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

# Voices: Email Sender Identities for Recovery Messages

> A voice is an email sender identity and tone the strategy model uses to write recovery emails. Only email has voices; SMS sends fixed templates.

A **voice** is a sender identity — a name, a role, and a tone — that Airdun uses when writing recovery **emails**. You define a small set of voices; the strategy model picks the right one for each email in a sequence and writes in that voice.

<Note>
  **Voices are email-only.** SMS is voiceless: it sends fixed, pre-approved templates with no generated copy, so there is nothing to voice. The in-app wall has its own templates too. You will only find voices on the Email channel.
</Note>

## What a voice contains

| Field           | Required | Notes                                                                                                                              |
| --------------- | -------- | ---------------------------------------------------------------------------------------------------------------------------------- |
| Role            | Yes      | Who this is, e.g. "Customer support", "Founder", "Billing team". Free text (1–80 chars). Suggestion chips are offered in the form. |
| Display name    | Yes      | The name the customer sees in the "From" line (1–120 chars).                                                                       |
| Job title       | No       | Adds a signature line when set.                                                                                                    |
| Tone of voice   | Yes      | A free-text brief (1–2000 chars) telling the model how this voice writes — formal, warm, direct, and so on.                        |
| Sending address | No       | An email address to send from. Must be on a verified sending domain. Leave empty to use the workspace default.                     |
| Primary         | —        | Exactly one voice per channel is primary.                                                                                          |

There is no avatar or photo field. A voice carries identity and tone, not artwork.

## How a voice is chosen

You do not map voices to steps by hand. When the strategy model designs an email sequence, it sees the full menu of your email voices — each one's role, display name, job title, and tone — and **picks a voice per email**, guided by that voice's tone and role. A warmer founder voice might open a late, softer ask; a billing voice might carry a firmer final notice. You steer the outcome three ways:

* **Which voices exist** — the model can only choose from the voices you have created.
* **Each voice's tone text** — this is your main lever; it is what the model reads to decide fit and how to write.
* **Which voice is primary** — the default when nothing else fits.

The chosen voice's identity and tone are snapshotted onto the step at generation time, so a later edit to a voice does not rewrite an email that has already been planned.

## Voices vs. sending domains

These are two different things, and it is worth keeping them straight:

* A **sending domain** is your workspace's authenticated email identity — the domain, with its SPF and DKIM records verified. This is set up once, separately, and is what makes email deliverable.
* A **voice** is a per-message persona. Its optional sending address must sit on an already-verified domain; if you leave it empty, the workspace default from-address is used.

So SPF/DKIM belong to the domain, not the voice. An email voice still requires your workspace to have a verified sending domain — without one, email generation cannot run.

## Managing voices

In your Airdun dashboard, go to **Auto-Pilot → Touchpoints → Email** ([app.airdun.com/auto-pilot/touchpoints/email](https://app.airdun.com/auto-pilot/touchpoints/email)). From there you can create, edit, and delete voices, and set which one is primary.

The create/edit form has: role (with suggestion chips), display name, job title (optional), sending address (optional), tone of voice, and a primary toggle. Editing and deleting are restricted to admins and owners.

<Note>
  You cannot un-set the current primary directly — promote another voice to primary and the previous one steps down. Deleting a voice cancels any pending steps still assigned to it, and promotes the next voice to primary if you deleted the primary one.
</Note>

<Tip>
  Give the model room to vary the sequence: create at least two email voices with clearly different tones (for example a warm relationship voice and a firmer billing voice). With only one voice, every email in every sequence speaks the same way.
</Tip>