Salesforce Agentforce-Specialist - Salesforce Certified Agentforce Specialist (AI-201) Spring 26 Update

The correct answer is B because a Knowledge-based Agentforce Data Library is the correct managed source for approved service and compliance content. Salesforce Knowledge already handles the article lifecycle, including publication state and article governance, while Agentforce Data Library makes that content searchable for AI grounding. Filtering to published articles prevents draft, archived, or outdated content from being used in customer-facing responses. Option A adds unnecessary Flow-based retrieval and bypasses the native semantic retrieval pattern. Option C may work for custom indexed content, but it is less efficient when the authoritative source is already Salesforce Knowledge. The clean architecture is to connect the data library to Knowledge, restrict the indexed scope, and allow the agent to retrieve semantically relevant approved articles.

The correct answer is C because Agent-to-Agent, or A2A, is the purpose-built protocol for communication and collaboration between AI agents across different vendors and platforms. In this scenario, Cloud Kicks has a third-party research agent and an Agentforce customer service agent. The requirement is direct cross-vendor agent communication, not just access to a tool or external endpoint. MCP is mainly used to connect agents with tools, resources, and external context. A generic API can integrate systems, but it is not the agent-interoperability protocol being tested here. Salesforce describes A2A as an open standard that allows AI agents from different platforms to securely communicate, collaborate, share context, and coordinate work.

To enable Sales Email prompt templates:

Permission Set: Assign the Prompt Template User permission set to the sales team to grant access to use pre-built templates.

Feature Activation: Enable Sales Emails in Salesforce Setup to activate the integration between prompt templates and email workflows.

Option B (Manager permission set): Required for creating/modifying templates, not for usage.

Option C (Data Cloud Admin): Unrelated to prompt template access.

1.Context of the Question

oUniversal Container (UC) has used prompt templates to update summary fields on record pages.

oNow, the admin wants to incorporate similar generative AI functionality within a Flow for automation purposes.

2.How to Call a Prompt Template Within a Flow

oFlow Action: Salesforce provides a standard way to invoke generative AI templates or prompts within a Flow step. From the Flow Builder, you can add an “Action” that references the prompt template you created in Prompt Builder.

oOther Options:

Invocable Apex: Possible fallback if there’s no out-of-the-box Flow Action available. However, Salesforce is releasing native Flow integration for AI prompts, making custom Apex less necessary.

Einstein for Flow: A broad label for Salesforce’s generative AI features within Flow. Under the hood, you typically use a “Flow Action” that points to your prompt.

3.Conclusion

oThe easiest out-of-the-box solution is to use a Flow Action referencing the prompt template. Hence, Option B is correct.

Salesforce Agentforce Specialist References & Documents

•Salesforce Trailhead: Use Prompt Templates in Flow

Demonstrates how to add an Action in Flow that calls a prompt template.

•Salesforce Documentation: Einstein GPT for Flow

To meet Universal Containers ' goal of improving efficiency and reducing agent handling time with AI-generated responses, the best approach is to enable Service Replies, Service AI Grounding, and Grounding with Knowledge.

Service Replies generates responses automatically.

Service AI Grounding ensures that the AI is using relevant case data.

Grounding with Knowledge ensures that responses are backed by Salesforce Knowledge articles, allowing agents to identify whether a response is coming from the LLM or Salesforce Knowledge.

Option C does not include Service Replies, which is necessary for generating AI responses.

Option A lacks the Grounding with Knowledge, which is essential for identifying response sources.

For more details, refer to Salesforce Service AI documentation on grounding and service replies.

The AgentForce Action Design Guide emphasizes that concise and clear action instructions are essential for predictable and reliable agent behavior. The documentation states: “Action instructions should be concise, clearly define the purpose of the action, and specify its expected inputs and outputs. Each action must be validated through testing in AgentForce Builder to confirm the LLM interprets and executes it as intended.”

This aligns with Option A, which focuses on clarity and testing.

Option B incorrectly focuses on label-intent matching, which is a naming best practice but not sufficient for accuracy.

Option C promotes overly detailed instructions, which can lead to LLM confusion and inconsistent results.

Therefore, Option A best aligns with Salesforce’s official best practices for creating and validating custom actions.

References (AgentForce Documents / Study Guide):

AgentForce Action Design and Testing Guide

AgentForce Builder Documentation: “Validating Custom Actions”

AgentForce Study Guide: “Creating Clear and Testable Action Instructions”

Hybrid reasoning in Agent Script means combining flexible LLM-driven conversation with deterministic procedural control in the same agent logic. This is exactly why option C is correct. Salesforce describes Agent Script as supporting business-critical logic that executes reliably while allowing conversational elements to remain flexible. The - > syntax marks executable procedural logic, so steps such as identity verification can be enforced before account access. Option A is wrong because hybrid reasoning is not about using multiple LLMs at the same time. Option B is also wrong because hybrid reasoning is a language and execution model, not a separation of capabilities between Canvas View and Script View. The core exam point is deterministic enterprise workflow control plus LLM interpretation.

