Service Request Resolution Workflow Advisor Enhancements

AI agent

The Service Request Resolution Workflow Advisor is enhanced with Human-in-the-Loop (HIL), follow-up handling, and rejection handling within the resolution workflow. These capabilities allow organizations to control when a Service Request is resolved autonomously, when it is routed for human review, and how follow-up or rejected interactions are handled without breaking the workflow.

With HIL, the workflow can route selected resolutions for CSR review through an Action Plan rather than sending a response directly to the customer. With follow-up handling, customers or CSRs can continue the interaction with additional details or clarifications while preserving context. With rejection handling, the workflow can refine the resolution when an outcome is not accepted, using bounded retries to avoid repeated execution and infinite loops.

These enhancements are particularly valuable during early stages of AI adoption, in sensitive or high-impact scenarios, and in environments where additional oversight is required for less mature processes or products.

What is introduced

Human-in-the-Loop (HIL)

  • HIL flag (default: false) is set via Automation Rules and passed with the Service Request to the resolution workflow
  • HIL = false: Resolution is generated and sent directly to the customer; the Service Request is marked resolved
  • HIL = true: Resolution is routed as an Action Plan to CSR for review before customer communication

Follow-up Handling

  • Customer interactions such as questions or additional details are routed back into the same workflow
  • CSR can refine, adjust, or validate responses before final resolution
  • The workflow maintains the same interaction context across follow-ups

Rejection Handling

  • Detects when a resolution is not accepted and avoids blind re-execution
  • Applies bounded retries (default: up to 3 attempts) to prevent infinite loops

Business Value:

  • Controlled automation: apply HIL for governance in sensitive or high-impact scenarios
  • Higher resolution accuracy: iterative follow-ups and rejection handling refine outcomes
  • Eliminates rework loops: avoids repeated, ineffective executions with bounded retries
  • Better customer experience: supports natural interactions (questions, clarifications) within the same flow
  • Operational efficiency: keeps review and retries within a single workflow

Steps to enable and configure

Prerequisite: Your environment must be on 26C or later.

Create an Automation Rule, configure its conditions, and add an AI Agent Action to invoke the Resolution Workflow. Configure the action inputs, deploy the rule, and test it with a service request. The example below enables human review by setting HIL to true.

If you use a custom workflow created from templates, configure and publish it before selecting it in the action.

Step 0 (Optional): Configure the AI Agent Team in Agent Studio

Follow this step only if you want to use a custom agent built from templates

  • Template name: Service Request Resolution Workflow Advisor
  • Navigate to: Tools > Agent Studio
  • If you use a template-based custom agent, clone both Service Request Resolution Assistant and Service Request Resolution Workflow Advisor templates, make the required changes, and then publish the agents
  • Adjust the agent team Name and Team Code to match your custom configuration
  • Use the same naming and code values consistently across Agent Studio and Automation Rules

Step 1: Configure Channel Id in the AI Agent workflow

This step ensures responses are routed through the correct communication channel (for example, Email)

  • Retrieve the channels in your environment using crmRestApi/resources/11.13.18.05/channels. Identify the applicable email channel with ChannelTypeCd = ORA_SVC_EMAIL and record its ChannelId. If multiple email channels exist, select the channel appropriate to your service process. The ChannelId can differ between environments.
  • Go to the Agent Studio, open the Service Request Resolution Workflow Advisor agent, and under Send Agent Response Node, replace the fetch message BO with Inbound SR Messages
  • Update ChannelId in the response payload with the identifier of the selected email channel.
  • If needed, validate the channel mapping before deploying to production

Example response payload:

let messageContentVar = $context.$nodes.REQUEST_INFO_EVAL.$output.Summary;
let partyViaEndPointVar = $context.$nodes.GET_SR_DATA.$output.PrimaryContactEmailAddress;
let partyIdVar = $context.$nodes.GET_SR_DATA.$output.PrimaryContactPartyId;
let title = $context.$nodes.GET_SR_DATA.$output.Title;
let srNumber = $context.$triggers.REST.$input.objectNumber;

const payload = {
  MessageTypeCd: "ORA_SVC_RESPONSE",
  ChannelTypeCd: "ORA_SVC_EMAIL",
  StatusCd: "ORA_SVC_COMMITTED",
  SourceCd: "ORA_SVC_REDWOOD_AGENT_UI",
  Subject: srNumber + " - " + title + " (AI Agent Generated)",
  MessageContent: messageContentVar
};

if (partyViaEndPointVar) {
  payload.channelCommunication = [
    {
      ChannelId: "<channelId>",
      PartyViaEndPoint: partyViaEndPointVar,
      RoutingCd: "ORA_SVC_TO",
      PartyId: partyIdVar
    }
  ];
}

return payload;


Note: Adapt this example to your workflow and validate it in a non-production environment.

