SALESFORCE AGENTFORCE

salesforce agentforce Practice Questions & Answers (Set 15) | CodeWme

📝 Instructions: Read the hint to know if you need to select one or multiple options.

#1 An Agentforce Specialist at Universal Containers (UC) is building with no-code tools only. They have any small come the are only touched periodically by a speciatized sales team, and UC wants to meximize the sales operations tears time. OC to help prep the sales team for calls by: Summarizing past purchases Displaying products the contact has shown interest in (with data captured via Dats Cloud) Providing a recap of past email and phone conversations that have transcripts Which approach should the Agentforce Specialist recommend to achieve this goal? Select 1

A. Use a prompt template grounded on CRM and Data Cloud data using standard foundation models.
B. Deploy UC's own custom foundational model on this data first.
C. Fine-tune the standard foundational model due to the complexity of the data.

✅ Answer: Use a prompt template grounded on CRM and Data Cloud data using standard foundation models.

Explanation:
Prompt Builder allows no-code grounding with CRM data (Merge Fields) and Data Cloud data (DMOs/Flows). Summarization is a standard capability of foundation models, requiring no fine-tuning.

#2 Universal Containers wants to implement a customer verification process where sensitive account information can only the accessed after the customer passes identity verification. The agent must enforce this security rule deterministically without alflowing the large language model (LLM) to bypass the verification requirement. What should an Agentforce Specialist recommend as the best solution? Select 1

A. Include detailed verification instructions in the agent's topic instructions explaining when customers showkt be verified and rely on the LLM to follow these guidelines consistently across all interactions.
B. Use context variables to store verification status in the messaging session and configure the agent to check these variables through natural language prompts during each sensitive action.
C. Create a custom variable IsCustomerVerified set by a verification action, then apply a conditional filter using the expression IsCustomerVerified equals true to all sensitive data actions, ensuring deterministic access control that the LLM can't alter.

✅ Answer: Create a custom variable IsCustomerVerified set by a verification action, then apply a conditional filter using the expression IsCustomerVerified equals true to all sensitive data actions, ensuring deterministic access control that the LLM can't alter.

Explanation:
Conditional filters on actions provide deterministic (Level 5) control. The reasoning engine enforces the rule programmatically, preventing the LLM from accessing the action unless the variable condition is met.

#3 An Agentforce Specialist is creating a custom agent action. The topic is selected correctly, but the ection is not Which setting should the Agentforce Specialist test and iterate on to ensure the action performs as expected? Select 1

A. Action Instructions
B. Action Scope
C. Classification Description

✅ Answer: Action Instructions

Explanation:
Action Instructions are the primary mechanism the reasoning engine uses to select a specific action within an active topic. If the topic is right but the action isn't picking up, the instructions likely need refinement.

#4 An Agentforce Specialist wants to troubleshoot an agent that is hallucinating weblinks, The agent has an template, which is using a knowledge retriever, to generate the output test that the agent will use. Which procesa is appropriate to find the root cause of the hallucination behavior? Select 1

A. Examine the prompt instructions and contents of the chunks shown in the resolved prompt output
B. Examine the topic name and classification description for hallucination guardrails.
C. Examine the topic instructions and ensure the word "ALWAYS" is used in the hallucination guardrails.

✅ Answer: Examine the prompt instructions and contents of the chunks shown in the resolved prompt output

Explanation:
The Resolved Prompt in the Plan Tracer/Prompt Builder shows the actual 'chunks' retrieved from Knowledge. Checking this allows you to see if the bad link is in the source data or if the LLM invented it.

#5 An administrator at Universal Containers is setting up e now Sales Development Representative (SDR) Agent. The egent's purp is to nurture cold leads before connecting them to the assigned sales rep To ensure the agent has all the necessary access to the leads in the North America sales region, what should the atministrenor det Select 1

A. Create a criteria-based sharing rule to grant access to targeted lead records to SDR Agent User.
B. Grant View All record permission of the Lead object to the 'Einstein Agent User profile.
C. Assign the user in the highest-level role within the North America role hierarchy as the SDR Agent User.

✅ Answer: Create a criteria-based sharing rule to grant access to targeted lead records to SDR Agent User.

Explanation:
Criteria-based sharing rules allow for precise, secure access control, granting the agent user access only to the specific subset of records (North America Leads) required for its job.

#6 Coral Cloud Resorts (CCR) uses Agentforce to assist cuntorners with booking end service beyes. CCR want to mee process so that High severity requests must be escalated to a human service rep. Lower severity requests should result in creating a support case for the guest. The requirement is to achieve the highest reliability and determinism in the response from the agent. Which approach should an Agentforce Specialist recommend? Select 1

A. Create a custom variable severityl.evel populated by a Triage action. Add filters so the "Escalate to human service rep" action only runs when severityLevel = 'High', and the "Create Support Case" action runs only when severityLevel != 'High'.
B. Use absolute keywords like "Always" and "Never" in Topic Instructions to enforce logic, such as "Always escalate when severity is high" and "Never create a support case when severity is high"

✅ Answer: Create a custom variable severityl.evel populated by a Triage action. Add filters so the "Escalate to human service rep" action only runs when severityLevel = 'High', and the "Create Support Case" action runs only when severityLevel != 'High'.

Explanation:
Using custom variables and conditional filters (Logic/State) ensures deterministic behavior. Relying solely on instructions (Option B) is probabilistic and less reliable for business-critical logic.

#7 An agent has been developed with multiple topics and agent actions that use flows and Apen Which option is available for deploying these to production? Select 1

A. Use only change sets because the Salesforce CLI does not currently support the deployment of agent-related metadata.
B. Deploy flows, Apex, and all agent-related iterns using either change sets or the Salesforce CLI/Metadata APL
C. Deploy the flows and Apex using normal deployment tools and manually create the agent-related items in production.

✅ Answer: Deploy flows, Apex, and all agent-related iterns using either change sets or the Salesforce CLI/Metadata APL

Explanation:
Agentforce metadata (GenAiPlugin, GenAiFunction, Bot, etc.) is fully supported by the Metadata API, allowing deployment via both Change Sets and the Salesforce CLI.

#8 Universal Containers wants to use an Al egent to answer questions about warranties. Warranty information has been uploaded as unstructured data in Data Cloud, When answering user questions, the results must be fitherstile by productive ranked by recent updates. Which approach should the Agentforce Specialist implement? Select 1

A. Apply semantic embeddings with default metadata filters to achieve the desired result.
B. Use the default retriever which automatically accounts for recency ranking.
C. Build a custom retriever in Einstein Studio with product line filters and recency ranking.

✅ Answer: Build a custom retriever in Einstein Studio with product line filters and recency ranking.

Explanation:
Default retrievers are 'black boxes'. To apply specific metadata filters (like Product Line) or custom ranking logic (like Recency), a Custom Retriever must be built in Einstein Studio.

#9 Support agents at Universal Containers are using Agentforce to find troubleshooting information. They've reported that the agent frequently provides knowledge articles that are outdated, even when newer versions of the articles are available. The administrator has confirmed that all articles are correctly chunked and indexed. Which configuration change in the Data Cloud hybrid search index best addresses this problem? Select 1

A. Switch the chunking strategy from section-aware to fixed-size.
B. Add a ranking factor for recency based on the LastModifiedDate field.
C. Disable the keyword index to rely solely on the vector index.

✅ Answer: Add a ranking factor for recency based on the LastModifiedDate field.

Explanation:
Adding a ranking factor (boost) based on 'LastModifiedDate' in the Data Cloud Search Index configuration ensures that newer articles are ranked higher than older, semantically similar ones.
Next Set (Set 16) →