In Superchat, you can create your AI Agent and prompt it here:
The prompting done here defines the general, global behavior of the AI Agent across Superchat. This is where you describe the Agent’s role, tone, goals, and overall rules, such as what topics it can handle, how it should speak to customers, and when it should normally escalate to a human. These instructions apply everywhere the AI Agent is used.
Additional Instructions
When you then implement your AI Agent via the AI Agent node into an automation, you have the option to add Additional Instructions.
Additional instructions only apply to this specific AI Agent node, meaning instructions given here will not be applied to other AI Agent nodes, even if using the same AI Agent. Additional instructions in the AI Agent node are local and context-specific.
This is useful when you want the AI Agent to perform specific tasks or share certain information only at this point in the automation and not in general.
Example
Say for instance you have an automation that first asks the customer for their order number and issue type.
At the AI Agent node that comes after these questions, you add a prompt like:
“You are an AI Agent handling support requests. At this point in the automation, check whether the customer has provided both an order number and an issue type. If both are present, escalate the conversation to a human agent. If any information is missing, ask only for the missing detail.”
This prompt applies only to this AI Agent node, so the escalation rule is enforced only after the automation has collected the required information, not in earlier or later parts of the flow.