Setting your AI receptionist's tone of voice, with examples

An AI receptionist's tone of voice is set with a short paragraph: how formal to be, how long replies should run, whether to use first names, and the words to avoid. Write it from how your best front desk person actually texts, test it against ten real customer messages, and change the paragraph rather than editing individual replies. Four or five sentences are enough.

What a tone paragraph covers

1. Formality. First names or titles. Contractions or not.
2. Length. One to three sentences is right for most texts.
3. Warmth. Friendly and brief, or calm and precise.
4. Words to avoid. "Absolutely", "awesome", "no worries", exclamation marks, emoji, whatever does not sound like you.
5. How to say no. The phrasing for "I cannot answer that, a person will".

Four examples

Dental practice: "Warm, calm and brief. Use the patient's first name. One to three sentences. No exclamation marks or emoji. When you cannot answer, say a team member will reply this morning."

Plumbing company: "Direct and practical. Short sentences. Ask for the ZIP code early. Never sound alarmed; if something is urgent, say a person is being alerted. No jokes."

Med spa: "Friendly and polished, like a good concierge. First names. Never discuss whether a treatment is right for someone; a provider will. One emoji at most, only if the client used one."

Law firm: "Professional and careful. Use full sentences and no contractions. Never characterise anyone's legal position. Always say that an attorney will discuss specifics."

Write it from real messages

Open the last twenty texts your front desk sent. Notice the length, the greetings, the sign-offs, the phrases they repeat. That is your tone. Customers already know it, and a reply that suddenly sounds like a brochure feels wrong to them.

Test it

Take ten real messages from last month (a price question, a reschedule, a complaint, a vague "hi") and run them through the assistant. Read the replies out loud. If a reply sounds wrong, change the paragraph, not the reply, because the paragraph shapes every future answer.

What tone cannot fix

A wrong fact. A missing answer. A question that should have gone to a person. Tone makes correct answers sound like you; it does not make answers correct.

Frequently asked questions

Should the AI sound like a person?

It should sound like your business, and say it is an assistant. Sounding human is not the goal; sounding like you is.

Should tone differ by channel?

A little. Web chat can be slightly more conversational than SMS. The same paragraph usually works for both.

How often should I revisit it?

After the first week, then whenever your team keeps editing the same kind of phrase.

What Frontiva does here

Each of Frontiva's AI agents takes a tone, goals and escalation keywords in plain language, and the playground lets you try real messages against your live configuration before a customer sees a reply. Your industry template starts you with a suggested tone. See approval mode for testing it with real customers safely.