Flows: your best answers, turned into a conversation anyone can run

If the knowledge base is what I know, flows are how I walk a customer through it — one question at a time, in order, without skipping the step that matters.

What a flow is

A flow is a guided conversation: questions, choices, outcomes. "Is the breaker tripped? If yes, do this. If no, check the float switch." You build it once, on a canvas, and from then on the answer to that problem is the same on a Monday morning and at 3am — because it isn't being recalled from memory, it's being executed.

The publishing model is the part that makes it safe to edit. Flows are versioned: once you publish a version, it's immutable. Every running session is pinned to the exact version it started on. So you can improve the flow on Tuesday without yanking the version a customer is standing in mid-conversation. Edits never break sessions in flight — that's a promise the database enforces, not a convention we hope everyone remembers.

Where flows run

The chat interface is the primary way to run one — a conversational runner you can open straight from the Flows list. But flows meet you where the work is: when an agent opens a ticket, the system reads the ticket's subject and body and suggests the flows most likely to fit, ranked — with your AI provider configured, I do the ranking; without one, a keyword-matching fallback does the same job in a dumber but honest way. And inside the live-chat console, a flow runs right beside the conversation, so an agent can walk a customer through it without switching screens.

Flows are also the stage where the offers module performs: when a customer's answers match an offer's targeting conditions, the offer appears at that step — which is why your cross-sells land at the exact moment they're relevant.

The shared brain

Flow decisions use the same condition system as offer targeting and ticket assignment rules — one rule format across the workspace. You learn it once. And every session records its steps and outcomes as analytics, so you can see which flows run, where people drop off, and which outcomes dominate — the troubleshooting path that actually resolves things becomes visible instead of anecdotal.

What I do here

With your AI provider connected, I can draft a flow from a description — and anything I draft has to pass the same structural linter the canvas uses before it's allowed to exist. I don't get to skip the rules I ask everyone else to follow.

Sessions are instrumented as they run — each step, each outcome, every offer shown — so the analytics tab shows where customers drop off and which outcomes actually resolve things. And flows don't only answer customers: an appointment type can kick one off automatically when someone books, running your intake questions before the visit instead of during it.

Everything stays versioned end to end — the canvas edit, the published version, and the session's pinned copy are three different things the database keeps distinct on purpose.

Related: flows are where offers appear at the right moment, and they run inline in live chat next to the conversation.

How this compares

When people say "automation" they usually mean trigger-action tools — Zapier made the category famous, and it's genuinely good at moving data between apps on events. That's a different job. An automation fires and disappears; a guided flow is a conversation someone walks through, step by step, with decisions that depend on their answers. Documentation wiki pages are the other common substitute — a checklist a person is expected to follow from memory.

This module is built for the interactive job. Flows are versioned charts — publish an edit and in-flight sessions keep running the version they started on, so "we changed the doc" never breaks a customer halfway through. They run chat-first, they run from tickets, they can even kick off from a booking. Suggestions on new tickets come from the same library, so the flow you wrote once quietly pre-answers future tickets. And the decisions inside a flow use the same condition DSL as cross-sell offers and routing rules — one rule language across the workspace. General-purpose automation is for plumbing; this is for the moment a person needs to be walked somewhere.

Try it

Flows live under Flows in your sidebar. Start from the demo flows if you want to see the shape, then write one for the question your team answers most. Open the chat runner, run it as if you were the customer, and check the analytics tab afterwards — you'll see your own session counted.