What Should an AI Receptionist Ask a New Enquiry?

Many businesses assume the goal of an AI receptionist is to ask as many qualification questions as possible. That is usually the wrong approach.

A caller who simply wants to speak with a business should not have to complete a ten-question verbal form. The AI should first understand why the person is calling, then ask only enough questions to help the business take the next step.

A good AI receptionist does not collect the most information possible. It collects the most useful information necessary for the next action.

This guide covers what an AI receptionist should ask a new enquiry, which questions to avoid, and how to design a simple call flow that supports your team and CRM. It builds on our explanation of how an AI receptionist works.

First Question: How Can I Help?

The first meaningful step in a useful AI receptionist conversation should normally establish intent — why the person is calling — before collecting any personal details.

Broad call types a business might receive include:

• new enquiry
• existing customer
• appointment
• quote
• support
• billing or admin
• a particular staff member
• something else

The AI should not immediately begin with "What's your full name?", "What's your email?", "What's your address?" or "What's your budget?" without first understanding why the caller picked up the phone. Establishing intent first makes the rest of the conversation feel natural rather than like a form being read aloud.

The Core Questions Most Businesses Actually Need

There is no universal script. However, a simple service-business enquiry often needs some combination of the following.

1. What Can We Help You With?

Understand the enquiry itself. This is the foundation every other question builds on.

2. What's Your Name?

Useful for identifying the caller and creating a contact record.

3. What's the Best Number to Reach You On?

Where necessary. If caller ID already provides the number, the workflow may confirm it rather than make the customer repeat the entire number unnecessarily.

4. Where Are You Located?

Only where location matters — for example a suburb, postcode or property location. This is useful for service-area businesses. Do not automatically request a full residential address if only the suburb is necessary at this stage.

5. When Would You Like Us to Contact or Help You?

Where relevant: callback timing, an appointment preference, or service timing.

6. Is There Anything Else We Should Know?

Use this carefully. Do not invite unnecessarily sensitive information. The AI can allow the caller to add useful context without turning the conversation into a huge questionnaire.

Every Question Should Have a Reason

Before adding a question to the AI receptionist, apply a simple test:

What will we do differently because we know the answer?

If the answer is "nothing", consider removing the question.

• If suburb determines whether the business services the customer — ask suburb.
• If email is never used during initial follow-up — do not automatically ask for email.
• If budget does not change the next step — do not ask budget merely because it sounds like lead qualification.
• If a full address is not required until a booking — collect it later.

Do not turn a phone call into a database-completion exercise.

Collect Only What You Actually Need

Collecting more personal information is not automatically better. Current Australian privacy guidance under Australian Privacy Principle (APP) 3 emphasises proportionality and a data-minimisation approach where the Australian Privacy Principles apply. An AI receptionist should therefore be designed to avoid unnecessary collection.

• Instead of a full residential address, ask which suburb you are in — where suburb is enough for service-area qualification.
• Instead of date of birth, do not ask unless the business has a genuine reason and an appropriate process.
• Instead of collecting extensive personal information 'in case it is useful later', only capture information required for the current business purpose.

This is general guidance, not legal advice. For current requirements, refer to the OAIC's guidance on the Australian Privacy Principles.

Does Every Caller Need to Give an Email Address?

No. Ask for email where it supports an actual next step.

• sending requested information
• appointment communication
• a quote process
• documentation
• agreed follow-up

But if the workflow is simply "phone enquiry → owner calls customer back", then name, phone number and the enquiry itself may be sufficient. Do not add friction simply to fill every CRM field.

Should an AI Receptionist Ask About Budget?

Sometimes. Not always.

Budget can be appropriate when it genuinely changes:

• service suitability
• project scope
• which team receives the enquiry
• the next step

But asking budget too early can feel unnecessary or intrusive. A plumber receiving a blocked-drain enquiry probably does not need to begin with "What's your budget?". A business handling larger project enquiries may legitimately need approximate budget information later in the conversation.

Ask it because the answer matters — not because the CRM has a budget field.

Qualification Should Help Routing, Not Create an Interrogation

Qualification can be useful. But "qualification" does not have to mean a long sales script.

Simple qualification may involve:

• service required
• service area
• appointment type
• new or existing customer
• broad project type
• timing

This may be enough to determine the right department, a callback, an appointment, an unsuitable enquiry, or a human escalation. Do not encourage unnecessarily aggressive qualification.

Ask Questions in a Natural Order

A good conversational pattern looks like this:

AI: How can I help today? → Caller explains enquiry → AI: Absolutely, may I get your name? → Relevant industry question → Location if required → Callback or appointment details → Confirm next step

Compare that with a bad pattern:

The AI immediately asks for name, email, phone, address, budget, business size, preferred service and timeline — before understanding why the caller called. The second feels like a form being read aloud because that is exactly what it is.

Don't Ask Five Questions in One Sentence

Avoid:

"Can I have your name, phone number, email address, suburb and preferred appointment time?"

Prefer:

"Can I start with your name?"

Then continue naturally. One question at a time helps caller comprehension, speech recognition, information accuracy and conversational flow.

Confirm Details That Matter

Some information should be confirmed before the call ends. Examples include a phone number, an email where used, an appointment time, a property address where required, and the spelling of unusual names where necessary.

