Agentforce Frequently Asked Questions (FAQs)

【Contract & Licensing】

Q: I have contracted for Agentforce, but I don't know where to start.

A: We offer various contents to help you leverage Agentforce. Please feel free to consult your Account Executive (AE) first. 
For self-paced learning, you can find various resources, including Trailhead modules and videos, in the following documentation.

-Design and Implement Agents

Q: What can I do with my Agentforce license? 

A: Available features vary depending on your specific license. Please confirm with your Account Executive.

-Agentforce for Sales
-Einstein Features in Sales Cloud
-Supported Editions for AI Solutions for Service

Q: How can I check my Agentforce credit consumption?

A: You can check this via the Digital Wallet. Please refer to the article below for details on how to check usage and who to contact. Note that Data 360 (Data Cloud) must be enabled to view credit consumption within your org.

-About Digital Wallet

Q: I want to know my current Agentforce contract status and pricing.

A: Please contact your Account Executive for details regarding contracts and pricing. Also, refer to the following articles:

-Agentforce and Generative AI Usage and Billing
-FAQ on Einstein Request and Flex Credit Rate Card Updates

Q. I tried to configure credit consumption prevention settings, but I cannot find the "Unmetered User-Based AI" permission set in my org.

A. The "Unmetered User-Based AI" permission set is only granted to organizations that have purchased add-on licenses such as "Agentforce for Sales" or "Agentforce for Service" (Agentforce for X), or "Agentforce 1 Edition" (Sales, Service, Field Service, Industries, etc.) which includes these capabilities.In organizations that do not include the above add-on licenses, the license and permission set will not be granted and will not be displayed.Therefore, if you have not purchased the applicable add-on license, there is no need to configure credit consumption prevention using that permission set.

For details about your specific contract situation or if you are considering adding an add-on license in the future, please contact your Salesforce account executive.

Additionally, if you have already purchased the applicable add-on license and the "Unmetered User-Based AI" permission set exists in your production org but is not found in your Sandbox org, you will need to perform a license sync with production on the Sandbox.

-Configure Unmetered User-Based AI Permissions

Q: Agentforce credit consumption is higher than expected, and I don't know why.

A: Please contact Salesforce Support with the information requested at the end of this article.

【Prompts & Topic Instructions】

Q: How do I write effective prompts?

A: Best practices and techniques are available in our documentation and Trailhead. We also provide various contents to help you leverage Agentforce; please consult your Account Executive. 
Note: Since prompt engineering is a general skill rather than a Salesforce-proprietary technology, Technical Support cannot provide specific guidance on prompt writing.

-Best Practices for Building Prompt Templates
-Prompt Engineering Techniques
-Prompt Fundamentals

Q: I don't know how to write instructions for Agentforce Agent Topics.

A: We recommend trying it out by referring to the standard topic/action instructions in your org, Trailhead (Agentblazer), and related documentation. 
Technical Support can guide you on how to configure topics/actions and general functional specifications. However, the specific content of instructions or procedures falls under general prompt engineering, and we cannot provide specific content creation assistance.

-How to Write Effective Natural Language Instructions for Agentforce
-Best Practices for Topic Instructions in the Legacy Builder
-Best Practices for Agent Action Instructions

【Troubleshooting】

Q: I created an Agentforce Agent, but it is not working as expected.

A: First, identify exactly which part of the behavior deviates from your expectations.
If the Agent is not selecting the correct topic or action: There may be an issue with the Instructions or the classification.
Note: If instructions are written in languages other than English (e.g., Japanese), the reasoning engine may not perform as expected.
If the Agent selects the correct topic/action but returns incorrect content: There may be issues with action inputs, permissions, or the underlying Flow/Prompt/Apex logic.

-Agentforce Considerations

Q: Agentforce responses are appearing as HTML code.

A: If "Show in Conversation" is enabled in the Action output settings and the output is HTML, it will be displayed as-is. Try turning off "Show in Conversation" and check the behavior again.

-Creating Custom Agent Actions in the Legacy Builder

Q: I cannot create an Agentforce Data Library (ADL), or the loaded info is not reflected in the Agent.

A: Please check the following:
-Is Data 360 (Data Cloud) enabled?
-Is the ADL status "Ready"?
-Run the following query in the Query Editor for the corresponding chunk dlm: 
SELECT "Chunk__c","SourceRecordId__c" FROM “<AGENT_DATA_LIBRARY_NAME>_chunk__dlm" WHERE "SourceRecordId__c" LIKE '%<UPLOADED_FILENAME>%' LIMIT 10
Does the Agent execution user have Read permission for the Data Space where the ADL was created?
-For Knowledge-sourced ADLs: Does the user have "Allow View Knowledge," Read access to the Knowledge object, and FLS (Field Level Security) Read access to all fields selected in the ADL?
-Are the search keywords related to the content within the files?

Q: I want to deploy Agentforce from Sandbox to Production.

A: Please consider metadata migration using Salesforce CLI or DevOps Center.

-Metadata Types | Agentforce Developer Guide
-AI Agent Deployment with Agentforce Overview

Q: I want to package Agentforce.

A: Currently, the Agent itself cannot be packaged directly. You must create an Agent Template, include it in the package, and create a new Agent from that template in the target org.

-Package Agentforce Metadata Components | Managed Packaging Guide

【Contacting Support】

Q: The issue is not resolved after reviewing the FAQ.

A: Please contact Support with the following information (add/remove details as necessary):

Subject: [Summary of the inquiry]
Description: [Details of the inquiry]
Target Org ID:
Username (Login Access Granted):
Agent Name:
Agent Topic Name:
Agent Action Name:
Agent Data Library Name:
Expected Behavior:
Current Behavior:
Steps to Reproduce:
Permission for Support to execute reproduction steps: [Yes/No]
Severity: 4
Product Name: Agentforce