> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kaisupport.com/llms.txt
> Use this file to discover all available pages before exploring further.

# What Kai is

> Kai reads the conversations your team already handled, and turns them into a support agent that drafts or replies.

Kai turns your team's past chats into a support agent. You upload a chat
export. Kai reads what your team already said to customers, and compiles it
into a **Brain**. Your team reviews the Brain. After that, Kai writes drafts in
your inbox, or answers customers on its own.

You do not write a knowledge base. You already wrote one, one conversation at a
time.

<Frame caption="The Overview page: what Kai did today, and what still waits for a person.">
  <img src="https://mintcdn.com/kai-support/RaOwCZDRMIUnRjWu/images/en/dashboard.png?fit=max&auto=format&n=RaOwCZDRMIUnRjWu&q=85&s=1f71201d5bd16e58095ededde51b1697" alt="The Kai Overview page" width="2880" height="2360" data-path="images/en/dashboard.png" />
</Frame>

## What a Brain is

A Brain is everything Kai is allowed to know and say. It has six parts, and you
can read and change each one:

| Part                    | What it holds                                                               |
| ----------------------- | --------------------------------------------------------------------------- |
| **Answer Library**      | The answers to the questions your customers ask most often.                 |
| **Conversation Flows**  | The steps Kai follows to guide a conversation to an outcome.                |
| **Human Handoff**       | The situations where Kai must stop and give the conversation to a teammate. |
| **Communication Rules** | What Kai must never say, and what it must always explain.                   |
| **Brand Voice**         | The wording and the tone that keep every reply consistent.                  |
| **Current Information** | Live prices, links and payment details used inside answers.                 |

Kai writes the first version of each part from your history. Nothing in it can
reach a customer on its own until a person approves it.

A Brain is versioned: `v1`, `v2`, and so on. A published version is frozen. To
change a published version, you open a copy with **Edit as a new draft**, and
the copy becomes `v1.1`, `v1.2`, and so on. The published version keeps serving
customers until you publish the new draft.

## The two modes

<CardGroup cols={2}>
  <Card title="Copilot" icon="pen-line">
    Kai writes a draft and a coaching note in your inbox. A human reads it,
    changes it if necessary, and sends it. Nothing reaches a customer unread.
  </Card>

  <Card title="Autopilot" icon="bolt">
    Kai sends the reply itself. It does this only for content a reviewer
    approved, or only for the intents you pick. Everything else still drafts or
    escalates.
  </Card>
</CardGroup>

There is a third setting, **Off**. Kai watches and records, but never acts. It
is useful while your team reviews the Brain.

Even on autopilot, Kai writes a draft instead of sending when one of these is
true:

* a human handoff rule fires;
* the reply quotes a number that Kai cannot verify;
* confidence is below the floor you set;
* a human already replied on that conversation.

Read more in [Copilot, autopilot and off](/product/modes).

## Where Kai works

Kai works in the inbox your team already uses:

* **Intercom.** Drafts arrive as internal notes. A handoff is an assignment.
  One Intercom workspace can carry the Messenger widget, WhatsApp, email and
  SMS, and each of them can have its own mode.
* **WhatsApp Cloud.** A WhatsApp Business number talks to Kai directly.
* **Kai's own inbox.** WhatsApp Cloud has no provider inbox, so your team
  answers from Kai. The inbox holds the conversation list, the thread, contact
  details, assignment and delivery status.

## Your first hour

<Steps>
  <Step title="Sign in and name your workspace">
    About two minutes.
  </Step>

  <Step title="Upload your chat history">
    A WhatsApp or an Intercom export is enough. Kai replaces every phone
    number, email address and payment reference with a placeholder before any
    model reads a word.
  </Step>

  <Step title="Wait for the compile">
    A few hundred conversations take a few minutes. You can close the tab.
  </Step>

  <Step title="Work the review queue">
    Approve, edit or reject what Kai proposes. Work from the top: the queue
    puts what blocks publishing first, then the most frequent items.
  </Step>

  <Step title="Connect a channel and stay in copilot">
    Your team reads every word Kai writes. The first drafts appear in the inbox
    within minutes of the first customer message.
  </Step>
</Steps>

The full walkthrough is in [Get started](/product/getting-started).

## Next

<CardGroup cols={2}>
  <Card title="How Kai works" icon="diagram-project" href="/product/how-it-works">
    The pipeline, one live message end to end, and what Brain versions mean.
  </Card>

  <Card title="Get started" icon="rocket" href="/product/getting-started">
    From sign-up to the first draft, step by step.
  </Card>

  <Card title="Kai for support agents" icon="headset" href="/product/agent">
    A seven-page training path for the people who work the inbox every day.
  </Card>

  <Card title="Copilot, autopilot and off" icon="sliders" href="/product/modes">
    What each mode does, and a safe order to turn them on.
  </Card>

  <Card title="Contacts API" icon="code" href="/contacts-api/overview">
    Push, update, assign and export the contacts of your workspace.
  </Card>
</CardGroup>