Salesforce Docs Portal

Assign a Task to an AI Agent in Agentforce Operations

Assign a task to an AI agent in Agentforce Operations and provide the inputs needed for reliable completion.

Required Editions

License Required

This feature is a workspace that’s external to your Salesforce org and requires an Agentforce Operations license.

To purchase an Agentforce Operations license, contact your Salesforce account executive.
Role or Access Needed
To manage tasks in Agentforce Operations: Admin role or Creator role

Before assigning the task, determine which agent is best suited for the expected output and prepare the required context and data. For a list of available agents and their requirements, see AI Agents Catalog for Agentforce Operations.

  1. Create a blueprint or edit an existing blueprint.
  2. In the blueprint, in a stage, click + Task.
  3. Enter a task name.
    Use a clear name that describes the work, such as Validate invoice pricing.
  4. For Start, select when to start the task.
    1. To start the task when the stage starts, select When the stage starts.
    2. To start the task after another task in the stage ends, select After another task ends. Then select the task to complete before this task starts.
    3. To start the task on a specific date, select On a specific date. Then click calendar icon or enter numbers for month, day, and year.
  5. (Optional) For Due, configure a deadline for the task to finish by.
    1. To set no due date, leave the default configuration.
    2. To set a due date relative to when the task starts, select Time to complete. Then enter a number and select Hours, Days, or Minutes.
    3. To set a due date from a date field created by an upstream task, select From a field. Then select the date field.
  6. For Assigned to, select an AI agent and configure related options.
    1. Select an AI agent to assign the task to.
    2. Optionally, to configure a fallback assignee in the case that the agent fails, for Fallback assignees, select who to reassign the task to. You can select users, teams, workflow owner, and workflow creator. You can also select a field from an upstream task that contains a user, team, workflow owner, or workflow creator.
    3. Optionally, to give other users access to details and notifications about this task, click + Add Cc. For Cc, select users, teams, workflow owner, workflow creator. You can also select a field from an upstream task that contains a user, team, workflow owner, or workflow creator.
  7. For Description add clear instructions for the AI agent.
    Write step-by-step instructions for multi-step work. Include identifiers, dates, business rules, expected output format, and any edge cases or nuances that can occur. To include a field value in the description, type @ followed by the field name.
  8. Optionally, for Documents, add files for the agent to review.

    Supported formats: .pdf, .docx, .xlsx, .png, and .jpg.

  9. Optionally, add Tags to classify the task.
    Tags added here apply to this task only and don't affect the blueprint. Unlike tags on blueprints and workflows, tags on tasks don't affect filtering or views in Agentforce Operations.
  10. Select fields from upstream tasks that the agent uses,
    1. In the Share information section, click + Select field, and then select the fields to add.
    2. Click Apply.

    Fields listed in the Select fields window include existing fields that have values stored in them by upstream tasks.

    For example, in Share information select the Invoice Price, Contract Price, and Invoice Date fields.
  11. In the Request information section, select fields for the agent to store values in.
    1. To select specific fields without creating a form, for each field, click + Add field or role, and then select a field.
    2. To add fields within a form, click + Add form, and then select or create a form.
    For example, in Request information, select the Price Match Result and Review Notes fields.
  12. Save your task and then reopen it.
    1. To save the task, click Add.
      Agentforce Operations tells you that the agent is missing information.
    2. Click Save anyway.
      The task appears in your blueprint.
    3. Reopen the task.

    Before it can run, a task assigned to an agent requires a saved plan.

  13. Test and save the agent's plan.
    1. Click the Test tab.
    2. Add inputs for each Shared Information field.
    3. Click Run test.
    4. Review the test data in the Requested Information section, and review the plan in the AI AGENT ACTIVITY panel. The plan shows the exact steps that the agent follows on every workflow run.
    5. If the plan looks correct for the inputs you provided, test again with different inputs to verify that the agent accurately handles all expected scenarios. If the agent produces unexpected results, adjust the task configuration and test again.
    6. When the plan performs according to expectations, click Save changes.

    AI agent tasks with missing or unsaved plans block blueprint publishing. For more information, see Get AI Agent Tasks Ready for Run Time in Agentforce Operations.

See Also