Journey Builder

A journey is the visual flow that defines how a conversational agent talks with your customers — connected steps laid out on a canvas. You open it from the agent's Journey tab and build it yourself: what the assistant says, what it asks, and where the conversation branches.

Optimizer agents run on a journey too, but theirs comes pre-built from the kit and runs on a schedule — you rarely touch it. See Create an optimizer agent for how that one works.

A typical flow

A conversation usually opens with a greeting, offers quick replies to steer the visitor, then branches into answering a question, collecting details, showing products, or handing off to a person.

Reuse a journey

You don't always start from scratch. When you create an agent, a Starter Kit can give you a working journey to begin from. And once you've built a flow you like, Save to Library keeps it as a reusable template you can apply to your other agents.

The node palette

You build the flow by dragging nodes (steps) from the Node Palette onto the canvas and wiring them together. The palette groups them the same way you'll see in the builder:

For a conversational agent you'll mostly use Conversation, Intelligence, Data, Actions, and Logic & Flow — the Analyst group and scheduled Triggers are for optimizer agents. A Product Card pulls from a data table, and AI steps answer from your knowledge.

You connect steps by dragging a line from one to the next. Most lead to a single next step, but some branch: a Condition goes down a yes or no path; an AI Categorizer sorts the visitor's message into one of your categories and follows that category's path; and a Loop repeats a path for each item, then continues once it's done.

Variables

A journey remembers and reuses values as it runs. Drop a {{variable}} into almost any field — a message, a prompt, a condition, an API call, or a Product Card filter — and it's replaced with the live value (you can use several in one string). Variables come from:

• a visitor's answer to a Question step;
• a Data Collector step, where the AI pulls structured details out of the conversation;
• a Set Variable step — a fixed value, the last message, an AI-extracted value, or an expression;
• an API step's response (including nested fields);
• the signed-in visitor's details after a Login step;
• metadata your widget passes in — the page URL, referrer, and UTM tags.

Variables are available across the whole conversation. To pass them into a sub-flow, map them in and out — or turn on Share Variables so the sub-flow reads and writes the parent's variables directly.

Conversation behaviour

A few flow-wide behaviours keep chats smooth:

• Quick replies — after an AI response, if no Quick Reply step follows, the AI automatically suggests a couple of contextual follow-up buttons (on by default). An explicit Quick Reply step always takes priority — so wire those in only where you want specific, controlled options, and let the AI handle the open-ended points.
• Start over — a visitor can say "start over" or "reset" to return to the beginning; earlier messages stay visible.
• New conversation — after a few messages, a button lets a visitor begin a fresh session (the previous chat is saved to your Conversations).

Handling failures

AI and action steps — an API call, a webhook, an email — can fail. Each one lets you choose what happens then:

• Stop — end the flow immediately.
• Continue — skip the failed step and move to the next one.
• Error branch — route down a dedicated recovery path you build.
• Fallback message — show the visitor a set message and carry on.

You can also enable Retry once, which retries the step a single time before applying your chosen strategy — handy for transient network or API hiccups.

Sub-flows

Sub-flows let you compose several journeys into one agent, keeping each flow focused and reusable:

• Create separate flows for distinct modules — for example Product Recommendations, Support Triage, or Lead Capture.
• Use a Sub-Flow node to route from one flow into another.
• Configure input / output variable mappings to pass data between the parent flow and the sub-flow.
• Enable Share Variables to let the sub-flow read and write the parent's variables directly, instead of mapping each one.

They're especially useful for keeping complex agents manageable — the Health check suggests splitting a flow into sub-flows once it grows large.

Working on the canvas

The journey's name sits at the top-left — rename it here (this is the journey's name, separate from the agent's). Beside it, a Setup indicator (for example, 4/5) tracks how many of the journey's required items are configured — like attaching a knowledge base or setting contacts — and a small usage estimate hints at how heavy each run is.

The toolbar also gives you:

• History — the journey's version history.
• AI Assist — describe a change in plain words and it proposes edits you can apply to the canvas (unavailable for a few advanced node types).
• Save to Library — save this journey as a reusable template (see Reuse a journey).
• Save — save your changes.

A live Health check (below) flags problems as you build.

The Health check

As you build, Health validates the flow continuously and flags issues both on the affected step and in a summary panel, each with a plain-language fix hint. There are three levels:

• Errors — must be fixed before you publish (they'd break the flow).
• Warnings — recommended fixes that improve reliability.
• Tips — optional enhancements (for example, "add a Data Collector step to capture leads").

It catches things like:

• a step that can't be reached from the start, or a flow with no clear ending;
• a Quick Reply or AI Categorizer branch whose route leads nowhere (or a categorizer with no low-confidence fallback);
• a Product Card with no catalog, a Handoff with no contact, or a Send Email step with no recipient;
• a step that expects a knowledge base when none is attached to the agent;
• a step that calls an outside service with no error handling, or a flow large enough to be worth splitting into sub-flows.

Related

• Conversational agents — set up the agent this journey powers
• Knowledge & Data — the content and data a journey draws on
• Deploying Your Agent