Loading
시스템 관리자에 대한 피싱 방지 MFA 및 전 직원사용자 MFA 적용 안내 더 많이 읽기
Build AI Solutions for Service
Service Plan Record Processing

Service Plan Record Processing

Learn how guidance plans and dynamic plans process record data and how Service Assistant monitors each record type for updates during the resolution workflow.

Required Editions

View supported editions.
Checkmark This article applies to: Service Assistant for Case and Messaging

Service Assistant is a proactive agent on the record. It builds the service plan from your grounding sources. Guidance plans and dynamic plans process record data differently. The key distinction is how Service Assistant monitors a record for new information and how it incorporates those updates during the resolution workflow. A guidance plan is drafted once and stays fixed until a service rep redrafts it. A dynamic plan updates continuously as the record changes. Service Assistant monitors the record in real time and runs a fresh search of your grounding sources at each step.

How Dynamic Plans Process Records

A dynamic plan is a real-time, conversational resolution workflow available for both cases and messaging sessions. Unlike the static checklist in a guidance plan, a dynamic plan generates each step one at a time as the case or session progresses. At each step, Service Assistant performs a fresh retrieval of all your grounding sources and the latest record data to produce precise, current guidance. Service Assistant can also surface and execute agent actions to help service reps automate steps directly in the plan.

Here's how a dynamic plan progresses.

  • Initiation: When a service rep opens an eligible record, Service Assistant summarizes the conversation and identifies the assigned subagent. The service rep selects Start Plan to begin the interactive workflow. If you turn on Auto Run Dynamic Plan After Summary for Messaging, the first step surfaces automatically after the summary generates, without the service rep selecting Start Plan.
  • Real-time step formation: Service Assistant creates each step in real time, performing a fresh retrieval of all grounding sources and the latest record data. It also grounds the step in the component feed, so each step builds on the work completed so far, including the earlier steps, agent chat, knowledge looked up, and actions run. It evaluates whether an agent action is available to automate the task.
  • Step execution: Steps are presented as manual tasks or as automated steps that use agent actions. When an action is found, Service Assistant surfaces it for the service rep to confirm and execute, or the action executes automatically based on your configuration.
  • Plan collaboration: At any point, service reps can collaborate through chat to ask questions, request clarifications, or research without leaving the plan.
  • Active monitoring and adaptation: Service Assistant keeps the plan current without manual refreshes, but what it monitors depends on the channel. For cases, it monitors the fields and objects set in your Service AI Grounding configuration. For messaging sessions, it monitors the entire conversation transcript. How quickly an update is incorporated depends on the channel and the type of change.
  • Subagent switching: Service Assistant detects the customer's intent as the case or messaging session progresses and can switch to the relevant subagent to provide guidance from that subagent and related knowledge information. If the resolution starts with one subagent's issue but shifts to a second issue, Service Assistant switches to the second subagent, gathers the information it needs, and resolves that part of the plan. When the second issue is resolved, Service Assistant can automatically switch back to the original subagent. In some instances, Service Assistant doesn't switch back on its own. Use the chat to redirect it, such as "Let's go back to the original issue."

How Case Dynamic Plans Process Records

Service Assistant uses Service AI Grounding to ground itself in the details of the active case the service rep is handling. This grounding lets Service Assistant summarize the case for the service rep using the fields and objects (Case Email, Case Comments, and Case Feed) you set in the configuration.

  • Initiation: By default, Subject and Description are included as required grounding sources and can't be modified. These fields are used specifically to assign a subagent to the case record using the data inside them. No other fields or objects set in your configuration are used to assign a subagent, so the amount and detail of the Subject and Description fields play an important role in matching the case to a related subagent.
  • Record context: You can set additional fields and objects as grounding sources, such as Case Comments, Case Feed, and Case Email. However, the fields and objects serve only to provide richer grounding content for creating the plan steps and help to retrieve relevant knowledge information to include in plan steps. The Subject and Description fields are the only grounding source used to identify the issue and assign a subagent to the case.
  • Active monitoring and adaptation: During a dynamic plan, Service Assistant uses the fields and objects to actively monitor the record. It tracks any changes or updates to the fields and objects and can update the steps in real time based on the new information. However, Service Assistant updates the step the service rep is currently on only when new information arrives from a case email. For all other fields and objects, Service Assistant tracks the new information and incorporates the updated information into the next step of the plan rather than the active step.
  • Action matching: As it creates each step, Service Assistant also uses the record detail and grounding sources to look for relevant agent actions. It matches the record data, subagent instructions, and knowledge information that ground the step against your action descriptions. When it finds a related action that matches the context, Service Assistant surfaces the action in the plan step so the service rep can automate the task. Clear, specific action descriptions help Service Assistant make the right match. See Grounding Service Assistant with Agent Actions.
  • Related records processing: When a case is created from a closed messaging session or a voice call, Service Assistant analyzes the record transcripts as a grounding source, so the case summary and plan steps include key details from the conversation. The plan starts the steps from where the conversation left off. Messaging sessions and voice calls are the only supported related record types, and each record must be closed to be used as a grounding source. Related record grounding is supported for both guidance and dynamic plans. See Grounding Service Assistant with Related Records (Optional) (Case Only).

