Automations: Journeys & AI Agents

Build conversations that run themselves. This guide explains the two ways to automate your customer chats — Journeys and AI Agents — what each one is best at, and how to decide which to use. No technical background needed.

In this section

PageWhat it isBest forJourneyA guided, step-by-step conversation you designMenus, forms, and predictable flows where you want full control of every replyAI AgentsAn AI assistant that understands and answers on its ownAnswering questions in the customer's own words, using your knowledge base


What is an automation?

An automation is a conversation that happens without an agent typing every message. It can greet customers, answer common questions, show menus, collect information, take orders, and pass the chat to a human when needed — instantly, at any hour, across WhatsApp, Instagram, and Web Chat.

You build automations visually: you add steps on a canvas and connect them together to decide what happens next. You have two kinds of automation to choose from.

Journey — a conversation you script

A Journey is a guided path you lay out yourself. You decide exactly what the customer sees at each step: a welcome message, a set of buttons, a menu to choose from, a question to answer. Based on what the customer taps or types, they move along the path you designed.

Think of it like a choose-your-own-adventure: every choice and every reply is one you planned in advance. Journeys are predictable, easy to follow, and great when the conversation has a clear structure.

Great for:

  • Welcome menus ("Sales / Support / Track my order")

  • Order forms and lead capture (collecting name, address, preferences)

  • Step-by-step guides and FAQs with fixed answers

  • Booking, registration, and simple self-service

  • Anything where you want to control the exact wording of every reply

AI Agent — an assistant that thinks for itself

An AI Agent uses artificial intelligence to understand what the customer means and reply in their own words — even when they don't follow a menu. You give it instructions and connect it to your knowledge base, and it answers questions naturally, pulling from the information you've provided.

Think of it like a trained assistant: instead of following a fixed script, it reads the customer's message, figures out what they need, and responds — while still able to hand off to a human whenever you want.

Great for:

  • Answering questions asked in free text ("do you deliver to Bandung on Sundays?")

  • Support that draws on a knowledge base or FAQ documents

  • Understanding intent when customers don't use your exact keywords

  • Conversations that are too varied to script step by step


Which one should I use?

Use this quick guide:

If you want to…ChooseShow buttons and menus with fixed choicesJourneyControl the exact reply at every stepJourneyCollect information with a form-like flowJourneySend timed follow-ups ("Still there?")JourneyAnswer questions in the customer's own wordsAI AgentPull answers from your knowledge baseAI AgentUnderstand what a customer needs without a menuAI Agent

Simple rule: if you can sketch it as a flowchart with clear choices, build a Journey. If it needs to understand and answer free-form questions, build an AI Agent.

You can use both together

Journeys and AI Agents aren't either/or. An AI Agent can understand a customer's request and then hand them into a Journey — for example, chat naturally about what they need, then drop into a step-by-step checkout. This lets you combine the flexibility of AI with the control of a script in one experience.


How building works (the basics)

Both are built in the same visual way, so once you learn one, the other feels familiar.

  • Steps (nodes). Each step does one thing — send a message, show buttons, ask a question, check a condition, hand off to an agent. You add steps from a menu and drag them onto the canvas.

  • Connections. You draw lines between steps to decide what happens next. A message step simply continues to the next step; a buttons step branches depending on which button the customer taps.

  • The starting step. Every automation has one step where it begins.

  • Saved details (variables). Information collected during the chat — like a name or an order number — is remembered and can be reused in later messages (for example, greeting the customer by name).

  • Waiting for a reply. Whenever the automation asks a question or shows buttons, it pauses and waits. When the customer responds, it picks up right where it left off.

Channels

Every automation is tied to a channel — WhatsApp, Instagram, or Web Chat. The available steps adapt to the channel (for example, WhatsApp offers product catalog steps and interactive lists; Instagram offers story and comment features). Build a separate automation per channel you want to serve.

Turning it on

  • Journeys start when a customer types a trigger keyword you set (like "menu"), when you make one your auto-reply, from an Instagram comment, or when an agent starts it manually.

  • AI Agents run as your channel's AI auto-responder, replying to incoming messages automatically.

Testing, publishing, and versions

  • Test your automation in a built-in preview chat before customers ever see it.

  • Publish when you're happy — this makes your latest design live.

  • Versions & rollback. Each time you publish, a snapshot is saved. If a change doesn't work out, you can roll back to a previous version.

  • Import / export. You can export an automation to a file and import it into another workspace — handy for reusing a design or moving between environments. (After importing a Journey, re-check its trigger keywords and auto-reply setting, as these are cleared on import for safety.)


A quick example

A support experience combining both:

  1. A customer messages your WhatsApp asking, "my order hasn't arrived, what do I do?"

  2. An AI Agent reads the message, recognizes it's about a delivery problem, and answers using your shipping-policy knowledge base.

  3. It offers a button: "Talk to a person."

  4. If tapped, it hands off into a Journey that collects the order number and creates a support ticket, then routes the chat to an available agent.

The customer gets an instant, natural answer and a smooth path to a human — with no one typing until it truly matters.


Where to go next