SF-Admin Agentforce AI — 15 Questions

Question 1

An administrator needs to ensure that an Agentforce Agent only accesses specific knowledge articles regarding internal HR policies. How should the administrator configure the agent's access?

Accepted Answer

Use the Agent Builder to restrict the agent's access to specific Knowledge Data Categories.. To control agent data access, administrators must configure Agent Actions and Topic triggers. By applying Knowledge Data Categories to the agent's permission set or using specific filters within the Agent Builder, you ensure the AI model only retrieves permitted documentation. This maintains data sovereignty and security by preventing unauthorized access to sensitive company information, which is critical for compliance and trust in AI-driven interactions.

Answer

Apply a Sharing Rule to the Knowledge object based on the Agent's User profile.

Answer

Configure a Permission Set specifically for the AI User with read-only access to all articles.

Answer

Update the Agent's system prompt to ignore any articles outside of HR policy.

Question 2

An administrator is setting up an Agentforce Agent to assist customers with order status queries. Which TWO components are required to enable the agent to retrieve real-time order data from a custom object?

Accepted Answer

An Apex class annotated with @InvocableMethod.. To enable an agent to perform tasks involving external or custom data, you must define the capability via Apex or Flow as an Agent Action. Once the logic is defined, it must be mapped to a Topic in the Agent Builder. This two-step process ensures the agent has both the functional instructions and the authorized access to execute the business logic securely.

Answer

A custom Lightning Web Component to display the order status.

Answer

A dedicated Permission Set license for the Agent user.

Answer

An API integration user with OAuth tokens.

Question 3

Refer to the exhibit. An administrator is troubleshooting an Agentforce agent that fails to trigger the 'lookup_order' action. Based on the provided policy JSON, what is the most likely reason for the failure?

Accepted Answer

The confidence score of the user query is below 0.85.. The exhibit defines a confidence threshold of 0.85. If the agent's intent classification confidence score falls below this value, the action will not trigger, defaulting instead to the defined fallback. This mechanism is vital for maintaining high accuracy in AI interactions, as it prevents the agent from executing actions when it lacks sufficient certainty about the user's actual intent.

Answer

The agent is missing the necessary read permissions for the order object.

Answer

The 'data_scope' is set to 'current_user' which restricts access to shared orders.

Answer

The fallback array is empty and preventing execution.

Question 4

Which THREE factors should an administrator consider when choosing between a Flow-based action and an Apex-based action for an Agentforce agent?

Accepted Answer

Complexity of the underlying business logic and data calculations.. Flows are excellent for low-code automation and maintainability, while Apex offers high performance and complex data processing capabilities. Choosing the right tool depends on the complexity of the task, the need for external API integrations, and the administrative expertise required to maintain the code over time. Understanding these tradeoffs ensures that the agent's backend logic remains scalable, secure, and easy to debug.

Answer

The requirement to display a UI component to the end user.

Answer

The number of users currently active on the Agentforce platform.

Question 5

Which TWO of the following are considered best practices for managing data security when using Agentforce?

Accepted Answer

Implementing standard Salesforce record-level security (e.g., OWDs, Role Hierarchy).. Data security is paramount in AI implementations. Restricting data visibility through object-level security and sharing rules ensures that the AI only interacts with data the user is authorized to see. Furthermore, using sensitive data masking and auditing logs ensures compliance and visibility into how AI is accessing information, protecting sensitive PII (Personally Identifiable Information) from being inappropriately exposed in agent responses.

Answer

Granting the Agent user 'Modify All' access to ensure it can see all necessary data.

Answer

Including all company data in the Agent's knowledge base for faster retrieval.

Answer

Hard-coding API credentials in the Apex controller for data retrieval.

Question 6

Refer to the exhibit. An administrator has configured an Agentforce action using an invocable flow. After deployment, audit logs reveal that the agent is returning order records belonging to restricted accounts that the chatting user lacks permission to view. What is the root cause of this security vulnerability?

Accepted Answer

The securityContext is configured to run without sharing, bypassing user-level object and record permissions.. The security context of an Agentforce action dictates whether it respects user-level permissions. Setting the context to run without sharing overrides record-level security, causing the flow to bypass sharing rules and return unauthorized data to the conversational context, violating principle of least privilege.

Answer

The maxRetries parameter is set too high, causing the action to cache unauthorized query results from previous sessions.

Answer

GroundingEnabled is set to true, which forces Data Cloud to merge disparate customer records into a single profile.

Answer

The actionName uses an asynchronous invocable method that executes outside the active user session context.

Question 7

An administrator is preparing to deploy an Agentforce agent. Which TWO of the following are necessary prerequisites before the agent can be enabled for production use?

Accepted Answer

Test the agent thoroughly within a Sandbox environment.. Deploying Agentforce requires verifying that the agent has the correct permissions and that the underlying data sources are optimized. Testing in a sandbox is essential to validate that the LLM response accurately reflects the business logic. These steps ensure that the agent remains compliant with corporate policies and provides high-quality interactions before interacting with real customer data in a production environment.

Answer

Verify the agent has a dedicated User License assigned.

Answer

Install the Agentforce AppExchange package.

Answer

Manually create an Apex class for every user interaction.

