trap

Candidates often assume that standard Profile or Permission Set access automatically limits AI retrieval. However, Agentforce requires explicit configuration within the Agent Builder to define data sources and retrieval filters.

Practice this question →
23
MCQmedium

A service manager wants to implement Agentforce to handle routine customer password resets autonomously. Which component must be configured to allow the Agent to perform this specific action within the Salesforce environment?

A.Enable Einstein GPT in the Setup menu.
B.Configure an Agent Action linked to a Flow.
C.Add the user to the Agentforce Permission Set.
D.Update the Agent's Topic definitions.
AnswerB

Agent Actions serve as the bridge between the Agent's reasoning engine and Salesforce automation tools like Flow. By associating a specific intent with a Flow, the Agent gains the capability to execute the password reset logic, ensuring that all security protocols defined in the Flow are consistently applied.

Why this answer

To enable an autonomous agent to perform actions like password resets, you must configure Agent Actions. These actions bridge the agent's natural language understanding with internal Salesforce functions, such as Flow or Apex. By mapping user intent to a specific Flow, the Agent can securely trigger the reset process without human intervention, ensuring that business logic is strictly followed during execution.

Exam trap

Candidates often suggest using a Prompt Template or a simple Topic instruction. However, these cannot execute business logic; only an Agent Action linked to a Flow can perform system changes.

Practice this question →
24
MCQeasy

What is the primary function of the 'Topics' feature in Agentforce?

A.To define the color scheme of the agent's chat interface.
B.To classify and categorize the agent's capabilities and instructions.
C.To manage the list of users who can interact with the agent.
D.To store the chat history of the agent for auditing.
AnswerB

Topics are the primary mechanism for organizing what an agent is allowed to do. They encapsulate specific sets of instructions and actions, which allows the AI model to quickly identify which capabilities are relevant based on the user's input, thus improving the efficiency and accuracy of the task execution.

Why this answer

Topics act as the logical framework for organizing an agent's capabilities. By grouping related instructions, actions, and knowledge resources under a specific topic, you allow the agent to understand context better and switch between specialized tasks efficiently. This modular approach makes agents more maintainable and easier to scale as the business requirements grow and the agent's functionality expands over time.

Exam trap

Candidates often confuse 'Topics' with 'Instructions'. While they are related, Topics are the structural containers used to organize the agent's capabilities, not the specific prompts themselves.

Practice this question →
25
MCQmedium

An administrator at a retail company is building an Agentforce Service Agent that must look up a customer's loyalty points balance from an external loyalty platform in real time. The external system exposes a REST API that requires a named credential for authentication. The administrator wants to make this data available to the agent as an action. Which approach should the administrator use?

A.Configure a Data Cloud data stream to ingest loyalty points nightly, then reference the Data Cloud object in the agent's topic.
B.Create an External Service registration from the REST API schema, then use the auto-generated invocable actions in the agent's topic.
C.Build a Screen Flow that calls the REST API with an HTTP Callout action, then expose the flow as an Agentforce action.
D.Create a custom Apex class with an @InvocableMethod that performs the callout, and register it as an agent action.
AnswerB

External Services reads an OpenAPI (or similar) schema and generates invocable Apex actions that can be added directly to an Agentforce topic. It natively supports named credentials for authentication, so the agent can call the external loyalty REST endpoint in real time without custom Apex wiring. This is the declarative, admin-manageable path for a registered external REST API.

Why this answer

External Services is the declarative way to consume a registered REST API as invocable actions. It generates the action definitions from the API schema and honors named credentials for authentication, so the agent can query the loyalty platform live. Building custom Apex or a Screen Flow adds code and UI overhead that the External Service registration avoids, and batch ingestion cannot satisfy a real-time balance check.

Exam trap

The trap here is assuming that any callout requires custom Apex, when External Services can auto-generate invocable actions from a REST schema.

Practice this question →
26
MCQeasy

What is the primary role of a 'Topic' in the Agentforce Agent Builder?

A.It serves as the authentication layer for external API calls.
B.It defines a specific capability or intent that the agent can perform.
C.It stores all customer interaction logs for auditing.
D.It determines the visual layout of the chat interface.
AnswerB

In the Agent Builder, a Topic acts as a handler for specific user intents. By defining a topic, you tell the agent what it is capable of doing—such as checking order status or scheduling an appointment—and provide the specific instructions needed to fulfill that capability.

