Conversation Nodes

Each step of the conversation is its own node — the greeting, each question, each qualification step, the objection handling, the close. Nodes connect to each other, and they can branch based on what the caller actually says.

This is where most of your time goes, and where the workflow earns its keep: you can move a question earlier, add a branch, or rewrite one step without touching the rest of the agent.

What a node holds

• The prompt for that step — what the agent is trying to accomplish
• Interruption handling — how the node behaves if the caller talks over the agent
• A toggle for whether the node also draws on the shared business context

Why nodes instead of a single script

A real conversation is not linear. Callers answer out of order, ask their own questions, or need to be routed differently based on what they say. Modelling each step as its own node means:

• You can reorder steps without rewriting the whole prompt
• You can branch a node based on the caller's actual response — see Connections & Routing
• You can test one node in isolation when something about that specific step is wrong

Next

• Connections & Routing
• Extraction Variables
• End Call