Question 8

An agent is providing incorrect answers regarding product pricing. Which THREE of the following actions should the administrator perform to improve the agent's accuracy?

Accepted Answer

Add a specific Knowledge base that contains current pricing.. To resolve pricing accuracy, the administrator must ensure the grounding data is correct, the prompt is clear, and the retrieval logic is focused. By using high-quality knowledge articles, refining the instructions, and testing with diverse user queries, the AI can better understand the specific business context of product pricing. These steps ensure the agent relies on authoritative internal data rather than broad, potentially outdated training data.

Answer

Increase the number of conversational turns in the history.

Answer

Enable the agent to search the public internet for pricing.

Question 9

An administrator at a healthcare company has deployed an Agentforce Service Agent to handle patient appointment scheduling. The agent is configured with a Topic called 'Appointment Scheduling' that includes a Flow action to create appointments. During testing, the agent successfully creates appointments but sometimes provides incorrect information about available time slots because it is not querying the real-time schedule. Which configuration should the administrator review to ensure the agent retrieves current availability?

Accepted Answer

The Topic's data source configuration to include the Appointment object and relevant fields.. The agent needs to query the Appointment object to provide real-time availability. This is achieved by adding the Appointment object as a data source in the Topic configuration. The administrator must ensure the object and fields like Start Time and Status are included, and that the running user has read access. Without this, the agent cannot retrieve current slots and may hallucinate or use stale data.

Answer

The Flow action's input and output parameters to include a query for available slots.

Answer

The Einstein Trust Layer data masking rules for the Appointment object.

Answer

The agent's language model temperature setting to reduce creativity.

Question 10

An administrator is configuring an Agentforce Service Agent for a bank. Compliance requires that the agent never reveal a customer's full account number in chat, even when the agent retrieves it from a Salesforce record to verify identity. The administrator wants to enforce this at the agent level rather than relying on each action's code. Which configuration should the administrator use?

Accepted Answer

Configure a Data Masking rule on the Account object so the account number field is masked in Agentforce responses.. Data Masking rules in Agentforce redact sensitive field values before they reach the model or the customer, providing a deterministic guardrail that does not depend on prompt wording. Instructions are advisory, field-level security blocks retrieval entirely and breaks identity verification, and validation rules only govern record saves. Masking at the data layer is the compliant, agent-level control.

Answer

Set field-level security on the account number field so the agent's running user cannot read it.

Answer

Create a validation rule on the Account object that prevents the account number from being displayed in chat.

Answer

Add an instruction to the agent's system prompt telling it to mask account numbers before responding.

Question 11

An administrator at a financial services company is configuring an Agentforce Agent to handle customer inquiries about loan balances. The agent must not reveal any personally identifiable information (PII) such as Social Security numbers or account numbers in its responses. The administrator needs to enforce this restriction at the agent level. Which feature should the administrator use to prevent the agent from including PII in its replies?

Accepted Answer

Configure the agent's instructions to explicitly prohibit sharing PII.. Agent instructions are the primary mechanism to define an agent's behavior and boundaries. By explicitly prohibiting the sharing of PII, the administrator can guide the agent to avoid including such information in its responses. This is a direct, agent-level configuration that aligns with Agentforce's design for customizable, natural language-based control.

Answer

Implement a Flow that filters PII from the agent's responses before they are sent.

Answer

Set the agent's user profile to restrict access to PII fields.

Answer

Enable Data Masking on the agent's user profile.

Question 12

A Salesforce administrator is building an Agentforce Service Agent to assist customers with password reset requests. The agent needs to verify the customer's identity by asking for their email and a one-time code sent to their phone. The administrator wants to ensure that the agent can only access the verification code from a secure external system and cannot store it in Salesforce. Which action type should the administrator use to meet this requirement?

Accepted Answer

Apex action. Apex actions are the most appropriate choice because they enable custom server-side logic, including external callouts, while allowing the administrator to handle sensitive data in memory without persisting it in Salesforce. This satisfies the requirement to verify the code externally and avoid storing it, which is critical for security and compliance.

Answer

Flow action

Answer

Prompt Template action

Answer

Standard action

Question 13

An administrator is creating an Agentforce Agent to help customers track their shipments. The agent needs to greet customers, ask for their order number, and then provide the current shipping status. Which component of Agentforce is used to define the conversational flow and the agent's responses?

Accepted Answer

Dialog flows. Dialog flows are specifically designed to orchestrate the conversation between the agent and the user. They enable the administrator to define the sequence of prompts, collect user input, and determine the agent's responses based on that input. This makes them the ideal component for structuring the shipment tracking interaction described.

Answer

Instructions

Answer

Topics

Answer

Actions

Question 14

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

Accepted Answer

Agent Builder Preview.. 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.

Answer

Salesforce Optimizer.

Answer

Einstein Discovery Predictions.

Answer

Deployment Manager.

Question 15

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

Accepted Answer

Provide clear, concise instructions defining the agent's persona and allowed scope.. 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.

Answer

Use a complex, multi-page instruction document to cover every possible edge case.

Answer

Allow the agent to dynamically learn from user input to improve its tone over time.

Answer

Hard-code all possible responses into the Agent Builder to ensure perfect tone.