--- schema_version: "1" id: 9133e06aa50b126524ea2ef61e07453ff6931e68f4ba7646fa2a91431b66b6f3 title: Grounding Service Assistant with Subagents collection: admin version: current locale: en-us portal_url: https://docs.salesforce.com/admin/current/en-us/service/264-0-0/sp_topics_start.md canonical_url: https://help.salesforce.com/s/articleView?id=service.sp_topics_start.htm&release=264.0.0&type=5 source_site: help.salesforce.com content_type: documentation content_hash: b92a33ce55308cf6546c2161d7c80d0904dbc647442ac11bbb0fd2d352d0c7fa description: Learn the role of subagents in Service Assistant and review guidelines for creating them. products: - Service Cloud - Service Intelligence release: "264" --- # Grounding Service Assistant with Subagents Learn the role of subagents in Service Assistant and review guidelines for creating them. ### Required Editions

- [Subagents and Service Assistant](https://help.salesforce.com/s/articleView?id=service.sp_topics_start.htm&release=264.0.0&type=5) - [Guidelines for Writing Service Assistant Subagents](https://help.salesforce.com/s/articleView?id=service.sp_topics_start.htm&release=264.0.0&type=5) - [Guidelines for Writing Service Assistant Instructions](https://help.salesforce.com/s/articleView?id=service.sp_topics_start.htm&release=264.0.0&type=5) - [Sample Service Assistant Subagent](https://help.salesforce.com/s/articleView?id=service.sp_topics_start.htm&release=264.0.0&type=5) ![Note](https://sf-zdocs-cdn-prod.zoominsoftware.com/tdta-service-support_intelligence-264-0-0-production-enus/dcd06a7b-b5ab-448b-a767-4e2e90481508/images/icon_note.png) Note - This section provides reference guidance for the creation of subagents in Service Assistant. Setup guidance follow the general Agentforce subagents setup, which is included in [Set Up Service Assistant](https://help.salesforce.com/s/articleView?id=service.sp_start_setup.htm&type=5&language=en_US). - Service Assistant is only supported in the [legacy Agentforce Builder](https://help.salesforce.com/s/articleView?id=ai.agent_builder_explore.htm&type=5&language=en_US). Starting in July 2026, the requirement that new agents can only be created in the new Agentforce Builder doesn’t apply to Service Assistant. You can continue to create and configure a Service Assistant agent in the legacy builder until further notice. See [Create Agents in the New Builder Only Starting July 2026](https://help.salesforce.com/s/articleView?id=release-notes.rn_agentforce_legacy_agent_creation.htm&release=262&type=5&language=en_US). - Beginning in April 2026, agent topics are now called subagents. There are no changes to functionality. During this transition, you may see a mix of the new and previous terms in our documentation. See [Topics Are Now Subagents](https://help.salesforce.com/s/articleView?id=release-notes.rn_einstein_agentforce_subagent_new_name.htm&release=260&type=5&language=en_US). This video outlines the best practices and configuration guidance found in this article and in [Subagent Best Practices](https://help.salesforce.com/s/articleView?id=service.sp_topics_pb.htm&language=en_US&type=5) and [Subagent Strategy](https://help.salesforce.com/s/articleView?id=service.sp_topics_strategy.htm&type=5&language=en_US). It provides an overview of how to create subagents and how to optimize them when you use knowledge grounding. Review all of the information in this article and subsequent articles for more specific details and guidance. ## Subagents and Service Assistant In Agentforce, AI agents use subagents, actions, and instructions (known as agent assets) to categorize user interactions and determine what specific tasks they need to complete for users. AI agents typically assist customers in resolving general issues and finding self-service solutions without the need for a service rep, like changing a password or making an appointment. In contrast, Service Assistant is an agent that gives specific and detailed guidance to your service reps so they can resolve customer cases directly. To make sure the guidance in a service plan is accurate and specific, you need to create subagents and instructions that precisely detail your service policies for each case type, such as “Return Request” or “Credit Card Declined.” Having detailed and specific subagents help Service Assistant determine what instructions are relevant for inclusion in a service plan based on case details. Keep in mind that we don't provide any preconfigured subagents for Service Assistant. You must create you own subagents and instructions following the provided guidelines. ## Guidelines for Writing Service Assistant Subagents We suggest reviewing these guidelines on how to write both subagents and instructions for Service Assistant. The guidelines highlight how to structure your subagents and instructions for Service Assistant so that the resolution guidance in a service plan is relevant to the case and your policies. Here are a few guidelines for creating subagents for Service Assistant.

