Bot Flow Concepts
Before building a flow, it helps to understand the core building blocks:- Trigger — What starts the bot. A trigger can be a specific keyword the customer sends (e.g. “hi”, “pricing”, “support”), any new incoming conversation, or a message received outside of business hours.
- Step — A single action in the flow. Each step does one thing: send a message, ask a question, evaluate a condition and branch, call a webhook, or collect a variable.
- Variable — Data collected during the conversation and stored for later use — for example, the customer’s name, email address, or order number. Variables can personalize responses and be saved back to the contact record.
- Handoff — Transfers the active conversation from the bot to a live agent in the Team Inbox, along with all conversation context so the agent can pick up without asking the customer to repeat themselves.
Creating a Bot Flow
1
Open the Chatbots module
Go to Chatbots → New Flow.
2
Name and assign the flow
Give the flow a descriptive name and select the WhatsApp channel it applies to. A flow only intercepts conversations on its assigned channel.
3
Set the trigger
Choose Keyword and enter one or more trigger words (e.g. “hi”, “pricing”, “support”), or select Any New Conversation to capture every new incoming chat.
4
Build the flow on the visual canvas
Add steps by dragging and dropping blocks onto the canvas:
- Send Message — send a text, image, video, or document
- Ask Question — prompt the customer for input and save their response to a variable
- Branch — split the flow based on a condition (e.g. variable value, contact tag)
- Set Variable — manually assign a value to a variable
- Assign to Agent — trigger a handoff to a specific agent or team
- Webhook — call an external URL with conversation data
5
Connect the blocks
Draw arrows between blocks to define the conversation path. Branch blocks create multiple outgoing paths based on conditions.
6
Test the flow
Click Test Flow to simulate the conversation in a preview panel. Walk through each path to verify the logic before going live.
7
Activate the flow
Toggle the flow to Active. It will immediately begin intercepting conversations that match the trigger conditions.
AI-Assisted Responses
Flows can include an AI step that generates a contextual, natural-language reply using your choice of AI model — ChatGPT, Claude, or Gemini. The AI step can be pointed at a knowledge base (a set of documents or FAQs you upload) or driven by a custom prompt you define. This lets the bot handle open-ended questions — product details, policy questions, troubleshooting — that are too varied to script manually. When the AI step reaches its confidence threshold or the customer’s query falls outside its scope, the flow automatically hands off to a human agent with the full conversation history intact.Keyword Triggers
Keyword triggers are case-insensitive and match partial words by default — so a trigger of “price” will fire if the customer types “pricing”, “Price list”, or “what’s the price?”. Use Exact Match mode if you need the trigger to fire only on a precise word or phrase. Multiple keywords can be assigned to the same flow, so you don’t need a separate flow for every variation of a word.