Conversation flow
Ordered sequence of automated messages that a chatbot executes after a trigger, with output to human and clear action at the end.
Conversation flow is the ordered sequence of automated messages that a chatbot runs after a trigger. A good flow is short, gives output to a human at any point, and ends with a clear action (tag applied, schedule set, link sent).
The anatomy of a flow that works
- Trigger — keyword, time, first message from the contact, or external event (webhook).
- Initial message — identifies that it is automation ("I'm the virtual assistant"), sets expectations ("I'll direct you in 1 minute"), offers immediate value.
- Branches — 2-3 paths based on the answer. Never more than 4 options: customer loses patience.
- Maximum depth of 3 levels — flow with 5+ levels turns into a maze.
- Output for human on every node — always "0 - speak to attendant". Without this, the customer hits his head and disappears.
- Closing action — tag applied, link sent, card moved in Kanban, attendant notified.
Errors that kill a stream
- Bot that pretends to be human (customer finds out and loses trust).
- Messages are too long (not read).
- Dead-end loops for human.
- Branches that depend on the client typing exactly one word (no typo tolerance).
- Flow that does not end with clear action (client waits for something to happen).
Most common flow types on WhatsApp
| Scenario | Typical structure |
|---|---|
| Welcome | Greeting → 3-option menu → routing |
| Lead Qualification | 3 short questions → automatic tag → handoff |
| After hours | Time notice → demand capture → return promise |
| FAQ | Keyword detection → ready answer → "did it help?" → end or human |
| Scheduling | Service question → shows schedules → confirms → creates entry via webhook |
See: complete chatbot guide and automation.