How Messaging Dynamic Plans Process Records

Unlike case processing, which tracks specific field updates, Service Assistant grounds itself directly in the full lifecycle of the conversation transcript.

  • Initiation: Service Assistant generates a service plan only after the conversation has messages to ground in. Monitoring begins automatically when the session meets your configured eligibility criteria, the conversation contains enough context for Service Assistant to identify the issue, and the messaging session record is owned by a human service rep. If the record is owned by any other user, Service Assistant doesn't process the record or generate a plan. There's no standard threshold for the number of messages needed to start the service plan. What counts as enough context varies and depends on the clarity of the messages and the intent Service Assistant can identify in the chat. Problem messages, such as vague, off-topic, or conflicting messages, can disrupt this process and delay when the plan starts.
  • Record context: Service Assistant draws conversation and customer context strictly from the transcript and the MessagingSessionId context variable. It processes conversation data only from the current messaging session, and ignores previous sessions even when they're linked to the same conversation record. Service Assistant references only text in the conversation. It doesn't process file uploads, images, or choice responses.
  • Active monitoring and adaptation: Grounding in a messaging session is an active, fluid process. As the conversation unfolds, Service Assistant monitors each new message in the transcript and continuously adapts the plan, so the steps always reflect where the conversation currently stands. For each step it creates, Service Assistant performs a real-time search against all your grounding sources to make sure the step is the most relevant to the current context. As soon as the context shifts, Service Assistant adjusts the plan and creates a new step to match.
  • Action matching: As it creates each step, Service Assistant also uses the conversation context and grounding sources to look for relevant agent actions. It matches the transcript, subagent instructions, and knowledge information that ground the step against your action descriptions. When it finds a related action that matches the context, Service Assistant surfaces the action in the plan step so the service rep can automate the task. Clear, specific action descriptions help Service Assistant make the right match. See Grounding Service Assistant with Agent Actions.
  • Related records: Service Assistant for Messaging doesn't support related record grounding. It doesn't automatically receive context from the messaging session record or related records such as Contact or Case. To include related record data in the service plan, build an agent action that takes currentRecordId as input and queries the fields you need. For messaging sessions, use EndUserContactId to retrieve the associated Contact for customers already in your CRM, or MessagingEndUserId for customers identified through a messaging channel. This action returns the relevant fields for Service Assistant to produce more context-based plan steps.
  • Session end: When the messaging session ends, the service plan ends, and Service Assistant no longer provides proactive, step-by-step assistance. Service Assistant and the chat box stay available for about 24 hours after the session ends, so service reps can complete wrap-up tasks such as asking for a session summary, drafting a follow-up email, asking knowledge questions, running actions, or getting general help using the chat. All interactions are saved to the component feed.

How Voice Dynamic Plans Process Records

Like messaging processing, Service Assistant grounds itself directly in the voice call transcript rather than in specific record fields. After the plan starts, Service Assistant continues to analyze and monitor each new transcript segment and generates one step at a time using your grounding sources.

