Platform

An AI voice agent you build, test, and put on a number

Lay the conversation out node by node, hear every branch in the browser, then point a phone number at it. The agent that answers is the one you tested.

Incoming call

What this is for

Voice AI usually arrives as one of two things: a demo you cannot change, or an SDK you have to build a telephony stack around. The interesting work sits in the gap. Drawing a conversation that branches sensibly, hearing it before a customer does, and putting it on a real number without becoming responsible for media servers.

How you build one

  1. 01

    Draw the conversation

    Nodes and branches in a visual editor. A node is something the agent says or does. A branch is a way the conversation can go. No code.

  2. 02

    Give it what it needs to know

    Your documents for the answers, your calendar for the slots, your systems for the lookups. It works from those rather than from whatever a model happens to believe.

  3. 03

    Hear it before anyone else does

    Test the whole thing in the browser, on any branch, as often as you like. What you approve is what picks up the phone.

  4. 04

    Point a number at it

    Connect a phone number to the agent in the dashboard. Woise hosts the agent, so there is nothing for you to run.

Related

AI voice agent: common questions

What is an AI voice agent?

Software that holds a spoken phone conversation. It listens, works out what the caller means, does something about it, and speaks back. That something can be a lookup, a booking or a transfer. Unlike an IVR, it does not need the caller to pick from a menu.

Do I need to write code?

No. Flows are built in a drag-and-drop editor and tested in the browser, and Woise hosts the agent so there is no infrastructure to stand up. You connect a phone number to the agent in the dashboard.

Can it do things, or only talk?

It acts. Reading a live calendar, looking up an order, writing a booking, transferring a call: those are nodes in the flow rather than things it describes doing.

How do I know what it said on a call?

Recording, full transcript and an outcome for every call. Nothing depends on the agent reporting its own behaviour honestly.

Is it only voice?

Voice is the larger part of it, but the same agents run in chat. Flow, knowledge and records are shared across both, which is the groundwork for handling more than one channel with one agent.

Still stuck? Talk to our team.