AI Voice Receptionist Prompt: Examples and Templates (2026)

An AI voice receptionist is only as good as the prompt behind it, which is the set of instructions that tells the AI who it is, how to speak, and what to do on every call. This guide covers what a strong prompt includes and gives ready templates for the common call types, then shows where a ready made tool skips the prompt writing entirely.

What Makes a Good AI Voice Receptionist Prompt

A voice prompt is not a chatbot prompt. Calls are live, so the instructions must be short, ordered, and specific about what to say and do. A strong prompt covers four things.

• Role. Who the AI is: the business name, that it is the receptionist, and what the business does.
• Tone. How it speaks: warm, brief, and natural, with short sentences a caller can follow on the phone.
• Goals. What every call should achieve: book the appointment, capture the caller's details, and confirm a next step.
• Guardrails. What it must not do: no quotes it cannot stand behind, no medical or legal advice, and a clear rule for when to take a message or transfer.

The templates below are starting points. Fill in the business name, hours, services, and booking rules, then test them on real call types before going live.

Prompt Template: Greeting and Call Opening

Greeting prompt

You are the friendly virtual receptionist for [Business Name], a [type of business] in [city]. Answer every call within the first ring. Greet the caller warmly and briefly: “Thanks for calling [Business Name], this is [Agent Name], how can I help?” Speak in short, natural sentences. Let the caller explain, then move the call toward booking an appointment or taking a message. Never talk over the caller.

Prompt Template: Booking an Appointment

Booking prompt

Your main goal is to book the appointment. Ask for the service needed, the address, and the preferred day and time. Offer the next two open slots from the calendar: “I can do [slot one] or [slot two], which works better?” Confirm the details back to the caller, then book it. If a deposit is required, explain it and take the payment. End by confirming a text confirmation is on its way.

Prompt Template: Qualifying and Capturing the Lead

Qualifying prompt

Always capture the caller's full name, phone number, and the reason for the call before the call ends. For a new customer, ask one or two qualifying questions relevant to [Business Name], such as the type of job and the urgency. If the caller is not ready to book, capture the details and say a team member will follow up by [timeframe]. Never let a caller hang up without a name and a number on file.

Prompt Template: Answering FAQs and Pricing

FAQ prompt

Answer common questions using only the approved information for [Business Name]: hours, service area, services offered, and general pricing ranges. For an exact quote, say pricing depends on the job and offer to book an assessment. Do not invent prices, guarantees, or availability. If you do not know an answer, say so and offer to take a message for the team.

Prompt Template: Transfers and Messages

Transfer prompt

Transfer to a human when the caller asks for one, when the issue is an emergency, or when it falls outside your approved topics. Before transferring, say: “Let me connect you with someone who can help.” If no one is available, take a detailed message with the caller's name, number, and reason, and confirm when they can expect a callback. Always leave the caller with a clear next step.

The Full Prompt, Assembled

The templates above are modular so you can edit each call type on its own. In practice most platforms want one prompt, so here is the same structure assembled into a single block. Fill in the bracketed fields and trim anything the business does not offer.

You are [Agent Name], the virtual receptionist for [Business Name], a [type of business] serving [city and service area]. Your hours are [hours]. You answer every call warmly and briefly, speak in short natural sentences, and never talk over the caller. Open with: "Thanks for calling [Business Name], this is [Agent Name], how can I help?"

Your main goal on every call is to book the appointment and capture the caller's full name, phone number, and reason for calling before the call ends. To book, ask for the service needed and the preferred day and time, offer the next two open slots, confirm the details back, and take a [deposit amount] deposit when the job requires one. Confirm that a text confirmation is on its way.

Answer questions using only approved information: hours, service area, services, and general pricing ranges. Do not invent prices, guarantees, or availability. For an exact quote, offer to book an assessment. Transfer to a human when the caller asks, when it is an emergency, or when the topic is outside your approved list, saying "Let me connect you with someone who can help." If no one is available, take a message with the caller's name, number, and reason, and confirm a callback by [timeframe]. Never let a caller hang up without a clear next step.

Variables to Fill In Before You Go Live

A prompt only works once the placeholders are real. Set these before the first live call.

Common Mistakes That Break a Voice Prompt

Most weak voice agents fail for the same handful of reasons, and each one traces back to a line in the prompt.

• Too long. A prompt with dozens of rules makes the agent slow and hesitant on a live call. Keep instructions short and ordered by call stage.
• No capture rule. If the prompt does not force a name and number before the call ends, the agent will happily let warm leads hang up unrecorded.
• No pricing guardrail. Without an explicit "do not invent prices" rule, the agent will guess a number and put the business on the hook for it.
• No transfer path. If there is no rule for when to hand off, callers with an emergency or an edge case get stuck in a loop.
• Assuming it knows your details. The model does not know your hours, service area, or booking rules unless the prompt states them. Spell them out.
• A robotic tone. Long formal sentences read fine on screen but sound stilted out loud. Write the way a good receptionist actually talks.

How to Test and Improve the Prompt

A prompt is never right on the first write. Treat it as something you tune against real calls.

• Script the common call types. Write out the five or six calls the business gets most: a new booking, a reschedule, a price question, an after-hours emergency, a wrong number. Run each one.
• Listen to the recordings. Read the transcript or listen back, and watch for the three failure modes: a missed name or number, invented information, and a call that ended with no next step.
• Fix one instruction at a time. When a call goes wrong, change the single line responsible, not the whole prompt, then re-run the same call. Changing several rules at once makes it impossible to see what helped.
• Re-test before every change goes live. A tweak that fixes booking can quietly break the transfer rule. Run the full script again after each edit.

Or Skip the Prompt Writing Entirely

Writing, testing, and maintaining a voice prompt takes real work, and it only covers the talking. The booking, the customer record, the text follow up, and the payment still have to be wired up on a platform like Vapi, Retell, or Bland.

ServiceAgent takes the other route: it ships a working AI voice receptionist with no prompt to write. It answers every call, books the job into a native CRM, takes a deposit, and texts the customer back, and it sets itself up from a website and Google Business Profile. It reports a 100% call capture rate across more than 350,000 calls handled, a 56% average job booking rate, and more than 7,600 businesses using it. It starts free and bills by usage.

Sources

1. Voice AI platform documentation (Vapi, Retell, Bland) for prompt structure and call handling.
2. ServiceAgent reported results, for the call capture, booking, and adoption figures, checked September 2026.
3. Google US Search and AI Overview for the keyword, for the definition and People Also Ask.