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

# Create and launch an outbound sequence

> Build a sequence with the Companion: add people, write and personalize steps, set waits, senders and business hours, launch, and learn what happens when people reply.

You build a sequence by describing it to the Companion: who to reach, what to send and in what order. The Companion drafts it, and you review and adjust each part on the sequence's tabs before you launch.

> [!NOTE]
>
> **Outbound → Sequences → New sequence** in your [OpenCX dashboard](https://platform.open.cx/outbound/sequences). Requires **Outbound sequences** write, and **Companion** write to start the chat. Read the [introduction](https://docs.open.cx/outbound/introduction.md) first for what each channel needs.

## Start the sequence

**1. Click New sequence**

The Companion opens and asks: *Who should we reach, and what should we say to them?*

**2. Describe it**

Say who the people are, what you want from them, and which channels to use, for example "email everyone who asked about the annual plan last month, then call the ones who don't reply after three days".

**3. Review the draft**

The sequence opens next to the chat on **Overview**, as a **Draft**. The stage bar marks the next tab that still needs something before you can launch: **People**, **Steps** or **Settings**.

Keep asking the Companion for changes at any point. You can also edit steps, waits and settings directly on the tabs.

## Add people

People appear on the **People** tab as the Companion enrolls them. Ask it to find them for you:

* **From your own data**, such as a contact segment, contacts who match a filter, or people who were in touch over a period.
* **From what customers said**, such as everyone who complained about late deliveries. The Companion looks through past sessions for the topic.
* **From a spreadsheet** you attach to the chat. Its columns are kept with each person, so steps can use them.

People can also be enrolled from a workflow with **Add Contact to Sequence**, or from the [API](https://docs.open.cx/api-reference/sequences/add-contacts.md). Adding someone who isn't a contact yet creates the contact, which needs **Contacts** write.

Some people are not enrolled: anyone who already left this sequence, anyone whose addresses are all on the outbound blocklist or on channels their contact restricted, anyone without a valid phone number, email or WhatsApp ID, and, with the default settings, people already in another sequence or messaged by a past one.

### Qualify the list

The plan on **Overview** can list **Must have** criteria, which everyone has to clear, and **Nice to have** criteria, which only rank people. With criteria, the Companion can look up what OpenCX knows about each person and score them; the People tab shows the progress, such as *Judging 12 of 35*.

The **Fit** column shows how much of the plan each person matches, out of 100. People who clear every must-have send; people short of one are held back. Click a score to see the checklist behind it. To send to someone who was held back, choose **Include anyway** from their row menu. People who were never scored always send.

## Write the steps

The **Steps** tab shows the steps in order, with the wait before each one.

| Step | What you can change here |
| - | - |
| **Email** | **Edit template**: the subject line and body. Type `{{` to place one of the step's variables, then **Save for everyone**. `{{signature}}` is filled from the sending mailbox. |
| **WhatsApp** | Nothing: the message is an approved template and can't be edited. Ask the Companion to change the template or what fills its variables. A badge shows a template that is waiting on approval, paused or no longer usable. |
| **Phone** | **Add a brief** or **Edit brief**: what the AI agent should accomplish on the call. Without one, the agent's own instructions carry the whole call. Under **Who calls**, switch to **Your team** to have your agents call instead; see [Call from a sequence](https://docs.open.cx/phone/human-calling.md#call-from-a-sequence). |

### Variables and fallbacks

A variable is filled for each person, from their name, phone or email, or from data that came with them, such as a spreadsheet column. Hover a variable to see what fills it. A variable can have a fallback, used when a person has no value. Without a fallback, that person is held back at the step with the status **Needs data** instead of receiving a message with a gap.

### Personalize for one person

Open **Personalize for your contacts** on the Steps tab and pick a person to see every step as they will get it. Click **Personalize** on a step to rewrite it for them only. **Back to the template** undoes it. A person held back on **Needs data** goes back in the queue once you save a personalized step for them.

