Agents and Human Tasks - FAQs
Building and publishing Agents
How do I enable Agent Builder?
An administrator must activate the feature, and the AI Command Center license must include it. Check feature activation and your permissions if the builder is unavailable. See enabling Agent Builder and Features.
How do I create my first Agent?
Create the automation in AI Command Center and design its workflow in Agent Builder. Add the required activities, configure their properties and data, then test the workflow before publishing. See Agent Builder getting started.
How do I pass data through an Agent workflow?
Use the Process Data drawer to configure model data, workflow parameters, form data, and other supported data sources. Choose the data category that matches the value's purpose and scope. See the Process Data drawer and workflow parameters.
How do I add a form and bind it to workflow data?
Create or select a form in the Project Explorer, then configure its data binding so fields exchange values with the workflow. Configure validation for fields that require it. See forms, form data binding, and field validation.
How do I manage Agent versions and sharing?
Use the Agent management page to review versions and configure sharing permissions. Grant the permissions needed for viewing, editing, or execution. See Agent version management and sharing.
Can an external application trigger an Agent?
Yes. Configure a webhook for the Agent or use the REST endpoint to create an automation instance. Configure the required inputs and authentication for the chosen method. See webhooks and create process instance API.
Can an Agent run on a schedule or when an email arrives?
Yes. Configure the appropriate trigger and its criteria for the Agent. See time triggers and email triggers.
Human tasks
Where do I find and complete my assigned tasks?
Open Inbox → Tasks, select the task, and choose Open Task. Complete the form and use its action button. Available operations depend on task status and permissions. See My Tasks.
How do tasks assigned to a group work?
A group member can take assignment of a new task and open it. An assigned task can be released to the pool so another member can take it, subject to the available operations. See assigning and releasing tasks.
Can someone handle my tasks while I am away?
Yes. Create a time-bound delegation to another user and choose the task scope, applicable processes, and date range. Managing delegations for others requires the appropriate permissions. See delegations.
What happens if I cancel a human task?
The My Tasks guide states that cancellation advances the workflow to the next activity with the current task's form data left blank. Review downstream handling before cancelling. See task cancellation.
Monitoring and maintaining execution
Can I suspend, resume, or cancel a running Agent automation?
The Automation page exposes these actions according to the automation state and your permissions. Inspect the instance details and choose the appropriate action. See Agent automation operations.
Where do I find failed activities and execution data?
Open the Agent automation's details to review its status, activity history, logs, and data. During design, use the Output and Error drawers for feedback. See automation details, Output drawer, and Error drawer.
How do I configure workflow error handling?
Configure an activity's On Error policy or add a Custom Error Handler where supported. Inside a custom handler, the Process Data Explorer exposes the failing activity's error details for recovery and logging. Test the failure path as well as successful execution. See On Error policies, custom error handlers, and error data.
Can an existing automation continue on a newer Agent version?
Yes, Automation Migration moves suspended instances from one Agent version to another. Configure source and target versions, review activity mapping, and validate the migration before resuming execution. This is separate from exporting definitions between environments. See automation migration.
For chat, voice, and enterprise knowledge, see Conversational Agents and Knowledge FAQs.