Writing voice: make AI drafts sound like you

Set how the AI assistant writes emails and texts for your business in Settings → CRM Settings → How we write, including your name, role, sign-off, tone, example messages and emoji use, what it may say about your business (where you work, hours, licences, guarantees), how the tone check works, and what happens to a message with a formatting problem.

Updated

Your writing voice is how the AI assistant sounds when it drafts emails and texts for your customers and leads. Set it once and every draft follows it: your name and sign-off, how casual or formal you are, and the way you actually talk. Writing voice settings apply to the AI features on Starter and Pro.

Where do I set my writing voice?

Go to Settings → CRM Settings and find the How we write card ("The voice the assistant uses for every email, text, reply and follow-up it drafts for your customers and leads."). Only owners and admins can change it, and it applies to your whole business.

SettingWhat it does
Signed by (first name)The name drafts are signed with. If you leave it blank, your first name from your account is used.
Role (optional)Your role, such as "owner", added to the signature.
Sign-offHow drafts end, such as "Thanks,".
ToneA slider from Casual to Professional, in five steps. The default is casual, like a busy local owner emailing a peer.
How we talk (optional)A sentence or two in your own words, for example "Short and straight. We say folks, never premium. No sales talk."
Emails or texts you like (optional, up to 2)Paste an email or text you wrote that sounds like you.
We use emojisTurn on if emojis fit your style; they're then used sparingly.

The card shows a preview of your signature. Click Save, or Reset to defaults to start over. Resetting the voice doesn't touch the business facts below it.

What can the assistant say about my business?

Only what's in your records or what you've told it. The assistant never makes up a service area, how long you've been in business, licences, insurance or guarantees. On the same How we write card, under What the assistant can say about us, you tell it what's true:

SettingWhat it does
Where we workThe cities, towns or areas you take jobs in, separated by commas, for example "Calgary, Airdrie, Cochrane and Okotoks". This is what lets the assistant answer "do you cover X?".
Places we don't cover (optional)Places you don't go. The assistant never says you work there, and tells people who ask that you don't.
Other facts the assistant can state (optional)Hours, years in business, licences and insurance, guarantees, anything else it may tell customers. Write it the way it is: "licensed, not bonded" means it can say licensed and won't say bonded.

This section has its own Save button. After you save, the card shows how your service area was read ("The assistant can say you work in: …" and "It won't say you cover: …"), so you can check it.

A few things worth knowing:

  • A place counts when it's one you listed. "Greater Victoria" also covers "Victoria", but "North Vancouver" isn't "Vancouver".
  • A place you listed as not covered is never claimed, even if you've done a job there before.
  • Your mailing address and the cities of your past jobs still count, as they did before.
  • If something is pending or planned ("insurance pending", "applying for a licence"), the assistant doesn't state it yet.

You can also just tell the assistant, for example "we also cover Nanaimo" or "we're licensed and insured". It shows you a card with the change, and nothing is saved until you confirm.

Why did the assistant remove a sentence from a draft?

When a draft says something about your business that isn't in your records or your facts, that sentence is taken out before you see the card, and the card tells you what was removed and where to add it, for example:

Removed an unsupported claim from the Email: “We also serve Kelowna.” — location/service area “Kelowna” isn't in your business details. If you work there, add it under Settings → CRM Settings → How we write → “Where we work” (or tell me and I'll add it).

Add the fact there (or tell the assistant), then ask for the draft again.

What does the writing voice apply to?

Anything the AI drafts for your customers and leads: email campaigns and outreach, text campaigns, one-off emails and texts, replies to your business inbox, follow-up steps, the personal lines in outreach emails, and drafts from background tasks.

It doesn't change the AI text receptionist, your automations, missed-call text-back, or your notification templates, which have their own settings.

What style rules do drafts follow?

On top of your settings, every draft follows a shared style guide so it reads like a real person wrote it:

  • Outreach emails are short (roughly 50 to 90 words, and follow-ups shorter), ask one clear question, and use a short subject line.
  • No exclamation marks in outreach emails.
  • Overused sales and filler phrases are avoided, such as "reaching out", "spotless", "free, no-obligation" and "warm regards".

What is the tone check?

Every draft gets an automatic tone check before you see it. The card shows the result: "Tone check: ✓" when it's clean, or "Tone check: N notes" when something might sound off. If a draft clearly misses the mark, the assistant rewrites it once before showing you the card.

What if a message has a formatting problem?

Campaign emails and texts, one-off emails and texts, Inbox replies and follow-up steps are checked just before they're queued or sent, so a broken message doesn't reach a customer. That means things like "[object Object]", a placeholder nobody filled in (such as {first_name} or {{company}}), or a greeting with no name ("Hi ,"). Automations, the AI text receptionist, missed-call text-back and notification templates aren't part of this check.

  • When it's safe, the message is fixed on its own: a line that is only the leftover is dropped, a placeholder with a fallback uses the fallback, and "Hi ," becomes "Hi,".
  • When the problem is in the middle of a sentence, the message is held instead of sent, with the reason, for example "Held: the message had a formatting problem — unfilled placeholder {{company}}."

Where you see it: a one-off email, reply or text isn't sent and the card (or the Inbox or text screen) shows the reason. In an email campaign, that person's row in Deliveries reads Held with the reason while everyone else's email goes out; fix the wording or that person's details and queue the campaign again, and the held email is sent. In a text campaign, the result says how many texts were held and why. The tone check also points these out on the card before you confirm.

Can I ask for a rewrite?

Yes. While the card is waiting for your OK, just say what to change: "make it shorter", "less corny", "more professional" or "sound more like me". You get new wording for the same recipients and schedule, and the old card is marked Replaced.