### Waits

Click the wait between two steps to change it, in minutes, hours or days, up to 90 days. The first step has no wait: it is due as soon as the sequence launches, or as soon as someone is added to a sequence that is already sending. Each wait counts from when the previous step went out, or was skipped or failed, and is then moved into business hours if needed.

## Choose the settings

| Section | What it controls | Default |
| - | - | - |
| **Sends from** | The **WhatsApp number**, the **AI agent** for calls, and the **Mailboxes** for email. Pick as many mailboxes as you like: each sends a few emails a day, so more mailboxes get through the list faster. A change applies from the next send. | None |
| **Replies** | What the AI should do when someone replies on WhatsApp or email: the goal, the tone, when to hand off. Calls follow the phone agent's own instructions. | Empty |
| **Business hours** | **Only send during business hours**, and **Hours to use**: **Default office hours** or a schedule from **Settings → Office Hours**. A message due outside them waits for the next opening, in the person's own timezone when it's known. | On, default office hours |
| **Not messaging the same person twice** | **Skip people already in another sequence** and **Skip anyone a past sequence already messaged**. Both count every sequence in the workspace, and both are checked when people are enrolled. | Both on |

Keep both skip settings on for cold outreach. Turn them off only on purpose, for example for a re-engagement campaign or an announcement to existing customers.

## Launch

Click **Launch** at the top right. A toast confirms how many people started sending.

**Launch** stays greyed out until the sequence has at least one step, something to send from on every channel it uses, and no rejected or disabled WhatsApp template; hover it to see what is missing. A template still waiting on approval doesn't block the launch: sends start once it is approved.

If the sequence calls numbers in countries with known calling rules, a **Quick check before launching** dialog lists them under **Country rules**. Tick **I have read this and consent**, then **Launch**. The summary is a reminder to check those rules, not legal advice.

After launch, people added to the sequence start right away. Use **Pause** to stop all sending and **Resume** to continue; everyone keeps their place.

## When someone replies

1. The reply opens a session in the [inbox](https://platform.open.cx/inbox) with the message the person is answering, and a **Reply to campaign** banner naming the sequence and step. Click it to open the sequence. For an answered AI call, the banner reads **Call from sequence**.
2. The person leaves the sequence with the status **Replied**, and no further steps go to them, even if they had already finished.
3. When the AI answers, it treats the session as a continuation of the sequence and follows your **Replies** instructions.

On email, only a reply to the sequence's thread counts. A new email someone writes to the mailbox arrives as an ordinary session and doesn't stop their steps. An automatic reply, such as an out-of-office message, counts as a reply, and the AI doesn't answer it. A WhatsApp step waits while the person has an open WhatsApp session with your team, and tries again an hour later.

To act on replies automatically, use the **Sequence Replied** workflow trigger; **Continue Sequence for Contact** puts someone who replied back on their next step. See [Skip a sequence call after a reply](https://docs.open.cx/phone/follow-up-workflows.md#skip-a-sequence-call-after-a-reply).

## Troubleshooting

| Symptom | Likely cause | Fix |
| - | - | - |
| **Launch** is greyed out | No steps, no sender for a channel, or a rejected WhatsApp template | Hover **Launch**, then fix what it names on **Steps** or **Settings**. |
| No mailbox to pick under **Mailboxes** | No mailbox on a live sending domain yet | Add mailboxes under [Outbound → Senders](https://docs.open.cx/outbound/senders-and-domains.md). |
| People show **Needs data** | A variable has no value and no fallback for them | Add a fallback, give them the data, or personalize the step for them. |
| None of the above | | Ask the Companion in the sequence's chat, or contact OpenCX support. |

***

## Related Documentation

> [card] [Track results](https://docs.open.cx/outbound/track-results.md)
>
> Statuses, replies and exits.

> [card] [Senders and domains](https://docs.open.cx/outbound/senders-and-domains.md)
>
> Mailboxes for email steps.

> [card] [Outbound calls](https://docs.open.cx/phone/outbound-calls.md)
>
> Phone agents and number pools for call steps.

> [card] [Sequences API](https://docs.open.cx/api-reference/sequences/index.md)
>
> Enroll, author and launch from code.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.