For Einstein Sales Emails to generate personalized follow-up emails, it is crucial to ground the email content with the most up-to-date and accurate information. Grounding refers to connecting the AI model with real-time data. The most appropriate technique in this case is Ground with Record Merge Fields. This method ensures that the content in the emails pulls dynamic and accurate data directly from Salesforce records, such as lead or contact information, ensuring the follow-up is relevant and customized based on the specific record.

Record Merge Fields ensure the generated emails are highly personalized using data like lead name, company, or other Salesforce fields directly from the records.

Apex Merge Fields are typically more suited for advanced, custom logic-driven scenarios but are not the most straightforward for this use case.

Automatic grounding using Draft with Einstein is a different feature where Einstein automatically drafts the email, but it does not specifically ground the content with record-specific data like Record Merge Fields.

Salesforce Einstein Sales Emails Documentation: https://help.salesforce.com/s/articleView?id=release-notes.rn_einstein_sales_emails.htm

The most critical technical consideration when pairing a prompt template ' s grounding data with a chosen Large Language Model (LLM) is the relationship between the two. The correct action is to review the model limitation in Prompt Builder versus the grounding data size (C).

Every LLM has a fixed context window limit, typically expressed in tokens (the model ' s units for processing text). This token limit defines the maximum amount of input data (the prompt template text + all the dynamic grounding data) and output data the model can handle in a single request.

The grounding data, which is pulled dynamically from Salesforce records (e.g., related lists, long text fields, Flow outputs), varies significantly in size from one record to the next. If the combined size of the prompt and the dynamic data for a specific record exceeds the LLM ' s token limit, the generative AI request will fail with a " token limit exceeded " error. The Agentforce Specialist must proactively design the template to limit the amount of data retrieved (e.g., using Flow to summarize related lists or querying only essential fields) to ensure it stays within the chosen model ' s capacity.

Option A is incorrect because the Einstein Trust Layer ' s token limit primarily relates to PII masking and is a security-related capacity, not the fundamental model ' s context window. Option B is incorrect because OFFSET is a SOQL query function used for pagination, which is irrelevant to ensuring the total size of the final assembled prompt (template + data) fits within the model ' s token limit.

Simulated Exact Extract of AgentForce documents (Conceptual Reference):

" A major challenge in prompt template design is managing the Large Language Model (LLM) token limit against the volume of grounding data. The specialist must always Review the model limitation in Prompt Builder versus the grounding data size before activation. LLM context windows (token limits) are fixed per model, but dynamic prompt components—such as merge fields from related lists or long text area fields—can cause the total size of the prompt to vary significantly by record. To prevent random token limit failures, the prompt instructions and grounding logic (Flow/Apex) must be explicitly constrained to retrieve only the essential data required to answer the query, ensuring the combined input stays well below the LLM ' s defined capacity. "

Simulated Reference: AgentForce Prompt Builder Best Practices Guide, Section 4: Performance and Scalability, p. 92.

UC’s custom prompt for summarizing case records is underperforming, and we need to identify a likely cause. Let’s evaluate the options based on Agentforce and Einstein Generative AI mechanics.

Option A: The prompt template version is incompatible with the chosen LLM.Prompt templates in Agentforce are designed to work with the Atlas Reasoning Engine, which abstracts the underlying large language model (LLM). Salesforce manages compatibility between prompt templates and LLMs, and there’s no user-facing versioning that directly ties to LLM compatibility. This option is unlikely and not a common issue per documentation.

Option B: The data being used for grounding is incorrect or incomplete.Grounding is the process of providing context (e.g., case record data) to the AI via prompt templates. If the grounding data—sourced from Record Snapshots, Data Cloud, or other integrations—is incorrect (e.g., wrong fields mapped) or incomplete (e.g., missing key case details), the summaries will be inaccurate. For example, if the prompt relies on Case.Subject but the field is empty or not included, the output will miss critical information. This is a frequent cause of poor performance in generative AI and aligns with Salesforce troubleshooting guidance, making it the correct answer.

Option C: The Einstein Trust Layer is incorrectly configured.The Einstein Trust Layer enforces guardrails (e.g., toxicity filtering, data masking) to ensure safe and compliant AI outputs. Misconfiguration might block content or alter tone, but it’s unlikely to cause summaries to lack appropriate information unless specific fields are masked unnecessarily. This is less probable than grounding issues and not a primary explanation here.

Why Option B is Correct:

Incorrect or incomplete grounding data is a well-documented reason for subpar AI outputs in Agentforce. It directly affects the quality of case summaries, and specialists are advised to verify grounding sources (e.g., field mappings, Data Cloud queries) when troubleshooting, as per official guidelines.