Do not repeatedly confirm every sentence. Confirmation should reduce genuine errors, not make the conversation frustrating.

Phone-Number Handling

If telephony provides caller ID, the receptionist may be able to use or confirm that number. For example: "Is the number you're calling from the best number for us to contact you on?" This may be more natural than requiring the customer to repeat the entire number. Do not assume caller ID is always reliable or available.

Each Useful Answer Should Have Somewhere to Go

The point of capturing information is that it helps the business after the call. Each useful answer should land somewhere structured:

• Caller name → CRM contact
• Phone → contact field
• Suburb → service-area field or note where appropriate
• Enquiry type → pipeline, tag or opportunity context
• Preferred timing → note, task or appointment request
• Call summary → CRM note

This is where the conversation connects to CRM and lead management. The goal is not simply to have an impressive AI conversation — the information needs to help the business respond.

A Simple Connected Workflow

Customer calls → AI understands enquiry → Relevant questions asked → Information captured → CRM updated → Team notified → Human follows up or booking proceeds

This kind of straightforward flow is covered by simple business automation. It does not require a complicated autonomous AI system. Where appropriate, acknowledgements and reminders can also be handled through SMS and email follow-up.

The AI Doesn't Have to Answer First

A business can potentially let its normal phone ring first. The questions only begin because the team could not answer:

Customer calls → Team gets first opportunity to answer → No answer after configured delay → AI receptionist answers → Questions begin

Three rings may be used as an illustrative example, but the actual forwarding delay depends on the telephony provider and configuration. For a fuller comparison of this setup against voicemail, see our guide on AI receptionist vs voicemail.

The Right Questions Depend on the Business

Different industries receive different enquiry types, so the call flow should be configured around the business — not copied from a generic template.

Plumber

• What can we help with?
• Which suburb is the property in?
• Is this a repair, service or quote enquiry?
• What are you experiencing?
• What's the best number for a callback?

Do not allow the AI to diagnose plumbing, gas or safety problems. See how this fits plumbing businesses.

Electrician

• What are you calling about?
• Which suburb or property is involved?
• Is this an existing job or a new enquiry?
• What's the best callback number?
• What has the caller observed?

Do not ask questions designed to diagnose the electrical fault. Do not instruct the caller to access switchboards, test wiring or perform repairs. See electricians.

HVAC / Air Conditioning

• Is this installation, servicing or repair?
• Residential or commercial?
• Suburb?
• System type if the caller knows it?
• What has the caller noticed?

Do not diagnose equipment. See air conditioning and HVAC businesses.

Automotive Workshop

• Service or repair enquiry?
• Vehicle make and model?
• Registration if the workshop genuinely needs it at this stage?
• What has the customer noticed?
• Preferred booking timing?

Do not determine whether a vehicle is safe to drive. See automotive workshops.

Real Estate Agency

First determine whether the caller is buying, selling, seeking an appraisal, renting, an existing tenant, or a landlord. Then ask only the relevant questions.

Example seller enquiry:

• name
• phone
• suburb or property
• appraisal request
• preferred callback

Do not automatically ask for financial information. See real estate agencies.

Dental Clinic

Keep this example especially conservative. Routine administrative questions may include:

• new or existing patient
• appointment request
• preferred timing
• contact details

Do not use AI reception to diagnose symptoms, recommend treatment, collect extensive medical history unnecessarily, or make clinical decisions. Clinical questions require human handoff. See dental clinics.

One Script Shouldn't Handle Every Caller

A single rigid script cannot sensibly handle every caller. Branching around intent keeps the conversation relevant:

How can I help? → New enquiry (ask relevant new-enquiry questions) / Existing customer (identify relationship and reason) / Appointment (handle appointment workflow) / Particular staff member (route or capture callback) / Complex or sensitive issue (human handoff)

This is why AI reception should be configured around intent rather than one huge script.

Questions an AI Receptionist Should Avoid Unless They're Actually Needed

• unnecessary date of birth
• extensive personal information
• sensitive information unrelated to the enquiry
• full home address when suburb is enough
• income
• financial details
• budget for every enquiry
• marketing questions before resolving the reason for calling
• medical information unrelated to an administrative dental request

These are not forbidden categories. Only request information where there is a genuine business need and an appropriate process for handling it.

Don't Trap the Caller in the Script

An AI receptionist should not endlessly repeat a question because the caller declines to answer. Depending on the business process, it may:

• skip non-essential questions
• explain why information is useful
• collect a callback request
• transfer to a person
• end the intake appropriately

Not every CRM field needs to be completed.

How Many Questions Is Too Many?

There is no universal numerical maximum. The right number depends on the enquiry. A simple callback enquiry may only require a few details. A more structured booking or quote may legitimately require more.

Measure the flow by one question:

Does every question help complete the next step?

If not, simplify it. Do not arbitrarily claim "five questions is optimal" unless genuine research supports it.

Sometimes the Most Useful Output Is the Summary

After the call, the business may receive something like the following (a fictional example, not a real customer):

Structured capture like this is far more useful to a busy team than a scattered voicemail.

These principles underpin the AI receptionist service — the conversation is designed around the business's actual next step, not around collecting data for its own sake.