Step 2: Create an Automation Rule

Create a rule and configure the following fields:

  • Automation Rules Template Name: Enter a descriptive name for the rule.
  • Entity Name: Select the entity for which the rule will run.
  • Stripe Code: Verify that the displayed stripe matches the intended application.
  • Description: Describe the purpose of the rule.
  • Event Type: Select Attribute Changes for an event-driven invocation.
  • When this happens: Configure the triggering event and conditions for your scenario.

Ensure Enabled is selected when the rule is ready for testing.

Step 3: Configure the AI Agent Action

Under Automation Rules Variables, hover over the play icon and click the plus icon.

  • To reuse a configured action, select it from the action catalog.
  • To create an action, choose Create Action, select AI Agent Action, and configure the fields below.

For either option, configure the action attributes for this rule.

  • Configure the action with the fields:
    • Action Type Object Name: Service request
    • Application: Oracle Fusion Applications
    • Authentication Context: Oracle AI Agent Studio
    • Family: CX
    • Product: Service
    • Agent Team Name: Select Service Request Resolution Workflow Advisor, or select your published custom workflow.
    • ActionType: Rest
    • Operation: Create
    • Action URL: Automatically populated when you select the Agent Team Name. For the delivered workflow, verify that the URL contains ORA_SERVICE_REQUEST_RESOLUTION_WORKFLOW_AGENT_TEAM_RA. For a custom workflow, verify that it contains your published Resolution Workflow code.
    •  Description: Automatically populated.
    • API User Type Code: Integration User
    • API User: Provide the username that will be used for authentication with AI Agent Studio

  • Configure these action attributes (Resolution Workflow Inputs):

    • parameters: Select JSON as the Attribute Source. In Attribute Value, enter a JSON object containing objectNumber, objectType, and the inputs required for your scenario.
    • conversationId: Select Parent as the Attribute Source and enter SrNumber as the Attribute Value.

The Example Configuration: Enable Human Review section below provides an example payload and explains these attributes.

Asynchronous Mode / Polling

Required for asynchronous execution and response polling from the AI Agent (Some details will be auto-populated)

  • Set to Async with Polling
  • Maximum Polling Count: 3
  • Success key: status
  • Success value: COMPLETE
  • Polling URL: Automatically populated when you select the agent or workflow.
  • Request Id key: jobId

Step 4: Complete and Test the Automation Rule

  • Verify the conditions, selected action, and action attributes.
  • Add other actions if required by your business process.
  • Save and deploy the rule in a nonproduction environment.
  • Test with a service request that meets the conditions and verify the expected result.

After validation, deploy the configuration to production using your organization’s deployment process.

Example Configuration: Enable human review

  • Conditions:
    • SR Severity = High
    • Channel = Email
  • Action attributes:
    • Attribute 1 (Resolution Workflow Inputs)
      • Attribute Name: parameters
      • Attribute Source: JSON
      • Attribute Value: Enter the following example JSON, adapted to your scenario  
{ "objectNumber": "{{$SrNumber}}", "objectType": "ServiceRequest", "HIL": true }
  • Attribute 2 (Conversation Identifier)
    • Attribute Name: conversationId
    • Attribute Source: Parent
    • Attribute Value: SrNumber

      • objectNumber and objectType identify the service request.

      • HIL enables human review.

      • The separate conversationId attribute associates subsequent invocations for the same service request with the same conversation, preserving interaction context.

Expected Result: The proposed resolution is routed to a customer service representative for review.

Note: Adapt the example conditions and inputs to your business process, and validate the rule as described in Step 4.

Follow-ups from UI

Configuration is required only for customization

  • For customization:
    • Use skipValidation = true, skipClassification = true, HIL = true
    • conversationContext should match the original AI workflow request message ID
    • For OOB scenarios, this is handled automatically

Tips and considerations

  • Use HIL selectively for high-risk or low-confidence scenarios
  • Gradually expand autonomous resolution as confidence improves
  • Configure and validate the workflow in a sandbox environment and promote to production after validation

Key resources

Use this documentation to ingest Product and Category Dataset as required: Prepare Category and Product Data Using Ingestion AI Agent for Service Request Automation AI Agents

Access requirements

  • Roles and permissions required to access Service Automation Rules (Formerly Service Workflow)
  • Roles and permissions required to access AI Agent Studio