Guideline	Details	Example
Proivde subagents with a clear and specific names.	Subagents should have a name that speficies the exact Case type you're trying to address.	Instead of using “Customer Inquiry”, use “Order Status Inquiry”. Customer Inquiry is too broad as it covers a wide range of cases.
Eliminate overlapping subagents.	Make sure your subagents are distinct from one another. Overlapping subagents make it hard for the agent to classify cases meaningfully.	Instead of including both “Manage Reservations” and “Confirm Reservation” as separate subagents, rename them to make the differences explicit. You can avoid overlap by consolidating them into one “Reservation Management” subagent.
Keep subagents singular.	Subagents work best when they encompass only one definition or concept. Combining different ideas into one subagent can result in inappropriate service plans.	Instead of combining two concepts into a single subagent like “Returns and Exchanges,” separate each concept into its own subagent, “Returns” and “Exchanges."
Provide encompassing subagent descriptions.	Use a single high-level subagent for a wide range of related issues by explicitly listing variations and keywords in the subagent's description field. This way you don’t need to create separate subagents for each specific variation of transactions. When you have subtypes defined in the description, Service Assistant is better able to match a single subagent to multiple transaction declined case types.	Subagent: Transaction Declined Description: Guide service reps in helping customers resolve declined credit card transactions. Issues are related to authorization failures, including declining reason codes such as insufficient funds, suspected fraud, incorrect card details, travel blocks, and daily spending limits Subagent: Return Request Description: Guide service reps through processing a customer's return request, including verifying eligibility and initiating the return in the system. Questions are related to return requests, which includes proof of return, return windows, product condition, and refund status.

## Guidelines for Writing Service Assistant Instructions When writing instructions for Service Assistant, think about the steps a service rep must complete for a particular case. Translate those steps into instructions with specificity and accuracy to your service policies and practices. Here's a set of sample instructions for a "Process Return" subagent with details that follow. ![Service Assistant instructions example](https://sf-zdocs-cdn-prod.zoominsoftware.com/tdta-service-support_intelligence-264-0-0-production-enus/dcd06a7b-b5ab-448b-a767-4e2e90481508/support_intelligence/images/service_planner/sp_topic_guidelines.png)

INSTRUCTION GUIDELINE	CONTEXT
(1) Use clear, conditional statements	Include conditional language in your instructions to consider the possible scenarios that can occur when working a case. This ensures that all the relevant information is accounted for and available for the agent to decide what to include in a plan. If the information isn’t included in the instructions, it won’t be in the plan.
(2) State the actions ‌that follow preceding instructions and be clear about expected outcomes.	Instructions should match the standard workflow for the specific issue. In the example, the instruction states what occurs given the previous instruction. Include language such as “When you have…”, and “After…”. For instructions that are mandatory for working the case, use language such as “Always include this as a step in the plan”. This makes sure that the information is always included in a plan. Keep in mind that you don’t need to add instructions in a logical order. Agentforce agents reference each instruction as standalone information, which allows them to understand and apply instructions without additional context. Service Assistant can logically organize all instructions into a sequenced service plan.
(3) Specify your service policies and practices	When writing instructions, be specific and refer to your company's established guidance for the given case type. Cover all your service policies and standards for a given case and outline the criteria and conditions that must be met for the instruction to apply to the given case. Make sure that each instruction only refers to one piece of information from your policy or procedures for working the Case type. Don’t combine multiple pieces of information in one instruction. Be sure to include your guidelines for tone, and specify any sample scripts or templates available to the service rep to aid them in working the case.
(4 - not shown in image) Omit sensitive and personally identifiable information.	Avoid including sensitive or confidential information in your instructions. That way, you make sure the information is safe to share and referenced by Service Assistant without concerns about privacy or security. Includes sensitive information: If the customer's discount code isn't working, create a support ticket and assign it to the lead specialist, Jane Doe, for review. Omits sensitive information: If the customer's discount code isn't working, create a support ticket and assign it to the lead specialist for review.