Why this answer

A Topic represents a specific intent or category of customer inquiry that the agent is trained to understand and act upon. By grouping related phrases, instructions, and actions under a Topic, you organize the agent's capabilities logically. This allows the AI model to quickly identify the user's intent and select the appropriate process to follow, which is fundamental for providing relevant and structured support.

Exam trap

Candidates often confuse 'Topics' with 'Instructions' or 'Actions.' They mistakenly believe a Topic is an automated script or a specific API integration rather than a high-level intent classification for the agent.

Practice this question →
27
MCQeasy

A service team at Ursa Major Solar wants its Agentforce Service Agent to hand a conversation to a human representative whenever a customer expresses frustration, without the customer repeating any details. The administrator must configure this escalation so the human sees the full transcript and context. Which capability should the administrator configure?

A.A prompt template instruction telling the agent to apologize and continue assisting the customer.
B.A Case escalation rule that reassigns the case owner after the agent closes the conversation.
C.Omni-Channel escalation with the agent as a routing source, passing the conversation to a queue.
D.An Einstein Trust Layer audit trail that logs the frustrated sentiment for later review.
AnswerC

Agentforce Service Agents integrate with Omni-Channel so an agent can escalate an active messaging session to a human queue while preserving the transcript and context. Routing the session as a work item means the representative receives the full conversation history in the messaging component, satisfying the requirement that the customer need not repeat details, which is exactly the escalation behavior described.

Why this answer

Escalating a live agent conversation to a person requires a routing mechanism that carries session context, which is provided by Omni-Channel treating the agent session as a routable work item. Case rules, audit logging, and prompt wording all operate on records or model behavior rather than on the active messaging session, so none of them can deliver the transcript to a human without the customer repeating information.

Exam trap

The trap here is treating case escalation rules or polite prompt wording as equivalent to transferring a live conversation with its transcript.

Practice this question →
28
MCQmedium

How does an Agentforce Agent maintain context during a multi-turn conversation?

A.By storing all user history in a long-term database record.
B.By passing the conversation history as context to the LLM.
C.By relying on browser cookies to save the user's session state.
D.By periodically refreshing the agent's training model.
AnswerB

To maintain context in a multi-turn conversation, the system sends the previous turns of the current session to the LLM. This provides the AI with the necessary background to interpret follow-up questions accurately, ensuring that the responses remain coherent and contextually relevant throughout the user's interaction.

Why this answer

Agentforce leverages Large Language Models that maintain a 'conversation window' or 'state' by passing the history of the current session back to the model with every request. This allows the AI to remember what was previously said and provide relevant, consistent follow-up answers. Managing this state effectively is crucial for a natural, human-like interaction, allowing the agent to resolve complex queries that require multiple exchanges of information.

Exam trap

Candidates often guess that the AI stores data in a persistent database or a custom object. They fail to realize the LLM requires the history to be passed as context.

Practice this question →
29
MCQhard

Northern Trail Outfitters wants its Agentforce Service Agent to summarize a long customer chat and store the summary on the related case for human agents. The administrator needs the summary to be generated by the model and then written to a field. Which approach should the administrator use to accomplish both steps reliably?

A.Add the summary as a static instruction in the Topic so the agent repeats it at the end of every session.
B.Use a Prompt Template to generate the summary text, then pass its output into an Agent Action that updates the case field.
C.Enable Einstein Activity Capture to automatically summarize the chat and update the case.
D.Configure a Data Cloud calculated insight to write the summary directly to the case field.
AnswerB

A Prompt Template produces the model-generated summary, and an Agent Action can take that output as input and write it to the case field. Combining generation with a deterministic write action accomplishes both steps reliably. This is the supported pattern for generating content and persisting it to a record.

Why this answer

Generating a summary and persisting it requires two capabilities: model-based text generation and a deterministic record update. A Prompt Template produces the summary, and an Agent Action can consume that output and write it to the case field. Calculated insights, Einstein Activity Capture, and static instructions cannot generate a dynamic summary and store it on the record.

Exam trap

The trap here is expecting a single feature to both generate model text and write it to a record, when generation and record updates are handled by separate Agentforce capabilities.

Practice this question →

Ready to test yourself?

Try a timed practice session using only Agentforce Ai questions.

Start 20-question session