Flows overview

A flow is your agent's game plan for a conversation. When a customer says hello, the flow decides what happens next: what to say, what to ask, when to look something up, when to think with AI, and when to call in a human.

You build flows on a visual canvas. You drag in actions, connect them with lines, and that's your conversation. No code, no deploys — when you save, it's live.

Why flows exist

You could hand every conversation straight to the AI and let it figure things out. For many businesses, that's a perfectly good start (and we'll show you how).

But real businesses have rules. Always greet first-time visitors by name. Route refund requests to the billing team. Ask for an order number before looking anything up. Never let the AI promise a discount. Flows are how you encode those rules — the structure around the intelligence.

The sweet spot is a blend: let the AI be smart where it should be, and keep firm rails where it matters. Flows give you both.

The anatomy of a flow

Every flow is made of a few pieces:

• A Start — the proactive opener. It runs once, before the customer has said anything — the moment a conversation begins. Use it to greet someone the instant they open the chat. (If the customer's very first action is to send a message, there's nothing to "open," so Start is skipped.)
• A Fallback — the main handler for what the customer says. When a customer sends a message and the flow isn't already waiting for a specific answer, it lands here. This is where your answering logic lives.
• Actions — the steps. Each does one job: send a message, ask the AI, capture an email, search your knowledge, branch on a condition, hand off to a human.
• Connections — the lines between actions that decide what comes next.

The two-word rule worth memorizing: Start = before they type, Fallback = when they type. This trips up most newcomers, so it's the first thing to get right.

Here's the simplest useful flow you can build:

Fallback  →  Agentic AI

That's it. Connect the Agentic AI action to the Fallback — now every customer message is handled by an autonomous agent that reads the message, searches your collections, remembers details, looks up orders, and hands off when needed. A single block on the Fallback can run an entire support agent.

The action toolbox

Actions are grouped by what they're for. You don't need to learn all of them — most flows use a handful. Here's the lay of the land:

• Control — the skeleton of the flow: Start (the proactive opener), Fallback (the handler for customer messages), idle timers, and the resolving/resolved states.
• Messaging & capture — talking and listening: Send Message, Capture User (name/email/phone), Assign User.
• AI & data — the smart stuff: Ask AI, Agentic AI, User Intent, Extract Data, Search Documents, Reply From Documents.
• Integration — reaching outside Vivollo: API Call, Update Conversation, and the help-desk (Connexease) actions.

Every action — what it does, when to use it, and what each setting means — is in the Actions reference.

How a flow runs

There are two ways a flow gets going:

• The conversation opens. If a conversation begins before the customer has typed anything, Vivollo runs the Start — your chance to greet them first.
• The customer sends a message. This is the common case. Their message enters at the Fallback, unless the flow was already waiting for a specific answer — for example, right after you asked a question with a Send Message or a Capture User, their reply continues that step instead. Tap a button and the flow follows wherever that button points.

From its entry point, Vivollo walks forward action by action, following your connections. Some actions just do their job and move on (like sending a message). Others pause and wait for the customer, then continue based on what they said.

A flow ends when it reaches a Resolved state, hands off to a human, or simply runs out of steps. If the customer messages again later, their message comes back in through the Fallback.

Reusing work with variables

As a conversation unfolds, it collects information: the customer's email, an extracted order number, a detected intent, the answer from an API call. Vivollo stores these as variables, and you can drop them into later steps using double braces:

Thanks {{customer_name}}! Your order {{order_id}} is on its way.

Variables are what make a flow feel personal instead of robotic. You'll see them throughout the Actions reference.

Where to go next

Now that you've got the map, let's build something real:

1. Build your first flow — a guided, start-to-finish flow you can copy.
2. Conditions & branching — send different customers down different paths.
3. Reusable flows & timing — chain flows together and handle silence gracefully.
4. Rich messages — buttons, carousels, and forms that make conversations feel like an app.