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

# Sequences

> The step-by-step outreach flow your Outreach Agent runs.

A **sequence** is a chain of steps that runs for every prospect enrolled in an Outreach Agent. Send an invite → wait N days → send a DM → wait → send an email — and so on.

## The step types

<CardGroup cols={2}>
  <Card title="Send" icon="paper-plane">
    Send a message on a specific channel: LinkedIn invite, LinkedIn DM, email, WhatsApp message, LinkedIn comment, LinkedIn reaction, LinkedIn profile view.
  </Card>

  <Card title="Wait" icon="clock">
    Wait N hours or days before the next step.
  </Card>

  <Card title="Wait for event" icon="hourglass">
    Wait for a specific event (they replied, they accepted the invite, they opened the email) before continuing.
  </Card>

  <Card title="Branch" icon="code-branch">
    Split the flow based on a rule ("if they replied → escalate; if not → try email").
  </Card>

  <Card title="AI reply" icon="wand-magic-sparkles">
    When a reply comes in, Guhan drafts a response for you to approve — then sends automatically or waits for your OK.
  </Card>

  <Card title="Hand off" icon="user">
    Pause the sequence and send you a task in the Inbox. Useful for high-stakes moments.
  </Card>
</CardGroup>

## Two ways to build a sequence

### Simple (stair view)

A vertical list of steps. Great for **linear** sequences with no branching.

```
Step 1  →  LinkedIn invite
Step 2  →  Wait 3 days
Step 3  →  LinkedIn DM
Step 4  →  Wait 4 days
Step 5  →  Email
Step 6  →  Wait 7 days
Step 7  →  Follow-up email
```

Read top to bottom. Click any step to edit it.

### Advanced (canvas)

A visual node graph. Necessary if you want to **branch** based on prospect behavior.

```
                  ┌── replied ──→  AI reply → Meeting request?
                  │                    ├── yes → Book call
LinkedIn invite ──┤                    └── no  → Hand off to me
                  └── didn't reply ─→ Wait 4d → LinkedIn DM
                                          ...
```

Drag nodes onto the canvas, wire them together, click to configure.

You can toggle between views for the same sequence — the underlying data is the same.

## Send steps: what actually gets sent

Every **send step** has:

* **Channel** — LinkedIn invite, DM, email, WhatsApp, comment, reaction, profile view
* **Body / brief** — you write a short **brief** describing what to say, not the literal words. Guhan writes the actual message per prospect.
* **Subject line** (email only) — same treatment; you write intent, Guhan personalizes
* **Sender account** — which of your connected accounts sends this step (defaults to the agent's default)
* **Require approval** (optional) — hold the send until you click Approve in the Inbox

<Note>
  **Brief, not body.** In older systems you'd write `Hi {{firstName}}, saw that {{companyName}} raised...` with token-replacement. Guhan doesn't do that — you write "Congratulate on the funding, offer help with SDR ramp" and Guhan writes a fresh message per prospect referencing their actual funding round.
</Note>

## Wait steps

* **Fixed wait** — "wait 3 days" then go to the next step regardless
* **Wait for event** — "wait until they reply OR 7 days pass, whichever first"

Common pattern: `Send invite → wait for invite acceptance (up to 14 days) → send DM if accepted`. Guhan handles the "if they didn't accept, don't try to DM" automatically.

## Branch steps

Split the flow based on rules. Example:

* If prospect has an email → try email next
* If not → try WhatsApp next

Rules can check: title, company, headline, ICP score, whether they have an email/phone, whether they're on Do Not Contact, whether they came from a specific List, days since added, and more.

## AI reply steps

When a reply comes in during a sequence, Guhan can:

1. **Classify** the reply — is it a "yes", a "not now", a "not interested", a meeting request?
2. **Draft** an appropriate response — brief, on-brand, references what they said
3. Either **auto-send** it OR **hand off** to you for approval

You configure how many rounds Guhan will handle before pulling you in (default: 1 exchange).

## Editing a live sequence

When you edit a sequence on a live Outreach Agent, Guhan publishes a **new version**. What happens:

* **New prospects** enrolled after the edit get the new version
* **Prospects already in-flight** keep their old version (so they don't get double-messages or skipped steps)
* **You can pause** the agent, edit, publish, and the pause resumes on the new version

## Related

<CardGroup cols={2}>
  <Card title="Build a sequence" icon="pen-to-square" href="/agents/outreach-agents/build-sequence">
    Step-by-step canvas walkthrough.
  </Card>

  <Card title="AI reply drafts" icon="wand-magic-sparkles" href="/inbox/ai-reply-drafts">
    How Guhan drafts replies for you.
  </Card>
</CardGroup>