## Sample Service Assistant Subagent To better understand how to create subagents for Service Assistant, here’s an example “Return Request” subagent. The example is detailed, but the amount of information in your subagents and instructions varies based on the case type and complexity. To save time creating subagents and instructions, we recommend using Agentforce Data Libraries. The feature adds information from your knowledge articles as steps in a service plan, reducing the amount of details you need to include in instructions. See [Grounding Service Assistant with Knowledge (Optional)](https://help.salesforce.com/s/articleView?id=service.sp_knowled_topics.htm&release=264.0.0&type=5 "Use Agentforce Data Libraries to ground Service Assistant in your knowledge.").

SUBAGENT PART	EXAMPLE	CONTEXT
Label	Return Request	A name for the subagent.
Description	Assist service reps with handling customer questions related to return requests. Returns requests can also be called return processing, returns, refunds, return inquiries,‌ and authorize return. Questions are related to return transactions, which include proof of return, refund status, return documentation, return policy, return window, and return resolution steps.	A statement of the subagent’s purpose and the types of requests that are categorized under it. Include any company-specific terms related to the subagent that the customer may provide in the case.
Scope	Your job is to assist service reps in resolving customer inquiries regarding returns processing. Ensure they adhere to all internal policies and procedures. You must not handle inquiries unrelated to the return of goods.	The scope sets the boundaries for your agent and tells the agent what it can and can't do with a subagent. Notice that the scope states, “You must not handle inquiries unrelated to the return of goods.” This ensures that the agent doesn't suggest information beyond your grounding sources.
Instruction 1	Make sure the customer has provided the ‌required information: valid return date, customer name, and receipt number or order number. This step must always be in a plan for cases dealing with returns.	The first instruction is a must-do for the service rep. It's the critical first step of a service plan that must be completed to resolve the specific case type effectively. When a step is essential to resolving a case, clearly state that it's required. Include language such as: Always Must Required Confirm Verify At all times
Instruction 2	After you have the required information, you must confirm the purchase in the return management system.	The second instruction serves as a follow-up to the first. It can always apply or be conditional. After you have … After completing the first instruction … After verifying … If …, then …
Instruction 3	If the customer provides the full Return Number, then you search the return in the Returns Management System (RMS) using the return number and return date. If the customer provides the order number, then you search the return in the Order Management System (OMS) using the order number and return date. If the customer provides only the customer name, then you search the return in Salesforce using the customer name and return date.	Consider the scenarios that can occur after the required steps are completed. These scenarios typically include conditional language. Try to cover all situations as much as possible. Conditional statements make instructions clear and comprehensive, ensuring that service reps have ‌all the necessary information to work the case. If …, then … When …, then … Whenever …, then … Provided that …, then … Should …, then … This format ensures that the instructions are clear and comprehensive, covering a wide range of potential scenarios.
Instruction 4	If the purchase can't be found in the system, first check the information provided (Order Number, Return Request Date, Item Description, Return Authorization Number, or Tracking Number) for any errors or misspellings. If the information is correct and the purchase still can't be found, tell the customer that the purchase isn't showing in the system. Ask them to give any other information that can help, such as the email address used for the purchase, the payment method, or any order confirmation emails they got. If the customer's item is confirmed in the system and meets the return time frame, confirm the reason for the return and ask them to provide photos or a detailed description of the issue.	Continue to build your instructions in a logical order. This provides a clear pathway in the service plan for service reps to follow.
Instruction 5	Check that the product is in the right condition to be returned. Ask the customer to send you photos or a clear description of the product’s condition. If the condition doesn’t meet the qualifier, tell the customer the return can't be processed.	Be clear about the expected outcomes after the information is gathered.
Instruction 6	After you've gathered all the information, send a response to the customer using the Return Request Template to indicate the return process. This step must always be in a plan for cases dealing with returns.	Make sure to provide instructions that state what to do after all other steps are completed.
Instruction 7	If the customer isn't happy with the return process or the solution, offer to raise the issue to a higher-level support team or manager. Document the customer's concerns and the steps taken to address them.	Provide instructions for situations where the customer isn't satisfied with the support provided. This helps your service reps anticipate customer needs and improve customer experience.

- **[Service Assistant Subagent Best Practices](https://help.salesforce.com/s/articleView?id=service.sp_topics_pb.htm&release=264.0.0&type=5)** Learn our recommended best practices for creating subagents in Service Assistant. - **[Subagent Strategy for Service Assistant](https://help.salesforce.com/s/articleView?id=service.sp_topics_strategy.htm&release=264.0.0&type=5)** Learn how to strategize the implementation of your Service Assistant subagents for optimal granularity and knowledge grounding.