After creating and embedding your AI Agent into a process, you can make it available to business users by running it in a public workspace. Public workspaces allow business users to access and run automations built by developers.

This topic explains the minimum setup, permissions, and configuration required to execute a process containing an AI Agent.

Note: AI Agents cannot be executed directly in a public workspace. They must be part of a process that is checked in and is available in the public workspace.

Prerequisites: Users and Roles

Running a process that includes an AI Agent in a public workspace requires specific roles and permissions. The following roles, described in Users and roles for Automation Co-Pilot, work together during process execution:

User type Roles/Permissions required Purpose in Co-Pilot Execution
Co-Pilot Admin AAE_Robotic Interface Admin Controls everything related to the Co-Pilot. This user creates teams, assigns processes, and configures scheduler users and device pools.
Co-Pilot User (Business user) AAE_Robotic Interface User Starts and runs public processes that include AI Agents. Executes processes through Automation Co-Pilot.
Scheduler User Scheduler permission (within a custom role) Acts as the bridge between the Control Room and Co-Pilot. The Scheduler receive requests and assigns jobs to Bot Runners. All public processes are consumed and executed under this user.
Bot Runner (Attended or Unattended) Custom role (with access to Model connections) The device that runs the job fired off from Co-Pilot. For AI Agent processes, the Bot Runner must have access to the Model connection via the custom role. This is a key difference compared to running normal processes.

Before proceeding, ensure you have created and configured an AI Agent in the Control Room and embedded it into a process. For more details about these above roles, see: Users and roles for Automation Co-Pilot and Create a process automation in Process Composer

Procedure: Configuration and Execution

Configuration involves settings in both the Control Room and Automation Co-Pilot.

Step1: Control Room Setup (Admin)
  1. Check in the process: Ensure the process containing the AI Agent is checked in to the public workspace.
  2. Verify dependencies: Confirm that all dependent automations (for example, Task Bots and API Tasks), are checked in and available.
  3. Confirm model connection access: Verify that a valid Model connection is available and shared with the required roles (for example, the Bot Runner or a user with the GenAI Admin role).
  4. Create a device pool: Public processes do not necessarily run on a local device. Create a device pool that includes:
    • Devices available to run public processes.
    • A consumer role (Co-Pilot User role) allowed to use the device pool.
    • The Scheduler uses this pool to determine which device will execute the process. For more information, refer to Manage a process in Automation Co-Pilot.
Automation Co-Pilot Management Configuration

The AAE_Robotic Interface Admin configures the AI Agent process settings.

  1. Access Co-Pilot Management: Log in to Co-Pilot as the Robotic Interface Admin (who has the AAE_Robotic Interface Admin role) and go to Manage.
  2. Create a Team:
    • Navigate to Manage > Team
    • Create a team and assign a team admin.
    • Select the Control Room roles that can run processes for this team. For more information, refer to Create a team and assign roles to team members.
      Tip: You cannot run any process or agent in Co-Pilot without first creating a Team.
  3. Assign the Process or Agent: Next, associate the public process (which contains the AI Agent) with the team you just created.
    • Navigate to Manage > Processes.
    • Select the checked-in public process (for example, List Reversal Agent).
    • Assign the process to the created team. If the process is not assigned to a team, it will not appear in the Co-Pilot user interface.
  4. Configure Scheduler User and Device Pool:
    • Select the Scheduler user. All public processes run as this user.
    • Ensure the Scheduler has the appropriate Scheduler permission and access to the configured device pool.
    For more information on setup guidance, refer to Create a process automation in Process Composer.
Agent Execution (Co-Pilot user)
  1. A business user (Co-Pilot User) logs into Automation Co-Pilot.
  2. The assigned process appears in the Co-Pilot interface if the user is part of the corresponding team.
  3. The user selects the process (for example, List Reversal Agent) and chooses Run.
  4. Enter required input values when prompted, then select Confirm.
  5. The Scheduler passes the execution request to the Control Room.
  6. The Control Room runs the process on an available device in the device pool using the Bot Runner role permissions.
  7. The AI Agent executes using its configured model and tools, returning results through Co-Pilot or the defined output path.

The AI Agent executes using its configured model and tools, returning the results through Co-Pilot or the configured output path.