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

# Replies, suppression and auto-send

> How replies are read and classified, what each category does, the do-not-contact list, and the auto-send setting that lets drafted replies go out on a timer.

## How replies are read

Replies arrive through the connected mailbox and are matched to the enrollment by thread. Each reply
is classified into one category, with a confidence score:

| Category                | What the engine does                                                                                           |
| ----------------------- | -------------------------------------------------------------------------------------------------------------- |
| **Interested**          | Fires the *positive reply* outcome. By default the enrollment finishes and everyone else at the company stops. |
| **Meeting request**     | Same as interested.                                                                                            |
| **Not interested**      | Fires the *not interested* outcome.                                                                            |
| **Out of office**       | The enrollment is paused and the next step is rescheduled past the return date when one is found.              |
| **Wrong person**        | The enrollment finishes.                                                                                       |
| **Unsubscribe request** | The enrollment is finished and the email is added to suppression.                                              |
| **Auto-reply**          | Ignored for routing.                                                                                           |
| **Unknown**             | The classifier could not tell. The sequence's reply action applies.                                            |

The sequence's **primary reply action** (finish, continue, or pause for N days) applies to a reply
from the prospect; the **secondary reply action** applies to a reply from anyone else on the
thread. Reactions in [Rules of engagement](/sequencer/rules-of-engagement) can override these per
outcome.

### Custom reply labels

Under **Sequencer › Settings** a team can add its own reply labels ("Asked for pricing", "Referred
me to someone"). The built-in categories cannot be edited, because the engine acts on them; custom
labels only trigger reactions you define, through the *a reply matched one of your labels* outcome.

## Suppression: the do-not-contact list

**Sequencer › Suppression** holds emails, domains and phone numbers that will never be enrolled.
The enrollment preview refuses them with the reason "suppressed".

| Type   | Matching                                                                                                                           |
| ------ | ---------------------------------------------------------------------------------------------------------------------------------- |
| Email  | Exact address, case-insensitive.                                                                                                   |
| Domain | Any address at the domain. You can paste `example.com`, `@example.com`, a full address or a URL; all are normalised to the domain. |
| Phone  | The number with or without the US country code.                                                                                    |

Entries carry a reason: **unsubscribed**, **bounced**, **complained**, **manual**, or **do not
call**. Unsubscribes, bounces and complaints are added automatically; a *do not call* disposition
in the dialer also adds the phone. Bulk import a list of addresses or domains with **Import**.

<Warning>
  Removing an entry that was added by an unsubscribe or a complaint re-allows outreach to someone
  who asked not to be contacted. Do it only when you are certain it was added in error.
</Warning>

## Auto-send for replies

By default every reply a rep or Ari drafts is a task: someone reads it and sends it. **Auto-send**
lets a team release drafted replies on a timer for specific reply categories, for example an
automatic thank-you and scheduling link to an *interested* reply if nobody has sent one within two
hours.

Settings under **Sequencer › Settings › Auto-send**:

| Setting                | Notes                                                                                                                                                                 |
| ---------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Enabled**            | Off for every team until an admin turns it on. Enabling with no categories selected is refused.                                                                       |
| **Categories**         | Which reply categories may auto-send. **Unsubscribe request, out of office, auto-reply, wrong person and unknown can never be selected.**                             |
| **Delay**              | How long a drafted reply waits for a human before it releases. Delays shorter than the sweep interval are refused rather than rounded, and the error names the floor. |
| **Minimum confidence** | Replies classified below this confidence are never auto-sent.                                                                                                         |
| **Daily cap**          | Maximum auto-sends per day for the team.                                                                                                                              |
| **Respect schedule**   | Release only inside the sending window.                                                                                                                               |

Every auto-send is logged on the enrollment as a send with the note that it was released
automatically, and the drafted reply stays visible in the task list until it goes.

<Note>
  Unknown is banned on purpose. If the classifier is unavailable, every reply is categorised as
  unknown with zero confidence; banning it means an outage results in nothing being sent rather
  than everything.
</Note>