  • Initiation: Service Assistant generates a service plan only after the voice call transcript has enough context to ground in. Using the transcript, Service Assistant evaluates the conversation against your subagent descriptions and scopes to identify the issue. When it has enough context, it assigns the matching subagent and starts the plan. Unlike case and messaging processing, voice processing doesn't show a plan summary or a Start Plan button. Service Assistant generates the first step of the plan right away, based on the assigned subagent's instructions. This process happens automatically. Monitoring begins when the record meets your configured eligibility criteria and the Voice Call record is owned by a human service rep. If the record is owned by any other user, Service Assistant doesn't process the record or generate a plan.
  • Record context: By default, Service Assistant draws conversation and customer context only from the transcript on the Voice Call record page. No setup is required beyond creating subagents and instructions for the call types that you intend to support. For access to the Voice Call record and transcript, Service Assistant requires the Agent Voice Access custom permission set, and you can install the component only on the Voice Call record page. Service Assistant can't access the voice call transcript when it's installed on other record pages, such as Case or Account. See Service Assistant for Voice Permissions.
  • Active monitoring and adaptation: Grounding in a voice call is an active, fluid process. As the conversation unfolds, Service Assistant monitors each new transcript segment and continuously adapts the plan, so the steps always reflect where the conversation currently stands. For each step it creates, Service Assistant performs a real-time search against all your grounding sources to make sure the step is the most relevant to the current context. As soon as the context shifts, Service Assistant adjusts the plan and creates a new step to match.
  • Action matching: As it creates each step, Service Assistant also uses the conversation context and grounding sources to look for relevant agent actions. It matches the transcript, subagent instructions, and knowledge information that ground the step against your action descriptions. When it finds a related action that matches the context, Service Assistant surfaces the action in the plan step so the service rep can automate the task. Clear, specific action descriptions help Service Assistant make the right match. See Grounding Service Assistant with Agent Actions.
  • Related records: Service Assistant grounds only on the voice call transcript. It doesn't automatically access related records attached to the voice call, such as an associated Contact or Account. To include related record data in the service plan, build an Apex action that takes VoiceCall.Id as input and queries the RelatedRecordId field to retrieve the associated record data. Service Assistant incorporates the returned data when it creates a plan.
  • Subagent switching: Service Assistant detects the customer's intent as the voice call progresses and can switch to the relevant subagent to provide guidance from that subagent and related knowledge information. Because a voice call is a fluid, live conversation, subagent switching can happen more often than in case or messaging processing, and the plan updates continuously in real time. If the resolution starts with one subagent's issue but shifts to a second issue, Service Assistant switches to the second subagent, gathers the information it needs, and resolves that part of the plan. When the second issue is resolved, Service Assistant can automatically switch back to the original subagent. In some instances, Service Assistant doesn't switch back on its own. Use the chat to redirect it, such as "Let's go back to the original issue."
  • Real-time step adaptation: Because of the active, real-time nature of a voice call, plan steps are more fluid than in case or messaging processing. Service Assistant adapts and regenerates steps in real time as the conversation unfolds, so the plan can change frequently to stay current with what's being said on the call.
  • Call end: When the voice call ends, the service plan ends, and Service Assistant no longer provides proactive, step-by-step assistance. Service Assistant and the chat box stay available for about 24 hours after the call ends, so service reps can complete wrap-up tasks such as asking for a call summary, drafting a follow-up email, asking knowledge questions, running actions, or getting general help using the chat. All interactions are saved to the component feed.

Transcript Quality for Voice

Service Assistant's ability to assign a subagent to the issue and create a step depends on the quality and speed of the transcription.

  • Plan generation can be delayed at the start of a call. Before Service Assistant generates the summary and first step of the plan, it requires enough messages in the transcript to understand the context of the call, including the issue to resolve. If the transcript hasn't caught up to the conversation, the summary can take longer to appear. You can see a message such as, "There's not enough information to draft a service plan summary. Add more details, then check back here." Plan processing starts when enough transcript text is available. What counts as enough information varies and depends on the clarity and speed of the transcription and how quickly the customer surfaces the reason for the call.
  • Plan quality depends on transcript quality. If your transcription provider drops words, introduces errors, or delivers text in large delayed batches rather than incrementally, the plan can be less accurate or less timely. Transcript quality also affects agent actions. If the transcript doesn't clearly capture the context that triggers a specific action, that action might not surface as a step in the plan. For example, a customer requests a refund but the transcription renders it as an unrelated phrase. Reps must be prepared to manually start agent actions or quick actions using the agent chat when the plan doesn't reflect what they're hearing on the call.

How Guidance Plans Process Records

A guidance plan is a static, checklist-style plan available for cases only. Like a case dynamic plan, Service Assistant grounds the plan in the fields and objects set in your Service AI Grounding configuration. But instead of building each step in real time, Service Assistant drafts the full checklist at once from the current record data, and the plan doesn't change until a service rep redrafts it.

  • Initiation: When a service rep opens an eligible case, Service Assistant summarizes the case and assigns a subagent from the case Subject and Description. The service rep selects Draft Plan to receive a static checklist of steps organized into resolution phases.
  • Step formation: When the plan is drafted, Service Assistant creates the full checklist from the latest record data and your grounding sources, including subagent instructions and knowledge articles. The plan doesn't regenerate as the case progresses. To get updated steps, a service rep redrafts the plan.
  • Step execution: The checklist is fully manual. Service reps work through the steps and mark each one off themselves. To surface task automation directly in a step, add quick actions, which show as buttons for service reps to launch the quick action workflows. Agent actions aren't surfaced in guidance plan steps, but service reps can run them through agent chat.
  • Plan collaboration: Agent chat is available as an opt-in extension. Turn on the chat feature on the Service Assistant Setup page. Service Assistant tracks which steps a service rep marks off, but service reps can't ask it to mark a step complete or run the checklist for them. Because the plan isn't proactive, a service rep runs an agent action by asking Service Assistant through chat, such as drafting an email, and then marks the related step complete.
  • Manual updates: The plan doesn't automatically adapt to case updates. When a case field or object set in your Service AI Grounding configuration changes, the Redraft Plan button shows, and the service rep redrafts the plan to incorporate the updates.
 
로드 중
Salesforce Help | Article